Journal of Intelligent & Robotic Systems (2018) 91:35–58

https://doi.org/10.1007/s10846-018-0805-9

Mobile Robot Control and Navigation: A Global Overview

Spyros G. Tzafestas1

Received: 7 February 2018 / Accepted: 26 February 2018 / Published online: 10 March 2018
© Springer Science+Business Media B.V., part of Springer Nature 2018

Abstract
The aim of this paper is to provide a global overview of mobile robot control and navigation methodologies developed over
the last decades. Mobile robots have been a substantial contributor to the welfare of modern society over the years, including
the industrial, service, medical, and socialization sectors. The paper starts with a list of books on autonomous mobile robots
and an overview of survey papers that cover a wide range of decision, control and navigation areas. The organization of
the material follows the structure of the author’s recent book on mobile robot control. Thus, the following aspects of
wheeled mobile robots are considered: kinematic modeling, dynamic modeling, conventional control, affine model-based
control, invariant manifold-based control, model reference adaptive control, sliding-mode control, fuzzy and neural control,
vision-based control, path and motion planning, localization and mapping, and control and software architectures.

Keywords Mobile robot · Autonomous mobile robot · Control · Path planning · Motion planning · Navigation ·
Localization · Mapping · Control architecture · Software architecture

Mathematics Subject Classification (2010) 68T40 · 70E60 · 93C85 · 70Q05 · 70B15

1 Introduction cognitive, and behavior-based control, and attempts to max-

imize flexibility of performance subject to minimal input
Terrestial (Ground) robots are distinguished in fixed-place dictionary and minimal computational complexity.
robots and mobile robots. Fixed-place robots are robots of Therefore, AMRs belong to the broad class of intelligent
which the base is fixed at a specific place, and hence they robots. Ronald Arkin [7], defines an intelligent robot as ‘a
have a workspace limited by their kinematic structure and machine able to extract information from its environment
the size of their links. Unlike fixed-place robots, mobile and use knowledge about its work to move safety in a
robots are robots that can move from one place to another meaningful and purposive manner’. In general, a robot is
autonomously, i.e., they have the special feature of mov- referred to in the literature as a machine that performs an
ing around freely within a predefined workspace to perform ‘intelligent connection between perception and action’.
given tasks and achieve desired goals. Today’s autonomous The navigation and control field of AMRs has achieved
mobile robots (AMRs) can move around safely in cluttered over the years high maturity, both in theory and practice,
surroundings, understand natural speech, recognize real and a large number of authored and edited books have
objects, locate themselves, plan paths, navigate themselves, been published in the international scene. Twenty authored
and generally think by themselves. The design of AMRs books are listed in Table 1, in which the authors’ names,
employs the methodologies and technologies of intelligent, publication years, and gross contents of them are provided
[1–20]. Ten edited books written by invited or conference
authors are [21–30]. Also, several special issues of
 Spyros G. Tzafestas
international journals, and numerous state-of-art papers
tzafesta@cs.ntua.gr
exist in the literature which present surveys of particular
1 areas within the autonomous mobile robotics field [31–
School of Electrical and Computer Engineering, National
Technical University of Athens, Zographou 15773, 52]. A brief summary of some of these papers is given in
Athens, Greece Section 2.
36 J Intell Robot Syst (2018) 91:35–58

Table 1 Books on
Autonomous Robot Navigation Author Year Content
and Control (1991–2017)
A.M.Meystel [1] 1991 Intelligent motion control. Evolution of autonomous mobile
robots (AMR). Autonomous mobility. Cognitive control of AMR.
Nested hierarchical control. Intelligent Modules (planner,
navigator, pilot).Cartographer.
J.C. Latombe [2] 1991 Configuration space (CS) of rigid object. Obstacles in CS. Road
map methods. Cell de composition (exact, approximate).
Potential field methods. Dealing with uncertainty. Movable objects.
J.L. Leonard [3] 1992 The navigation problem. Sonar sensor model. Model-based
localization. Map building. Simultaneous map building and
localization. Directed sensing strategies. Why use sonar?
J.L. Jones [4] 1995 TuteBot. Computational Hardware. Designing and prototyping.
Sensors. Mechanics. Motors. Robot programming. Robot
applications. Robot design principles. Unsolved problems.
H.R. Everett [5] 1995 Design considerations. Dead reckoning. Odometer sensors.
Doppler and inertial navigation. Typical mobility configurations.
Tactile and proximity sensing. Triangulation ranging. Active
triangulation. Range from focus. Return signal intensity. Guide
path following. Position location sensors. Ultrasonic and optical
position location systems.
J. Borenstein [6] 1996 Sensors for mobile robot positioning. Heading sensors. Active
beacons. Sensors for map-based positioning. Landmark
navigation. Active beacon navigation systems.
R.C. Arkin [7] 1998 Whence behavior? Animal behavior. Robot behavior. Behavior-based
architectures. Representation issues. Hybrid
deliberative/reactive architectures. Perceptual basis for
behavior-based control. Adaptive behavior. Social behavior.
Fringe robotics-Beyond behavior.
J. Canny [8] 1998 Robot motion planning problems. Motion constraints.
Elimination theory. The roadmap algorithm. Performance
improvements. Lower bounds for motion planning. Motion
planning with uncertainty.
X.Zhu [9] 2001 Outdoor mobile robots. Motion Control. Cooperative motion
control and architecture. Kinematic motion control.
U.Nehmzow [10] 2003 Robot hardware. Robot learning. Making sense of raw sensor.
Navigation. Novelty detection. Simulation. Modeling robot-
environment. Robot behavior analysis. Locomotion. Mobile
robot kinematics. Perception. Mobile robot localization.
Planning and Navigation.
R. Siegwart [11] 2005 Locomotion. Mobile robot kinematics. Perception. Mobile robot
localization. Planning and Navigation.
F. Cuesta [12] 2005 Fuzzy systems. Stability analysis. Bifurcations in simple
Takagi-Sugeno fuzzy systems. Intelligent control of mobile robots with
fuzzy perception. Stability of mobile robots with fuzzy reactive
navigation. Intelligent system for parallel parking of cars
and tractor-trailers.
K. Berns [13] 2009 Historical overview of autonomous land vehicles. Vehicle
kinematics. Sensor systems. Where am I? The localization
problem. Map building. Navigation strategies. Control
architectures. Software frameworks.
J Intell Robot Syst (2018) 91:35–58 37

Table 1 (continued)
Author Year Content

G. Dudek [14] 2010 Fundamental problems. Mobile robot hardware. Non-visual

sensors and algorithms. Visual sensors and algorithms.
Representing and reasoning about space. System control. Pose
maintenance and localization. Mapping and related tasks. Robot
collectives. Robots in practice. The future of mobile robots.
G. Cook [15] 2011 Mobile robot control. Robot attitude. Robot navigation.
Application of Kalman filtering. Remote sensing. Target tracking
including multiple targets with multiple sensors. Obstacle
mapping and its application to robot navigation. Operating a
robotic manipulator. Remote sensing via UAVs. Robot parts. Kinematics.
C.A. Berry [16] 2012 Introduction. Hardware. Control. Feedback control.
Representation. Control architectures. Software. Navigation.
Localization. Mapping. Simultaneous localization and mapping.
R. Tiwari [17] 2012 Graph based path planning. Common planning techniques.
Evolutionary robotics. Behavioral path planning. Hybrid
graph-based methods. Hybrid behavioral methods. Multi-robot-systems.
A. Kelley [18] 2013 Introduction. Math fundamentals. Numerical methods.
Dynamics. Optimal estimation. State estimation. Sensors for
state estimation. Control. Perception. Localization and mapping.
Motion planning.
S.G. Tzafestas [19] 2014 Mobile robots. Mobile robot kinematics. Mobile robot
dynamics. Mobile robot sensors. Mobile robot control.
Lyapunov-based method. Affine systems and invariant manifold
methods. Adaptive and robust methods. Fuzzy and neural
methods. Vision-based methods. Mobile manipulation modeling
and control. Mobile robot path, motion, and task planning.
Mobile robot localization and mapping. Generic systemic and
software architectures. Experimental studies. Mobile robots at work.
L. Jaulin [20] 2017 Three-dimensional modeling. Feedback linearization.
Model-free control. Guidance. Instantaneous localization.
Identification. Kalman filter.

2 Overview of Twelve Survey Papers A short outline of them is as follows:

on Mobile Robots • Tai and Liu [31], provide a survey of deep learning (DL)
landscape in robotics which includes the development
The AMR areas covered by these surveys are:
of deep learning in related fields, especially the
• Deep learning in AMRs. essential distinctions between image processing and
• Multiple cooperative AMRs. robotic tasks. DL methods are distinguished in two
• Self-organized pattern formation in AMRs. classes: perception, and control systems. DL may be an
• AMR learning paradigms and applications. answer for the future of robotics and serve for solving
• Performance measures of mobile manipulators. robust and generic robotic tasks.
• Multi-robot coordination and decision problems. • Cao, Fukunago, and Kalug [32], review the research on
• Gathering fat robots and patrolling by mobile robots. systems composed of multiple autonomous mobile robots
• Control architectures of AMRs. exhibiting cooperative behavior. Groups of robots are con-
• Autonomous search and pursuit-evasion. structed aiming at studying group architecture, resource
• Control of mobile robots with trailers. conflict of cooperation, learning, and geometric problems.
• Simulation tools and their selection via user selection. This paper gives a critical survey of existing works
• Geometric registration and configuration selection. along with a discussion of open problems in this field.
38 J Intell Robot Syst (2018) 91:35–58

• Yamauchi [33], provides a survey of pattern formation • Chung, Hollinger and Isler [39], present a survey of
of autonomous robots focusing on self-organization of recent results in pursuit evasion and autonomous search
mobile robots, especially the power of forming patterns. relevant to mobile robotic applications. A taxonomy of
The existing results show that the robot system’s search problems is given that highlights the differences
formation power is determined by their asynchrony, resulting from several assumptions on the searches,
obliviousness, and visibility. Besides the existing targets, and the environment. Then, a list of a number of
results, this paper presents a number of open problems fundamental results in the areas of pursuit-evasion and
related to pattern formation in mobile robotics. probabilistic search is provided, including a discussion
• Cunha [34] provides an overview of the state-of-art of of the field of implementations of mobile robotic
the different applications of machine learning method- systems. Finally, several current open problems in the
ologies in mobile robotics. He starts by presenting area are highlighted to explore avenues for future work.
the credit assignment problem, namely: temporal credit • David and Manivanan [40], review a number of existing
assignment, structural credit assignment, and task credit studies in the control of mobile robots with trailers,
assignment. Then, an overview is given of various which accomplish their task in a faster and cheaper way
learning paradigms characterized by solving each credit than an individual robot. The main issue of their study
assignment problem in a different way. is the complexity and stability of the complete system
• Bostelman, Hong and Marvel [35], give a survey of which is nonlinear and unstable. This paper provides a
available research results concerning the performance survey of various control strategies and algorithms for
measurement of mobile manipulators. The survey the backward motion of mobile robots with trailers, and
provides a literature review of mobile manipulation identifies some unsolved problems in this area.
research with examples and experimental applications. • Ivaldi, Padvis and Nori [41], overview the panorama
Also the survey presents an extensive list of planning of simulation tools that are presently used in robotics.
and control references with factors into performance They propose to evaluate user feedback as a way to
measurement. make an objective and quantitative comparison, which
• Yan, Jouandeau and Cherif [36], deal with the is actually difficult to be done since many of the tools
analysis of multi-robot coordination, especially with are not open source. To this end, they created an on-line
the literature on problems related to communication survey about the use of dynamical simulation tools, and
mechanisms, planning strategies, and decision making. analyzed the participants’ answers to get a descriptive
Having made great progress in basic problems of single- information fiche for the most relevant tools. This can
robot control, much of the current research is focused to be helpful to robotics workers in choosing the best
the study of multi-robot coordination. The authors also simulation tool for each particular application.
review a series of related problems. • Pomerleau, Colas and Siegwart [42], review the
• Bandettini, Luporini and Viglietta [37], provide a geometric registration algorithms in robotics, which
survey of the open problems of (i) gathering fat (non- associate sets of data into a common coordinate system
point) robots which may be opaque in the sense and have been extensively used in object reconstruction,
that other robots cannot ‘see through it’, and (ii) the inspection, and localization of mobile robots. They,
problem of boundary patrolling by mobile robots with focus on mobile robotics applications in which point
constraints only on speed and visibility. For at most clouds are to be registered, presenting a formalization of
four robots an algorithm exists in the literature, but geometric registration and casting algorithms proposed
the question is whether gathering is always possible in the literature. They review some applications of this
for more than 4 robots. A set of mobile robots with framework in mobile robotics that cover different kinds,
constraints only on speed and visibility is working of platforms, environments, and tasks. The ultimate
in a polygonal environment having boundary and goal of the authors was to provide guidelines for the
possible obstacles. The robots must perform a perpetual selection of geometric registration configuration.
movement, so that the maximum time span, in which a
In the next Sections 3 through 9, an overview of the following
point of the boundary is not watched by any robot, is
mobile robot control and navigation topics will be provided:
minimized.
• Medeiros [38], presents a survey of the flavor in existing • Mobile robot kinematics and dynamics.
robot control architectures and identifies attributes of • Mobile robot control (standard control, state feedback
intelligent robot control architectures. He discusses the linearized control, invariant manifold-based control).
NASREM architecture, the subsumption architecture, • Mobile robot adaptive and robust control (MRAC,
the LAS architecture, and the TCA architecture. sliding mode control).
J Intell Robot Syst (2018) 91:35–58 39

• Mobile robot fuzzy and neural control.

• Mobile robot vision based control.
• Mobile robot path and motion planning.
• Mobile robot localization and mapping.
Section 10 will review the major intelligent control and
software architectures for mobile robot systems.

3 Mobile Robot Kinematics and Dynamics Fig. 1 Kinematic structure of a unicycle

3.1 Mobile Robot Kinematics

An extension of the car-like WMR is the car-pulling trailer
Robot Kinematics deals with the configuration of robots
WMR, in which N single-axis trailers are attached to a
in their workspace, the relations between their geometric
car-like robot with rear-wheel drive.
parameters, and the constraints imposed in their trajectories.
The two classes of holonomic WMRs that have been
The Kinematic equations depend on the geometrical
mostly studied are: (i) the multi-wheel omnidirectional
structure of the robot. For example, a fixed robot can have
WMR with orthogonal (universal) wheels, (ii) the four-
a Cartesian, cylindrical, spherical, or articulated structure,
wheel omnidirectional WMR with mecanum wheels (roller
and a mobile robot may have one two, three, or more wheels
angle ±45◦ ).
with or without constraints in their motion. The study of
Representative references on mobile kinematics, where
kinematics is a fundamental prerequisite for the study of
several complexities (e.g., motion with slip, motion in
dynamics, the stability features, and the control of the robot.
uneven terrain, etc.) are handled, are [53–59].
The development of new and specialized robotic kinematic
structures is still a topic of ongoing research, toward the end
3.2 Mobile Robot Dynamics
of constructing robots that can perform more sophisticated
and complex tasks in industrial and societal applications.
Like kinematics, dynamics is distinguished in:
As in fixed-place robots, the fundamental mathematical
tool is the concept of ‘homogeneous transformations’, and • Direct dynamics.
the concept of ‘nonholonomic constraints’. The kinematics • Inverse dynamics.
of robots (articulated, mobile, etc.) involves the direct
Direct dynamics provides the dynamic equations that describe
kinematics (from the joint space qεR n to the task space
the dynamic responses of the robot to given forces/torques
pεR m ) and inverse dynamics (from pεR m to qεR n ):
τ1 , τ2 , ...., τm that are exerted by the motors.
p = f(q), q = f−1 (p) Inverse dynamics provides the forces/torques that must
be exerted to get desired trajectories of the robot links.
and, respectively, direct differential kinematics and inverse
In the inverse dynamic model the inputs are the desired tra-
differential kinematics:
  jectories of the link variables, and outputs the motor torques.
−1 ∂xi The dynamic equations of a WMR are derived by the
dp = Jdq, dq = J dp, J = ∈ R m×n
∂qj Newton-Euler and Lagrange equations which have been
fully studied in mechanics.
where p = [x1 , x2 ,. . . .,xm ]T .
The general Lagrange model of a multilink robot is:
The major nonhlonomic wheeled mobile robots (WMR)
are: (i) the differential drive WMR, (ii) the tricycle WMR,
and (iii) the car-like WMR, which have been extensively D(q)q̈ + h(q, q̇) + g(q) = τ , q = [q1 , q2 , ..., qm ]T (3)
studied. The basic kinematic model which is used for
modeling the above WMR is the unicycle (Fig. 1). This
where, for any q̇  = 0, D(q) is an n × n positive definite
model is described by the equations:
matrix. The corresponding model for a nonholonomic robot
ẋQ = vQ cos φ, ẏQ = vQ sin φ, φ̇ = vφ (1) with m nonholonomic constraints:
By eliminating vQ from the first two equations, we obtain
the nonholonomic constraint: M(q)q̇ = 0, M(q)an m × n matrix, has he form:
−ẋQ sin φ + ẏQ cos φ = 0 (2) D(q.)q̇ + C(q, q̇)q̇ + g(q) + MT (q)λ = Eτ (4)
40 J Intell Robot Syst (2018) 91:35–58

where λ is a vector Lagrange multiplier. Now, defining The first class of controllers uses the affine dynamic model
a matrix B(q) such as q̇(t) = B(q)v(t), and working on [74, 75]:
Eq. 4, we get the unconstrained Lagrange model:

m
D̄(q)v̇ + C̄(q, q̇)v + ḡ(q) = Ēτ (5) ẋ = go (x) + gi (x)ui , x ∈ R n , g(x) ∈ R n
i=1
where D̄ = BT DB, C̄ = BT DḂ + BT CB, ḡ = BT g, Ē = = go (x) + G(x)u, G(x) = [g1 (x); ..., gm (x)]
BT E.
The dynamic model (5) has found extensive use for where x = [x1 , x2 , ...xn ]T , u = [u1 , u2 , ..., um ]T (m ≤ n).
finding the dynamic models of WMRs of any kind (with The drift term go (x) represents the general kinematic
slip, uneven terrains, differential drive, omnidirectional, constraints of the system. Fundamental concepts of affine
etc.), and applying control schemes [60–63]. theory are:
The Newton-Euler dynamic model is simpler in the sense • Diffeomorphism: A function of the form:
that does not need extensive derivations like the Lagrange
model, and in many cases it is preferred [64–66].
z = ϕ(x), x ∈ R n , z ∈ R n
where ϕ(x) is a vector function (field) with the properties (i)
4 Wheeled Mobile Robot Control ϕ −1 (ϕ(x)) = x, ϕ(ϕ −1 (z)) = z for all x ∈ R n and z ∈ R n ,
and (ii) both functions have continuous partial derivatives of
4.1 Standard Controllers any order (smooth functions).
• Lie derivative: The function
The standard general robot controllers that have been studied
over the years are:
 
• ∂s(x) ∂s(x) ∂s(x) ∂s(x)
Proportional plus integral controller. = ∇s(x) = , , ...,
• Lyapunov function-based controller. ∂x ∂x1 ∂x2 ∂xn
• Computed torque controller. where s(x) = s(x1 , x2 , ...., xn ) ∈ R is a smooth scalar real-
• Resolved motion rate controller. valued function and f(x) ∈ R n is a vector field, is said to be
• Resolved motion acceleration controller. the Lie derivative of s(x) along the field f(x).
All these controllers have been extensively applied to WMRs. • Lie bracket: It is symbolized by [f, g](x) and defined as:
The control procedure of a WMR involves two stages, [f, g](x) = (∂g/∂x)f(x) − (∂f/∂x)g(x), x ∈ R n , where
namely [19, 67]: ∂f/∂x and ∂g/∂x are the Jacobian matrices of the fields
f (x) and g (x).
• Kinematic tracking/stabilizing control.
• Dynamic tracking control. The second class of mobile robot stabilizing control uses
the concept of invariant manifolds, and leads to nonlinear
The first uses the kinematic model of the robot, with a proper
controllers directly. A manifold is a topological space which
candidate Lyapunov function, and yields the control laws
is locally Euclidean, i.e., around every point there is a
for the linear and angular velocities (v, ω) of the robot. The
neighborhood which is topologically the same as the open
second uses the Lagrange model or the Newton dynamic model
unit ball in R n .
which may involve or not the motor dynamics and the
gear box. This procedure belongs to the general class of
Invariant Manifold A manifold M = {x ∈ R n : s(x) = 0},
back-stepping control [67–70]. A solution of the control of
where s : R n → R m is a smooth map, is invariant for
omnidirectional robots, which uses the resolved accelera-
the dynamic system ẋ = f(x, u) if all system trajectories
tion control scheme, was provided in [19, 71]. The case of
starting in M at t = to remain in this manifold for all t ≥ t0 .
parking control of a car-like WMR was studied in [72], and
This implies that the Lie derivative of s along the vector field
a formation (leader-follower) controller was derived in [73].
f is zero, i.e., Lf s(x) = 0, for all x ∈ M.
The concepts of invariant set and invariant manifold
4.2 Advanced Controllers
extend the concept of equilibrium point (which is an invari-
ant monoset), and enable the construction of Lyapunov
Here, we will deal with two kinds of advanced controllers,
functions for nonlinear systems. The invariant set-based
namely:
Lyapunov stability is based on the LaSalle local and global
• State feedback linearization-based controllers. invariant set theorems [75], the Krasovskii theorem, and the
• Invariant manifolds-based controllers. Brockett theorem [76].
J Intell Robot Syst (2018) 91:35–58 41

4.2.1 Mobile Robot State Feedback Linearization as t → ∞, and so: [z1 (t), z2 (t), z3 (t)]T → [ 0 0 0 ]T as
t → ∞. We see that M does not depend on k1 . Now,
There are two classes of linearization via state feedback: (i) to construct a stabilizing control law which makes M an
input-state linearization, and (ii) input-output linearization attractive manifold the control law (6a) must be enhanced
[75]. The nonholonomic mobile robots can be modeled by so as to satisfy the attractivity condition:
the affine system:
If s(z) < 0, then ṡ(z) > 0 z ∈ R 3
ẋ = f(x) + g(x)u, x ∈ R n , u ∈ R m
If s(z) > 0, then ṡ(z) < 0 z ∈ R 3 (7a)
For single-input systems (u ∈ R) there has been derived a
generalized linearizing feedback law through the derivation A possible enhancement is:
of a generalized controllable canonical form [74, 75].   
−k1 z1 − z2 H (s)/ z12 + z22 
This control law was adapted to mobile (differential drive, u= , z12 + z22  = 0 (7b)
car-like) robots in [77]. Then, the trajectory tracking −k1 z2 + z1 H (s)/ z12 + z22
control was solved for the resulting linearized systems, where the scalar mapping H(s) is selected so as:
using conventional linear state-feedback control [19, 77].
Analogous results were obtained in [59, 78]. sH (s) < 0,

4.2.2 Mobile Robot Invariant Manifolds-Based Stabilizing to assure the satisfaction of the attractivity condition (7a).
Control A function H(s) with this property is H (s) = −k2 s. It
can be easily verified that the closed-loop system with the
This approach leads directly to nonlinear controllers without controller (7b) and z12 (0) + z22 (0)  = 0 drives the unicycle to
prior feedback linearization. The two general models the origin, while avoiding the manifold
that have been used are: (i) the nonholonomic Brockett 
integrators (simple, double, extended), and (ii) the (2-n) M ∗ = z ∈ R 3 : z12 + z22 = 0, z1 z2 − 2z3  = 0
chain models. This approach treats in an elegant way the
nonholonomic constraints, and a rich literature exists with
a large repertory of different controllers [76, 79–85]. The 5 Mobile Robot Adaptive and Robust Control
simplest problem is that of stabilizing control of a unicycle
in chained model form (Fig. 1): Adaptive control is suitable for systems that involve slowly
ż1 = u1 , ż2 = u2 , ż3 = z2 u1 varying parameters or uncertainties/disturbances due to load
variation, fuel consumption, etc. [86, 87]. Robust control is
where u1 = vφ and u2 = vQ − z3 u1 . applied in cases where there are strong parameter variations
The problem is to find a static quasi-continuous state or uncertainties, under the assumption that bounds of these
feedback control law u = u(z), u = [u1 , u2 ]T , z = uncertainties are a priori unknown [87]. The adaptive
[z1 , z2 , z3 ]T . It can be easily verified that the control law: controllers (control laws or control algorithms) improve
u = [−k1 z1 , −k1 z2 ], k1 > 0 (6a) their performance as the adaptation evolves with time.
Robust controllers can face fast disturbances, fast variations,
makes the origin z = [z1 , z2 , z3 ]T = [ 0 0 0 ]T globally
and non-modeled characteristics, and are attempt to keep an
asymptotically stable. The resulting closed-loop system is:
acceptable performance right from the beginning. Almost
ż = f(z), f(z) = [−k1 z1 , −k1 z2 , −k1 z1 z2 ]T (6b) always, the adaptive control techniques require some linear
The manifold: parameterization of the dynamics of the nonlinear system
under control.
M = {z ∈ R 3 : s(z) = z1 z2 − 2z3 }
is an invariant manifold of the closed-loop system, 5.1 Adaptive Control

3
since: Lf s(z) = (∂s/∂zi )fi (z1 , z2 , z3 ) = z2 (−k1 z1 ) The two widely used adaptive control methods are:
i=1
+z1 (−k1 z2 ) − 2(−k1 z1 z2 )= 0. The time derivative of s(z) • The model reference adaptive control (MRAC) method.
along the trajectories of Eq. 6b is found to be: • The self-tuning control (STC) method.
ṡ(z) = z1 u2 − z2 u1 = 0
The typical adaptive control method that has been applied
This means that the trajectories, once on the surface (manifold) to wheeled mobile robots is the MRAC method [88–91].
M, remain there. Now, since z1 (t) → 0 and z2 (t) → 0 For example, in [90, 91] the problem solved is that
as t → ∞, for any trajectory on M we have z3 (t) → 0 of designing an adaptive feedback tracking controller for
42 J Intell Robot Syst (2018) 91:35–58

the differential drive WMR which is described by (τR = trajectory tracking problem x(t) = xd (t) is equivalent to the
right-wheel torque, τL = left-wheel torque): problem of remaining on the sliding surface S(t) for all t.
Thus, to assure the trajectory tracking x(t)→xd (t), the
v̇ = (1/mr)(τR + τL ) = (1/m)τa
condition s(x, t) = 0 should be maintained, which can be
ω̇ = (2a/I r)(τR − τL ) = (1/I )τb , τα = (τR + τL )/r, done if u(t) is selected such that outside the surface S(t) the
τb = (τR − τL )/r following sliding condition holds:
ẋ = v cos φ, ẏ = v sin φ,
(1/2)ds 2 /dt ≤ −γ |s|,
with state vector p = [x, y, φ]T . The tracking control laws,
for τa and τb , are found to be: where γ is a positive constant. The solution is found by
treating s 2 (x, t) as a Lyapunov function and assuring that
τa = m̂v̇d + Ka ṽc , τb = Iˆω̇d + Kb ω̃c it remains a Lyapunov function despite the presence of the
where ṽc = vd − vc , ω̃c = ωd − ωc , vd and ωd are the disturbance and the model uncertainty.
desired linear and angular velocity of the WMR, and vc , ωc Let a second-order system,
are the corresponding reference velocities. The reference
d 2 x/dt 2 = b(x) + u
models for vc and ωc are assumed to be linear, namely:
v̇r + βrv vr = 0, ω̇r + βrω ωr = 0, βrv > 0, βrω > 0. where x is the scalar output, u the scalar control input, and
Then, selecting a proper candidate Lyapunov function the function b(x) (possibly nonlinear or time-varying) is
and choosing the parameter dynamics such that V̇ ≤ 0, the approximately known with uncertainty bound ρmax , i.e.:
parameter adaptation laws are found.
In [92] the adaptive control problem is solved using the |b̂ − b| ≤ ρmax
input-output linearization for an m-input m-output affine
system. This approach is then followed to solve the same The resulting sliding-mode controller is:
problem for the differential drive WMR. In [19] the MRAC
problem is solved for an omnidirectional robot using the u = û − k sgn (s), û = −b̂ + ẍd − ˙
x̃, k = ρmax + γ
original Landau method [93].
which results in the desired sliding condition:
5.2 Robust Control
(1/2)ds 2 /dt ≤ −γ |s|
A powerful robust control method for nonlinear systems
is the sliding mode control (SMC) method [75], which In practice, to avoid chattering of the control u, the
was applied to mobile robots in [94, 95]. This method was signum (sgn) function is replaced by the saturation (sat)
originally applied to a single-input single-output canonical function. The robust control problem was also solved
nonlinear model: using the Lyapunov stabilization method [75]. The above
methods were applied to design robust control (sliding-
ẋ1 = x2 , ẋ2 = x3 ...ẋn−1 = xn mode, Lyapunov function-based control) of WMRs in [94,
ẋn = b(x) + a(x)u + d(t) 95]. For example, in the case of the differential drive robot,
we have a two dimensional sliding surface:
y = x1
where u(t) is the scalar input, y(t) the scalar output, and sT = [s1 , s2 ]
d(t) a scalar disturbance input.
The nonlinear function b(x) is not exactly known but and so the Lyapunov function should be selected as:
with some imprecision (error)| b(x)| which is bounded
V = (1/2)sT s = (1/2)s12 + (1/2)s22
from above by a known function of x. The tracking error is
x̃(t) = x(t) − xd (t), where x = [y, ẏ, ÿ, ..., y (n−1) ]T and
where ṡ = −Hs − sgn (s), sgn(s)T = [sgn(s1 ), sgn(s2 )]
xd (t) is the desired trajectory xd = [yd , ẏd , ÿd , ..., ydn−1 ]T .
and  = diag[λ1 , λ2 ]. The control law that satisfies the
The function a(x) is also known with uncertainty. The
corresponding sliding condition is:
solution is based on a time-varying sliding surface S within
the state space R n which is defined as: ui = ûi − ki sat (si /U )(i = 1, 2)
s(x, t) = 0, s(x, t) = (d/dt + )n−1
x̃(t)
where U is the thickness of the boundary layer, and ûi are
where is a positive constant that represents the control proper functions determined analogously to the scalar input
signal bandwidth. Under the condition xd (0) = x0 , the case. An alternative solution of the WMR sliding model
J Intell Robot Syst (2018) 91:35–58 43

control problem, using polar coordinates to represent the

position and orientation of the WMR, was given in [96].

6 Mobile Robot Fuzzy and Neural Control

Fig. 2 General structure of a fuzzy logic controller
Fuzzy logic (FL) and neural networks (NN) have found
wide application in the identification, planning, and control
of mobile robots. Fuzzy logic offers a unified approximate
(linguistic) way of drawing conclusions from uncertain data fuzzification unit (IFU), and (iv) an output defuzzification
using uncertain rules. NNs offer the possibility of learning unit (ODU) (Fig. 2).
and training either autonomously (unsupervised learning) or The two main types of fuzzifier are the singleton fuzzifier
non-autonomously (supervised learning), or via evaluation and the bell fuzzifier. The two most popular defuzzification
of their performance (reinforcement learning) [97–100]. methods are the center of gravity (COG) method and the
In many practical cases (including mobile robots) use is mean of maxima (MM) method.
made of combined (hybrid) neurofuzzy systems (NFSs) that Neural networks (NNs) are large-scale systems that involve
provide better performance. Fuzzy sets were coined by Lofti a large number of nonlinear processors called artificial
Zadeh [97], and constitute an extension of the classical neurons, described by a state, a list of weighted inputs from
concept of (crisp) set which has broken the Aristotelian other neurons, and an equation governing their operation
(true-non true, yes-no) dichotomy. [99] (Fig. 3a,b).
A fuzzy set A is defined as: In Fig. 3a, b, xi , i = 1,. . . ,n are the neuron inputs, b is
a bias constant, wi , i = 1,. . . ,n are the synaptic weights,
A = {(x, μA (x))|x ∈ X, μA (x) : X → [0, 1]},  is a summation element, and f(x) = 1/[1 + exp(−βx)]
is the log-sigmoid activation function, with β being a
where X is the so-called reference superset (or universe of
constant.
discourse), and μA (x) is the membership function which
The NN weights adjust their values through learning
takes values in the full closed interval between 0 and 1 (i.e.,
(minimization of a certain objective function via the
0 ≤ μA (x) ≤ 1). In the special case where μA (x) takes
gradient or the Newton-Raphson algorithm, and error back
only the values 0 and 1, then A reduces to a classical (crisp)
propagation). The optimal values of the weights are stored
set, e.g., A = {x1 , x3 , x5 } ⊂ X = {x1 , x2 , x3 , x4 , x5 } or
as the strengths of the neurons’ interconnections.
A = {(x1 , 1), (x2 , 0), (x3 , 1), (x4 , 0), (x5 , 1)}.
The fuzzy systems and the NNs are suitable for systems or
The three fundamental operations of fuzzy sets (inter-
processes that cannot be modeled with concise and accurate
section, union, complement) are defined as extensions of
the respective operations of classical sets. Also, the stan-
dard properties of sets (De Morgan, absorption, asso-
ciativity, distributivity, idempotency) hold here. The fuzzy
inference (or fuzzy reasoning) is an extension of the clas-
sical inference based on the modus ponens and modus
tollens rules. On the basis of these rules Zadeh has formu-
lated the so-called ‘max-min fuzzy composition’ inference
rule:

B =A◦R

where ‘o’ denotes the max-min operation, and:

A = {(x, μA (x))|x ∈ X}, B = {(y, μB (y))},

R = {(x, y), μR (x, y)|(x, y) ∈ X × Y }
μR (xi , yj ) = min{μA (xi ), μB (yj )}(Mamdani rule)
μR (xi , yj ) = μA (xi )μB (yj )(Larsen’s rule)
Fig. 3 a Model of an artificial neuron, b Typical activation function
The general structure of a fuzzy logic controller (FLC) (log-sigmoid function). Other activation functions are: hard-limit
involves four units: (i) a fuzzy IF-THEN rule base (FRB), (threshold), symmetric hard-limit, and tan-sigmoid (tanh). Source:
(ii) a fuzzy inference mechanism (FIM), (iii) an input www.inspirehep.net (/record/1300728/plots)
44 J Intell Robot Syst (2018) 91:35–58

mathematical models (e.g., pattern recognition, machine

vision, control systems, human-based operations, etc.). The
three primary features of NNs are: (i) use of large amounts
of sensory information, (ii) collective processing capability,
and (iii) learning and adaptation capability [99]. The two
NNs that were mostly used in decision and control systems
are the multilayer perceptron (MLP) networks, and the
radial basis functions (RBF) networks that have always one Fig. 5 Fuzzy robot control loop
layer of hidden nodes (Fig. 4a, b). Other NN models include
the recurrent (or dynamic) NNs, the self-organizing maps
(Kohonen NNs), the convolutional NNs, the Hopfield NNs, controller (FBC) [101]. This scheme is known as feedback
and the Boltzmann machine [99]. error learning neurocontroller (FELN) (Fig. 7).
The typical structure of a fuzzy robot control loop is Representative references where the fuzzy control
shown in Fig. 5 [19]. method was applied are [102–109]. In [102, 103] a direct
The general structure of a neurocontrolled robot with adaptive fuzzy tracking control scheme is presented, and in
supervised learning is shown in Fig. 6. [104] a decentralized fuzzy logic control (FLC) scheme for
Other kinds of neural control involve unsupervised multiple WMRs is described. The structure of the adaptive
learning and reinforcement learning. The most general type fuzzy tracking controller of [102, 103] is shown in Fig. 8.
of neurocontrol involves two NNs: the first is used as The control system of Fig. 8 involves two loops, namely:
feed forward controller (FFC) and the second as feedback (i) the kinematic tracking loop, and (ii) the dynamic control
loop, where the dynamic controller is replaced by an FLC
which receives crisp values that are fuzzified in the IFU
unit, and gives crisp values for the robot inputs (torques)
after deffuzification in the ODU unit (Fig. 2). The kinematic
controller remains a crisp controller, since it does not
involve any unknown (or potentially unknown) parameter.
All of its variables are known or measured.
In [109] the fuzzy local path tracking controller for a
Dubins car is described. The kinematic model of Dubins
car is found from the standard car-like model by omitting
the equation for the steering angle velocity. The resulting
controller is a multirate controller.
In [110] a fuzzy sliding mode control scheme, applied to
a WMR, is described. The WMR is assumed to move on a
surface g(x, y, z) = 0 along a continuously differentiable
path p(r) = [x(r), y(r), z(r)] of the center of gravity of the
robot with respect to a world coordinate frame. The sliding
mode controller is similar to that described in Section 4.2,
and has a diagonal structure, namely:

˜ )sgn(s)
uf uzzy = −Kf uzzy (x̃, ẋ,

Fig. 4 a Multilayer perceptron with two layers of hidden nodes, b

Radial basis function network (φi , i = 1, 2, . . ., mare the radial basis
functions, typically Gaussian functions). Source: Neurosolutions:
What is a neural network? Fig. 6 Structure of a robot controlled by a NN with supervised learning
J Intell Robot Syst (2018) 91:35–58 45

Fig. 7 General structure of FELN

with the condition: Fig. 9 Structure of hybrid PD-SMFLC control

Kf uzzy (x̃1 , ẋ˜1 , ) ≤ Kf uzzy (x̃2 , ẋ˜2 , )
for | x̃1 + ẋ˜1 | ≤ | x̃2 + ẋ˜2 |. In [110] a reduced complexity with respect to some landmarks. Tracking stability can be
sliding mode fuzzy logic controller (RC-SMFLC) is assured only if the vision sensing delays are sufficiently
presented, which can be described as: small and/or the dynamic model of the robot has sufficient
uf uzzy = −Kf uzzy (|s|sgn(s)) accuracy. Over the years many techniques were developed
for compensating this delay of the visual system in robot
where s is the distance from the diagonal. The RC-SMFLC
control. A rich literature has been oriented to the control
is applied in parallel with a standard PD controller, as shown
of nonholonomic systems in order to handle various
in Fig. 9.
challenging problems associated with vision-based control.
This technique was applied to provide a powerful
Vision-based robot controllers (VRCs) depend on
solution of the parallel car-parking control problem.
whether the vision system provides set-points as input to the
In [111], the fuzzy two-step (back-stepping) procedure
robot joint controllers or computes directly the joint level
of [102, 103] was applied using an MLP neural network
inputs, and whether the error signal is determined in task
controller in place of a fuzzy controller.
space coordinates or directly in terms of image features.
Finally, in [112] the control of a differential drive
Therefore, VRCs are classified in the following three
robot is described which uses radial-basis networks for the
categories [113, 114, 116, 117].
estimation task.
• Dynamic look-and-move system: Here the robot joint
controller is eliminated and replaced by a visual servo
7 Mobile Robot Vision-Based Control controller which directly computes the inputs of the
joints, and stabilizes the robot using only vision signals.
Vision is a powerful robotic sensor which can be used for Actually, most implemented VRCs are of the look-
environment measurement without physical contact [113]. and-move type because internal feedback with a high
Visual robot control or visual servoing is a feedback sampling rate provides the visual controller with an
control methodology that uses one or more vision sensors accurate axis dynamic model. Also, look-and-move
(cameras) to control the motion of the robot. Specifically, control separates the kinematics singularities of the
the control inputs for the robot motors are produced by system from the visual controller, and bypasses the low
processing image data (typically, extraction of contours, sampling rates at which the direct visual control can
features, corners, and other visual primitives). In robotic work.
manipulators, the purpose of visual control is to control • Position-based visual robot control (PBVRC): Here,
the pose of the robot’s end-effector relative to a target use is made of features extracted from the image and
object or a set of target features. In mobile robots, the used together with a geometric model of the target and
visual controller’s task is to control the vehicle’s pose the available camera model to determine the pose of the

Fig. 8 Structure of overall

adaptive fuzzy tracking
controller for differential drive
robot
46 J Intell Robot Syst (2018) 91:35–58

target with respect to the camera. Thus, the feedback The platform of a mobile robot is moving with a linear
loop is closed using the error in the estimated pose velocity v(t) = [vx , vy , 0]T and an angular velocity ω(t) =
space. [0, 0, ω]T (where ω = ωz ). Thus, Eq. 9 becomes ω × p =
• Image-based visual robot control (IBVRC): Here, [−yω, xω, 0]T , and Eq. 8 gives:
direct computation of the control signals is performed    
ẋ vx − yω
using the image features. IBVRC reduces the compu- = , ż = 0 (z = constant = 0) (12)
ẏ vy + xω
tational time, does not need image interpretation, and
eliminates the errors of sensors’ modeling and camera Equation 12, combined with φ̇ = ω, gives the overall
calibration. But its implementation is more difficult due equation:
to the complex nonlinear dynamics of robots. ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
ẋ 1 0 −y vx
⎣ ẏ ⎦ = ⎣ 0 1 x ⎦ ṙ, ṙ = ⎣ vy ⎦ (13)
The PBVRC and IBVRC control schemes have the
structure shown in Fig. 10 [19]. φ̇ 00 1 ω
Two fundamental concepts of VRC are (i) the kinematic If the WMR involves a steering angle ψ, then the
transformations, and (ii) the camera visual transformations. corresponding equation ψ̇ = ωψ should be added. The
Consider a fixed-base robot manipulator working in 3D camera visual transformations are typical derived using
space. The motion of its end-effector is described in world the perspective projection model, in which a point p =
coordinates x, y, z by a translational velocity v and an [x, y, z]T whose coordinates are expressed with respect to
angular velocity ω, where: the camera coordinate frame Ac , projects onto the image
plane point f = [xim , yim ]T given by:
v(t) = [vx , vy , vz ]T , ω(t) = [ωx , ωy , ωz ]T    
xim lf x
Let p = [x, y, z]T be a point rigidly attached to end- f(x, y, z) = = (14)
yim z y
effector, then:
where lf is the camera’s focal length (Fig. 11).
ṗ = ω × p + v, (8) Differentiating Eq. 14 we get:
where ω × p is the cross product of ω and p, i.e.:    
ẋ l /z 0 −xim /z
ḟ = im = Jc (xim , yim , lf )ṗ, Jc = f
ω × p = [ωy z − yωz , ωz x − zωx , ωx y − xωy ]T (9) ẏim 0 lf /z −yim /z
The combined velocity vector r = [vT , ωT ]T
is known as (15)
velocity screw (or velocity twist) of the robotic and effector. Now, combining Eqs. 10, 11, 14, and 15 we find:
In compact form (8) and (9) can be written as:
ḟ = Jc (xim , yim , z, lf )J0 (p)ṙ
ṗ = J0 (p)ṙ, J0 (p) = [ I3×3 | S(p) ] (10) = Jim (xim , yim , z, lf )ṙ (16)
where S(p) is the skew symmetric matrix [114]: where Jim (xim , yim , z, lf ) = Jc (xim , yim , z, lf )J0 (p) is
⎡ ⎤ called the image Jacobian which depends on the distance
0 z −y
S(p) = ⎣ −z 0 x ⎦ (11) z of the end effector (or the target point being imaged, in
y −x 0 general).
The image Jacobian matrix of a unicycle-type WMR with
a pinhole on board camera and a target with three feature
points in the camera field of view was derived in [115]
and has the form (16). A position-based visual controller
for path following by nonholonomic robots is described in
[116]. Several image based controllers for mobile robots
are presented in [117–121]. An image-based control scheme

Fig. 10 a PBVRC loop (up), b IBVRC loop (down) Fig. 11 Geometry of camera lens system
J Intell Robot Syst (2018) 91:35–58 47

of mobile robots with catadioptric cameras is presented in if the environment (obstacles, etc.) is static and perfectly
[122]. An online estimation scheme of the image Jacobian known to the robot. In this case, the path planning algorithm
matrix for uncalibrated stereo vision feedback is provided produces a complete path from the start point to the goal
in [123]. A stable vision-based controller for nonholonomic point before the robot starts its motion.
WMRs to keep a landmark in the field of view is pro- Motion planning is the process of selecting a motion
vided in [124]. An homography-based mobile robot control and the corresponding inputs such that all constraints
scheme with nonholonomic and field-of-view constraints (obstacle avoidance, risk avoidance, etc.) are satisfied.
is presented in [125]. The above references describe only Motion planning can be considered as a set of computations
a few of the published methods and schemes of mobile which provide subgoals or set points for the control of the
robot visual control. More methods can be found in the robot. These computations and the resulting plans are based
references cited therein, and in books on mobile robots on a suitable model of the robot and the environment in
(e.g., [14, 15, 19]). Two works concerning omnidirec- which it is moved. The process by which the robot executes
tional vision-based mobile robot control are described in (follows) the planned motion is the control process studied
[126, 127]. in Sections 4–7.
We recall that the motion of the robot can be described in
three different spaces:
8 Mobile Robot Path and Motion Planning
• Task or Cartesian space.
• Joints’ (motors’) space.
Robot planning is concerned with the general problem of
• Actuators’ space.
figuring out how to move to get from one place to another
place (path planning, motion planning) and how to perform A very broad classification of free (obstacle-avoiding)
a desired task (task planning) [128, 129]. Here we will be path planning involves three categories, which include six
concerned with path planning and motion planning. distinct strategies. These are the following:
Path planning of mobile robots is one of the basic
operations needed to implement the navigation of the robot. • Reactive control (‘Wander’ routine, circumnavigation,
These operations are: potential fields, motor schemas).
• Representational world modeling (certainty grids).
• Self-localization. • Combinations of both (vector field histogram).
• Path planning.
• Map-building and map interpretation. In many cases, the above techniques do not assure that a path
is found that passed obstacles although it exits, and so they
Robot localization provides the answer to the question need a higher level algorithm to assure that the mobile robot
‘where am I?’ The path planning operation provides the does not end up in the same position over and over again. In
answer to the question ‘how should I get to where I am practice, it may be sufficient that the robot detects that it is
going?.’ Finally, the map building/interpretation operation ‘stuck’ despite the fact that a feasible path way exists, and
provides the geometric representation of the robot environ- calls for help. In indoor applications, a maneuver for avoiding
ment in notations suitable for describing location in the an obstacle is a good action. Outdoor situations are more
robot’s reference frame. So far, it seems that there is not complex, and more advanced perception techniques are
a generic method for mobile robot positioning/localization. needed (e.g., for distinguishing a small tree from an iron pole).
The specific techniques that exist are divided in two A research topic that received much attention over the
categories: years is the piano-mover’s problem, which is well known to
most people that tried a couch or big table through a narrow
• Relative localization methods. door. The object has to be tilted and moved around through
• Absolute localization methods. the narrow door. One of the first research works on this
problem is described in Latombe [2].
Relative localization is performed by odometry or inertial
On the basis of the way the information about the robot’s
navigation. Absolute localization uses active bacons,
environment is obtained, most of the path planning methods
recognition of artificial landmarks, recognition of natural
can be classified into two categories:
landmarks, and model matching.
Path planning may be either local or global. Local path
1. Model-based approach
planning is performed while the robot is moving, taking data
2. Model-free approach
from local sensors. In this case, the robot has the ability
to generate a new path in response to the changes of the In the first category, all the information about the robot’s
environment. Global path planning can be performed only workspace is learned beforehand, and the user specifies the
48 J Intell Robot Syst (2018) 91:35–58

geometric models of objects, and a description of them in • Cell decomposition.

terms of these models. In the model-free approach, some of • Potential fields.
the information about the robot’s environment is obtained • Vector field histograms.
via sensors (e.g., vision, range, touch sensors). The user has to
Representative works on mobile robot path planning and
specify all the robotic motions needed to accomplish a task.
navigation using the above methodologies are the following:
The obstacles that may exist in a robotic work environ-
ment are distinguished into stationary obstacles and moving • Gasparetto [135]: Solution of 2-D constrained WMR
obstacles. Therefore, two types of path finding problems trajectory planning.
have to be solved, namely: • Hatzivasiliou and Tzafestas [136]: WMR path planning
in structured environment.
• Path planning among stationary obstacles.
• Garcia and Santos [137]: WMR navigation with
• Path planning among moving obstacles.
complete coverage in unstructured environment.
The path planning methodology for stationary obstacles • Safadi [138]: Local path planning using virtual potential
is based on the configuration space concept, and is field.
implemented by the so-called road map planning methods. • Ding, Jiang, Bian and Wang [139]: Local path planning
The path planning problem for the case of moving obstacles based on virtual potential field.
is decomposed into two subproblems: • Koren and Borenstein [140]: Potential field methods for
WMR navigation.
• Plan a path to avoid collision with static obstacles. • Borenstein [141]: Vector field histogram for fast WMR
• Plan the velocity along the path to avoid collision with obstacle avoidance.
moving obstacles. • Wang, Yong and Ang Jr. [142]: Hybrid global path plan-
This combination constitutes the robot motion planning ning and local WMR navigation in indoor environment.
[130–133]. • Garrido, Moreno, Blanco and Jurewicz [143]: WMR
Configuration q of a robot is an n-tuple of real numbers path planning using Voronoi diagram and fast marching.
that specifies the n parameters required to determine the • Garrido, Moreno and Blanco [144]: Exploration of a
position of the robot in physical space. The configuration cluttered environment using Voronoi diagrams.
space (CS) of the robot is the set of values that its configura- • Arney [145]: Autonomous WMR path planning by
tion q may take. The subset of CS of configurations that are approximate cell decomposition.
not in collision with any obstacles that exist in the robot’s • Katevas, Tzafestas and Pneumatikatos [146]: Approx-
environment is called the free configuration space CSf ree . imate cell decomposition with local node refinement
In terms of CS, the path planning problem of a robot is the WMR path planning.
problem of finding a path in the free configuration space • Olunloyo and Ayomoh [147]: WMR navigation using
CSf ree . Examples of solution of the path planning problem hybrid virtual force field.
of robotic manipulators via CSf ree can be found in many • Katevas, Tzafestas and Matia [148]: Global and local
references (e.g., [134]). path strategies for WMR navigation.
Typically, CSf ree path planning methods involve two • Katevas and Tzafestas [149]: The active kinematic
operations: histogram method for path planning of non-point
nonholonomically constrained mobile robots.
• Collision checking (i.e., check whether a configuration, • Zelinski, Jarvis, Byrne and Yuta [150]: Planning paths
or a path between two configurations, lies entirely in of complete coverage of an unstructured environment
CSf ree ). by a mobile robot.
• Kinematic steering (i.e., find a path between two configu- • Zelinski and Yuta [151]: Unified approach to WMR
rations q0 and qf in CS that satisfies the kinematic planning, sensing, and navigation.
constraints, without taking into account obstacles). • Choi, Lee, Baek and Oh [152]: Online complete
The robot navigation maps that are used to represent the coverage WMR path planning.
environment can be a continuous geometric description
or a decomposition-based geometric map or a topological
map. These maps must be converted to discrete maps 9 Mobile Robot Localization and Mapping
appropriate for the path algorithm under implementation.
This conversion (or decomposition) can be done by four Localization and mapping are two of the five basic operations
general methodologies, namely [2, 130]: needed for the navigation of a robot. Very broadly, other
fundamental capabilities and functions of an integrated
• Road maps (visibility graphs, Voronoi diagrams). robotic system from the task/mission specification to the
J Intell Robot Syst (2018) 91:35–58 49

motion control/task execution (besides path planning) are • Trilateration.

the following: • Triangulation.
• Map matching.
• Cognition of the task specification.
• Perception of the environment.
• In general, the sensor imperfections can be grouped in:
Control of the robot motion.
sensor noise and sensor aliasing categories [156–158].
The sensor noise is primarily caused by the environmental
The robot must have the ability to perceive the environment
variations that cannot be captured by the robot. Examples
via its sensors in order to create the proper data for finding
of this in vision systems are the illumination conditions,
its location (localization) and determining how it should go
the blooming, and the blurring. In sonar systems, if the
to its destination in the produced map (path planning). The
surface accepting the emitted sound is relatively smooth
desired destination is found by the robot through processing
and angled, much of the signal will be reflected away,
of the desired task/mission command with the help of the
failing to produce a return echo. Another source of noise
cognition process. The path is then provided as input to the
in sonar systems is the use of multiple sonar emitters (16–
robot’s motion controller which drives the actuators such
40 emitters) that are subject to echo interference effects.
that the robot follows the commanded path. The general
The second imperfection of robotic sensors is the aliasing,
structure of the interrelations between the above operations
that is, the fact that sensor readings are not unique. In
has the form of Fig. 12 [19].
other words, the mapping from the environmental states
The sensors have inaccuracies/uncertainties and so the
to the robot’s perceptual inputs is many-to-one (not one-
localization based on data provided by them needs to
to-one). The sensor aliasing implies that (even if no noise
employ stochastic estimation of the relevant parameters,
exists) the available amount of information is in most cases
variables and states. The three primary estimation methods
not sufficient to identify the robot’s position from a single
used in mobile robot localization and mapping are [153–
sensor reading. The above issues suggest that in practice
155, 168]:
special sensor signal processing/fusion techniques should
be employed to minimize the effect of noise and aliasing,
• Kalman estimation (filtering, prediction).
and thus get an accurate estimate of the robot position
• Bayesian estimation.
over time.
• Particle filter (PF).
Relative localization is performed by dead reckoning,
i.e., by the measurement of the movement of a wheeled
But other techniques such as fuzzy and neural estimators mobile robot (WMR) between two locations. This is
(approximators) have also been used [23, 170]. The basic done repeatedly as the robot moves and the movement
concept of localization is the dead reckoning (relative measurements are added together to form an estimate of
localization) which can be performed by simple WMR the distance traveled from the starting position. Since the
kinematic analysis. individual estimates of the local positions are not exact,
Absolute WMR localization can be performed by: the errors are accumulated and the absolute error in the
total movement estimate increases with traveled distance.
The term dead reckoning comes from the sailing days term
‘deduced reckoning’ [156].
For a WMR, the dead reckoning method is called
‘odometry’, and is based on data obtained from incremental
wheel encoders [159].
The basic assumption of odometry is that wheel revolu-
tions can be transformed into linear displacements relative
to the floor. This assumption is rarely ideally valid because
of wheel slippage and other causes. The errors in odometric
measurement are distinguished in:

• Systematic errors (e.g., due to unequal wheel diameters,

misalignment of wheels, actual wheel base is different
than nominal wheel base, finite encoder resolution,
encoder sampling rate).
Fig. 12 Structure of an autonomous mobile robot (interconnection of • Non-systematic errors (e.g., uneven floors, slippery
cognition/path planning, perception and control) floors, over-acceleration, nonpoint contact with the
50 J Intell Robot Syst (2018) 91:35–58

floor, skidding/fast turning, internal and external • Kim and Chung [171]: SLAM with omnidirectional
forces). stereo vision sensor.
Systematic errors are cumulative and occur principally in
indoor environment. Nonsystematic errors are dominating
in outdoor environments. 10 Intelligent Control and Software
The problem of placing a WMR at an unknown location Architectures for Mobile Robots
in an unknown environment, while the WMR is incre-
mentally building a consistent map of this environment, is Systemic (intelligent control) architectures are used to
known as SLAM (simultaneous localization and mapping). integrate controllers and high-level functional units for
The SLAM problem was first studied in [171] by Durrant- achieving overall intelligent performance of mobile robots.
Whyte who established a statistical basis for describing rela- Autonomous mobile robots should be extremely self-reliant
tionships between landmarks and manipulating geometric to operate in complex, partially unknown environments via
uncertainty. control systems assuring in real time that the robot will
To solve the SLAM problem one needs to use a total perform correctly its tasks despite the above constraints. In
(joint) state which incorporates the WMR’s pose (posi- addition, the software architectures have to face the high
tion/orientation) and every landmark position. This joint degree of heterogeneity among the subsystems involved,
state should be estimated and updated following each land- and deal with the strict operational requirements posed by
mark observation. For the environment observation/sensing the real-time interactions with the robot’s environment.
there is available a whole gamma of sensors (sonars, cam- The achievement of autonomous behavior is assured
eras, laser range finders, etc.) [157, 158]. The estimation by using techniques of intelligent control (IC) which
of the joint state can be performed by an extended Kalman started with the development of generic intelligent control
filter (EKF), or a Bayesian estimator (BE), or a particle architectures (ICAs). The principal ICAs are the following:
filter/estimator (PF). EKF is an extension of the Kalman
• Hierarchical ICA (Saridis) [172, 173]
filter that covers nonlinear stochastic models. Bayesian esti-
• Multiresolutional/nested ICA (Meystel) [174, 175].
mators describe the WMR motion and feature observations
• Reference model ICA (Albus) [176, 177].
directly using the underlying probability density functions
• Behavior-based ICAs, namely: subsumption ICA
and Bayes updating law. The PF (also called sequential
(Brooks) [178, 179], and motor schemas ICA (Arkin)
Monte Carlo estimator) is based on simulation [171]. An
[180, 181].
outline of the above methods is provided in [19].
• Task ICA (Simmons) [182–184].
Representative references devoted to the mobile robot
localization and mapping problem are: These architectures were expanded, enriched, or combined
over the years in several ways [185]. Most of the software
• Beke and Gurvis [160]: Mobile robot localization using
systems and integrated hardware-software systems devel-
landmarks.
oped for intelligent mobile robot control follow, in one or
• Hu and Gu [161]: Robot landmark-based localization.
the other way, one of these generic architectures or suitable
• Andersen and Concalves [162]: Vision-based mobile
combinations of them.
robot localization using triangulation.
• Castellanos and Tardos [163]: Multisensor fusion for • The hierarchical intelligent control architecture has
mobile robot localization and map building. three main levels, namely (Fig. 13):
• Guivant and Nebot [164]: Optimal simultaneous
1. Organization level which implements the higher-
localization and mapping.
level functions (e.g., learning, decision making).
• Rekleitis, Dudek and Milios [165]: Probabilistic coop-
erative localization and mapping.
• Bailey and Durrant-Whyte [166]: Simultaneous local-
ization and mapping (SLAM).
• Rekleitis, Dudek and Millios [167]: Multirobot collabo-
ration for robust exploration.
• Crisan and Doucet [168]: Convergence of particle
filtering methods.
• Rituerto, Puig and Guerrero [169]: Mobile robot
SLAM with omnidirectional camera.
• Rigatos and Tzafestas [170]: Fuzzy modeling and
multi-sensor fusion via extended Kalman filtering Fig. 13 Saridis’ hierarchical ICA
J Intell Robot Syst (2018) 91:35–58 51

2. Coordination level which consists of several

coordinators.
3. Execution level which involves the actuators, the
hardware controllers, and the sensing devices,
and executes the action programs issued by the
coordination level.
Saridis has developed a complete analytic theory for this Fig. 14 The subsumption architecture
architecture, formulating and exploiting the Principle of
Increasing Precision with Decreasing Intelligence (PIPDI)
using the information entropy concept [186].
• The motor schemas architecture (MSA) was more
• The Multiresolutional Intelligent Control Architecture strongly motivated by biological sciences and uses
(MICA) was developed by Meystel and first applied to the theory of schemas originated by Kant. Schemas
intelligent mobile robots. It follows the commonsense represent a means by which understanding is able to
model Planner-Navigator-Pilot-Execution Controller. categorize sensory perception in the process of realizing
The Planner delivers a rough plan. The Navigator knowledge of experience. The three representative
computes a more precise trajectory of the motion to definitions of the schema concept are [7]:
be executed. The Pilot develops online tracking open-
loop control. Finally, the Execution Controller executes 1. A pattern of action or a pattern for action.
plans and compensations computed by the planner, the 2. An adaptive controller (based on an identification
navigator, and the pilot. This scheme is implemented procedure).
in the form of the so-called multiresolution 6- 3. A perceptual entity that corresponds to a mental
box. Each level contains perception (P), knowledge entity.
representation, interpretation and processing (K), and The capabilities of schema-based analysis and design of
planning and control (P/C) operations. behavior-base systems are:
• The Reference model architecture (RMA) was devel-
oped and expanded by Albus and colleagues, and is 1. It can explain motor behaviors in terms of the
suitable for modular expansion. Its control structure concurrent control of several different activities.
involves the following: 2. It can store both how to react and how to realize this
reaction.
1. Task decomposition.
3. It can be used as a distributed model of computation.
2. World modeling.
3. Sensory processing. • The task control architecture (TCA) is a high-
4. Value judgment. level robot operating system with an integrated
set of commonly needed mechanisms to support
The various control elements are clustered into computational
distributed communications, task decomposition,
nodes arranged in hierarchical layers, each one of which
resource management, execution monitoring, and
has a particular function and a specific timing behavior. The
error recovery. A system based on TCA involves
main design issues addressed by RMA are: (i) real-time task
a set of specific modules and a general purpose
and software execution, (ii) smart interface/communication,
reusable control module. The modules communi-
(iii) information/knowledge base management, and (iv)
cate with each other and with the central control by
optimal allocation of resources.
passing messages. The TCA possesses many fea-
• The Behavior-Based Control Architectures (BBCA) are tures of the blackboard (BB) architectures [187],
based on the concept of agent and can be implemented but differs from them because (although it main-
using knowledge-based systems, neural, fuzzy or neuro- tains control information centrally), the actual data
fuzzy structures. The two most common behavior-based need to solve problems in distributed way among
architectures are the ‘subsumption’ architecture devel- the system’s processes.
oped by Brooks, and the ‘motor schema’ architecture
The control software architecture of a semiautonomous/
developed by Arkin. The subsumption architecture fol-
autonomous mobile robot must meet the following desirable
lows the decomposition of the behavior paradigm and
requirements (characteristics) [188, 189].
was first employed in the autonomous robot Shakey.
Complex actions subsume simple behaviors (Fig. 14). • Robot hardware abstraction/portability.
The reactions are organized in a hierarchy of levels where • Extendibility/scalability (capability to add new hardware
each level corresponds to a set of possible behaviors. modules and new software components to the system).
52 J Intell Robot Syst (2018) 91:35–58

• Reusability (e.g., software reuse of components, struc-

ture, framework, and software patterns).
• Repeatability (which means that running the same
program on the same input gives the same result).
• Run-time overhead (which is specified by memory and
CPU requirements, frequency and end-to-end latency)
The control system of a semiautonomous/autonomous
intelligent robot is required to possess the following general
features [38].

• Reactivity to the environment.

• Robustness against imperfect inputs and unexpected
events or sudden failures.
• Multiple sensor integration to compensate the limited
accuracy, reliability and applicability of individual sensors.
• Modularity (i.e., the system modules must be able to be
separately and incrementally designed, implemented, Fig. 15 a Hierarchical vs. centralized systems, b Reactive vs.
debugged, and maintained). deliberative (planning) systems
• Expandability (i.e., the ability to build the system
incrementally). Two important software architectures are: (i) the Jde
• Adaptability (ability to adapt to rapid and unpredictable (component-oriented) architecture which uses schemas
changes of the world state). combined in dynamic hierarchies to unfold the global
behavior [190], and (ii) the layered mobile robot control
In practice, to keep the overall system complexity at a
architecture which involves four hierarchical layers [191].
reasonable level, specific compromises are made such as:
Three representative research prototypes of integrated
• Reduction of the autonomy level and allocation of more mobile robots are: (i) the SENARIO robotic wheel chair
difficult tasks or decisions to a human operator. [192], (ii) the KAMRO (Karlsruhe Autonomous Intelligent
• Reduction of the environment complexity by changing Mobile Robot) [193], and (iii) the Munich ROMAN intel-
the environment so as to make it more robot friendly ligent mobile manipulator [194]. The SENARIO mobile
(e.g., by introducing landmarks for navigation, etc.). robot uses a ‘virtually’ centralized hierarchical control
architecture, and has two alternative operating modes, viz.,
Very broadly, the majority of mobile robot architectures
(i) semi-autonomous mode, and (ii) fully autonomous mode.
found in the literature can be classified according to the
The KAMRO robot uses a natural language (NL) human-
following three aspects [38].
robot interface and performs the following functions: (i)
• The way their modules are interconnected (hierarchical task specification and representation, (ii) execution rep-
vs. centralized architectures) (Fig. 15a). resentation, (iii) explanation of error recovery, and (iv)
• The way their modules and the environment communi- description and updating of the environment representation.
cate (reactive vs. deliberative control) (Fig. 15b).
• The function of the modules (functional vs. behavior
systems).
Three ways, of merging deliberative and reactive behavior
were suggested by Arkin, and are shown in Fig. 16 [7]:
• Hierarchical integration of planning and reaction.
• Planning to guide reaction (i.e., allowing planning to
choose and set parameters for the reactive control).
• Coupled planning and reacting (each of these two
concurrent actions guides the other).
One of the first robotic control schemes that were
designed using the hybrid deliberative (hierarchical) and
Fig. 16 a Hierarchical hybrid deliberative-reactive structure, b
reactive (schema based) architecture is the autonomous Planning to guide reaction systems, c Coupled planning and reacting
robot architecture (AuRA) developed by Arkin. scheme
J Intell Robot Syst (2018) 91:35–58 53

The Munich mobile manipulator ROMAN involves a task 8. Canny, J.: The complexity of Robot Motion Planning. MIT Press,
planner and coordinator, and a multi-modal robot interface Cambridge (1998)
9. Zhu, X., Kim, Y., Minor, M.A., Qiu, C.: Autonomous Mobile
for natural voice/speech based dialog between the human
Robots in Unknown Outdoor Environments. CRC Press /Taylor
user and the robot, and the vision-based system localization and Francis Group, Boca Raton (2017)
and navigation. 10. Nehmzow, U.: Mobile Robotics: A Practical Introduction.
Springer, London (2003)
11. Siegwart, R., Nourbakhsh, I.: Autonomous Mobile Robots. MIT
Press, Cambridge (2005)
11 Conclusions 12. Cuesta, F., Ullero, A.: Intelligent Mobile Robot Navigation.
Springer, Berlin (2005)
Autonomous mobile robots (AMRs) need to be able 13. Berns, K., Von Puttkamer, E.: Autonomous Land Vehicles: Steps
Towards Service Robots. Springer, Berlin (2009)
to move purposefully and without human help in real-
14. Dudek, G., Jenkin, M.: Computational Principles of Mobile
world environments (i.e., environments that have not been Robotics. Cambridge University Press, Cambridge (2010)
specifically engineered for the robot). Currently, there is 15. Cook, G.: Mobile Robots: Navigation, Control and Remote
still a gap between the available methodology/technology Sensing. Wiley, Hobolen (2011)
16. Berry, C.A.: Mobile Robotics for Multidisciplinary Study.
and the new application and market demands which
Morgan & Claypool, San Rafael (2012)
require fully autonomous robots. This gap continuously 17. Tiwari, R., Shukla, A., Kala, R.: Intelligent Planning for Mobile
motivates researchers and practitioners in robotics to Robotics: Algorithmic Approaches. IGI Global, Hershey (2012)
develop novel methods and techniques which overcome 18. Kelly, A.: Mobile Robots: Mathematics, Models, and Methods.
Cambridge University Press, Cambridge (2013)
the large uncertainties that are inherent in the real world
19. Tzafestas, S.G.: Introduction to Mobile Robot Control. Elsevier,
environments, by incorporating all necessary details and Inc., New York (2004)
temporary features, and facing unknown changes of spatial 20. Jaulin, L.: Mobile Robotics. ISTE Press-Elsevier, New York
relations between objects or sensory imprecisions and (2017)
inaccuracies. The outcome of this continuous effort is
exhibited by the large amount of published results, over Edited Books
the years, many of which have substantially contributed
toward the development and construction of AMRs with 21. Patnaik, S., Jain, L.C., Tzafestas, S.G., Resconi, G., Konar, A.
(eds.): Innovations in Robot Mobility and Control. Springer,
higher-level autonomy and social interaction capabilities Berlin (2005)
[195]. The present paper has attempted to provide a global 22. Tzafestas, S.G. (ed.): Advances in Intelligent Autonomous
spherical overview of a large number of publications in Systems. Kluwer, Dordrecht/Boston (1999)
the field, focusing particularly in the area of control and 23. Driankov, D., Saffioti, A. (eds.): Fuzzy Logic Techniques for
Autonomous Vehicle Navigation. Physica-Verlag, Heidelberg (2011)
navigation methodologies. Comprehensible descriptions of 24. Azad, A.K.M., Cowan, N.J., Tokhi, M.O., Virk, G.S., Eastman,
the methodologies can be found in the cited books, special R.D. (eds.): Adaptive Mobile Robotics. World Scientific, Singa-
issues, and papers. Some further references in the field that pore (2012)
deal with mobile robot modeling, exponential stabilization, 25. Katevas, N. (ed.): Mobile Robotics in Healthcare. IOS Press,
Amsterdam (2001)
path planning, path following, trajectory tracking, and 26. Rembold, U. (ed.): Intelligent Autonomous Systems (IAS-4). IOS
navigation problems are [196–212]. Press, Amsterdam (1995)
27. Faruk, E., Ceccareli, M. (eds.): Mobile Robots for Dynamic
Environments. ASME Press, New York (2015)
References 28. Fujimoto, H., Tokhi, M.O., Mochiyama, H., Virk, G.S. (eds.):
Emerging trends in mobile robotics. World Scientific, Singapore
(2010)
Authored Books 29. Topalov, A.V. (ed.): Recent Advances in Mobile Robotics. InTech,
Open Access, Austria (2011)
1. Meystel, A.M.: Autonomous Mobile Robots: Vehicles with 30. Buchlo, J. (ed.): Mobile Robotics, Moving Intelligence. InTech,
Cognitive Control. World Scientific, Singapore (1991) Austria (2000)
2. Latombe, J.C.: Robot Motion Planning. Kluwer, Boston (1991)
3. Leonard, J.L.: Directed Sonar Sensing for Mobile Robot
Navigation. Springer, Berlin (1992)
Survey Papers
4. Jones, J.L., Flynn, A.M., Seiger, B.A.: Mobile Robots: Inspiration
to Implementation. Peters A.K., Ltd/CRC Press, New York (1995) 31. Tai, L., Liu, M.: Deep learning in mobile robotics: from perception
5. Everett, H.R.: Sensors for Mobile Robots: Theory and Applica- to control systems–a survey of why and why not. J. LATEX Class
tions. Peters A.K. Ltd/CRC Press, New York (1995) Files 14(8), 1–16 (2015)
6. Borenstein, J., Everett, H.R., Feng, L.: Navigating Mobile Robots: 32. Cao, Y.U., Fukunago, A.S., Kalug, A.B.: Cooperative mobile ro-
Systems and Techniques. Peters A.K. Ltd/CRC Press, New York botics: antecedents and directions. Auton. Robots 4, 1–23 (1997)
(1996) 33. Yamauchi, Y.: A survey of pattern formation of autonomous
7. Arkin, R.C.: Behavior-Based Robotics. MIT Press, Cambridge robots: Asynchrony, obliviousness and visibility. J. Phys.-Conf.
(1998) Ser. 473012016 (2013)
54 J Intell Robot Syst (2018) 91:35–58

34. Cunha, J.: A survey on machine learning in mobile robots. Elect. Robotics and Automation (ISPRA’08), pp. 66–74, Cambridge
Telecommun. 5(3), 304–308 (2011) (2008)
35. Bostelman, R., Hong, T., Marvel, J.: Survey research for 56. Chakraborty, N., Ghosal, A.: Kinematics of wheeled mobile ro-
performance measurement of mobile manipulators. J. Stand. bots on uneven terrain. Mech. Mach. Theor. 39, 1273–1287 (2004)
Technol. 121, 342–366 (2016) 57. Ashmore, M., Barnes, N.: Omni-drive robot motion on curved
36. Yan, Z., Jouandeau, N., Cherif, A.A.: A survey and analysis of paths: the fastest path between two points is not a straight line.
multi-robot coordination. Int. J. Adv. Robotic Syst 10(12), 1–18 In: Proceedings of 15th Australian Joint Conference on Artificial
(2013) Intelligence: Advances in Artificial Intelligence (AI’02), pp. 225–
37. Bandettini, A., Luporini, F., Viglietta, G.: A survey of 236. Springer, London (2002)
open problems for mobile robots. Cornell University Library. 58. Huang, L., Lim, Y.S., Li, D., Teoh, C.E.L.: Design and analysis
arXiv:111.2259v1 [cs. RO] (2011) of four-wheel omnidirectional mobile robot. In: Proceedings of
38. Medeiros, A.A.D.: A survey of control architectures for autono- Second International Conference on Autonomous Robots and
mous mobile robots. J. Brazilian Comput. Soc. 4(3). (Online) (1998) Agents. Palmerston North, New Zealand (2004)
39. Chung, T.H., Hollinger, G.A., Isler, V.: Search and pursuit-evasion 59. Phairoh, T., Williamson, K.: Autonomous mobile robots using
in mobile robotics. Auton. Robots 31, 299–316 (2011) real time kinematic signal correction and global positioning robot
40. David, J., Manivanan, P.V.: Control of Truck-Trailer Mobile control. In: Proceedings of IAJC-IJME International Conference
Robots: a Survey. Intelligent Service Robotics. Springer, Berlin, on Engineering and Technology. Sheraton, Nashville, TN, Paper
(Online) (2014) 087/IT304 (2008)
41. Ivaldi, S., Padvis, V., Nori, F.: Tools for Dynamics Simulation
of Robots: a Survey Based on User Feedback. arXiv:1402.750v1
[cs.RO] (2014)
Mobile Robot Dynamics
42. Pomerleau, F., Colas, F., Siegwart, R.: A review of point cloud
registration algorithms for mobile robotics. Foundations and 60. Sidek, S.N.: Dynamic Modeling and Control of Nonholonomic
Trends in Robotics. Now Publishers 4(1), 1–104 (2015) Wheeled Mobile Robot Subjected to Wheel Slip. PhD Thesis,
Vanderbilt University, Nashville (2008)
61. Williams, R.L. ., Carter, B.E., Gallina, P., Rosati, G.: Dynamic
Journal Special Issues model with slip for wheeled omni-directional robots. IEEE Trans.
Robot. Autom. 18(3), 285–293 (2002)
43. Madhavan, R., Scrapper, C., Kleiner, A. (eds.): Special issue: 62. Moret, E.N.: Dynamic Modeling and Control of a Car-Like Robot.
characterizing mobile robot localization and mapping. Auton. MSc Thesis, Virginia Polytechnic Institute and State University.
Robot. 27(4), 431–448 (2009) Blacksburg (2003)
44. Petrovic, I. (ed.): Special issue: advanced mobile robotics. J. 63. Handy, A., Badreddin, E.: Dynamic modeling of a wheeled
Comput. Inform. Technol. 17(1), 1–120 (2009) mobile for identification, navigation and control. In: Proceedings
45. Tzafestas, S.G. (ed.): Special issue: autonomous mobile robots of IMACS Conference on Modeling and Control of Technological
in health care services. J. Intell. Robotic Syst. 22(3–4), 177–374 Systems, pp. 119–128, Lille (1992)
(1998) 64. Watanabe, K., Shiraishi, Y., Tzafestas, S.G., Tang, J., Fukuda,
46. Tzafestas, S.G. (ed.): Special issue: intelligent mobile robots. Adv. T.: Feedback control of an omnidirectional autonomous platform
Robot. 12(4), 313–481 (1998) for mobile service robots. J. Intell. Robot. Syst. 22, 315–330
47. Tzafestas, S.G. (ed.): Special issue: research on autonomous (1998)
robotic wheelchairs in Europe. IEEE Robot. Autom. Mag. 8(1), 65. Pin, F.G., Killough, S.M.: A new family of omnidirectional and
4–65 (2001) holonomic wheeled platforms for mobile robots. IEEE Trans.
48. Cetto, J.A., Freze, U., Tenorth, M. (eds.): Special issue: selected Robot. Autom. 10(4), 480–489 (1994)
papers from 6th European conference on mobile robots. Robot. 66. Moore, K.L., Flann, N.S.: A six-wheeled omnidirectional
Auton. Syst. 69, 1–98 (2015) autonomous mobile robot. IEEE Control Syst. Mag. 20(6), 53–66
49. Lopez-Nicolas, G., Mezouar, Y. (eds.): Special issue: visual cont- (2000)
rol of mobile robots. Robot. Auton. Syst. 62(11), 1611–1668 (2014)
50. Lee, S., Menegatti, E., Lee (eds.): Special issue: intelligent auto-
nomous systems. Robot. Auton. Syst. 62(11), 1669–1848 (2014) Mobile Robot Control: Standard Controllers
51. Nguyen, C.C., Zheng, Y.F. (eds.): Special issue: mobile robots. J.
Robot. Syst. (Currently: Journal of Field Robotics) 14(4), 229–340 67. Kanayama, Y., Kimura, Y., Noguchi, T.: A stable tracking control
(1997) method for a nonholonomic mobile robot. IEEE Trans. Robot.
52. Chrysostomou, D., Goher, K., Muscato, G., Tokhi, U., Virk, G.S. Autom. 7, 1236–1241 (1991)
(eds.): Special issue: real-world mobile robot systems. Int. J. 68. Tian, Y., Sidek, N., Sarkar, N.: Modeling and control of a non-
Indust. Robot. 44(4), 393–563 (2017) holonomic wheeled mobile robot with wheel slip dynamics. In:
Proceedings of IEEE Symposium on Computational Intelligence
in Control and Automation, pp. 7–14, Nashville (2009)
Mobile Robot Kinematics 69. Chang, C.F., Huang, C.I., Fu, L.C.: Nonlinear control of a wheeled
mobile robot with nonholonomic constraints. In: Proceedings
53. Muir, P.F., Neuman, C.P.: Kinematic modeling of wheeled mobile of 2004 IEEE International Conference on Systems, Man and
robots. J. Robot. Syst. 4(2), 281–329 (1987) Cybernetics, pp. 5404–5409, The Hague (2004)
54. Killough, S.M., Pin, F.G.: Design of an omnidirectional and 70. Ashoorizad, M., Barzamini, R., Afshar, A., Zouzdani, J.: Model
holonomic wheeled platform design. In: Proceedings of IEEE reference adaptive path following for wheeled mobile robots.
Conference on Robotics and Automation, pp. 84–90, Nice (1992) In: Proceedings of International Conference on Information and
55. Sidek, N., Sarkar, N.: Dynamic modeling and control of Automation (IEEE/ICIA’06), pp. 289–294, Colombo (2006)
nonholonomic mobile robot with lateral slip. In: Proceedings of 71. Watanabe, K., Shiraishi, Y., Tang, J., Fukuda, T., Tzafestas, S.G.:
Seventh WSEAS International Conference on Signal Processing Autonomous control for an omnidirectional mobile robot with
J Intell Robot Syst (2018) 91:35–58 55

feedback control system [Chapter 13]. In: Tzafestas, S.G. (ed.) In: Proceedings of International Conference on Information and
Advances in Intelligent Autonomous Systems, pp. 289–308. Automation, pp. 289–294, Colombo, (IEEE/ICIA2006) (2006)
Kluwer, Boston/Dordrecht (1999) 89. Alicja, M.: A new universal adaptive tracking control law for
72. Lee, S., Youm, Y., Chung, Y.: Control of car-like mobile nonholonomic wheeled mobile robots moving in R 3 space. In:
robots for posture stabilization. In: Proceedings of IEEE/RSJ Proceedings of IEEE International Conference Robotics and
International Conference on Intelligent Robots and Systems Automation. Detroit (1999)
(IROS’99), pp. 1745–1750, Kyongju (1999) 90. Pourboghrat, F., Karlsson, M.P.: Adaptive control of dynamic
73. Panimadai Ramaswamy, S.A., Balakrishnan, S.N.: Formation mobile robots with nonholonomic constraints. Comput. Elect.
control of car-like mobile robots: a Lyapunov function based Eng. 28, 241–253 (2002)
approach. In: Proceedings of 2008 American Control Conference, 91. Gholipour, A., Dehghan, S.M., Ahmadabadi, M.N.: Lyapunov-
Seattle (2008) based tracking control of non-holonomic mobile robot. In:
Proceedings of 10th Iranian Conference on Electrical Engineering,
vol. 3, pp. 262–269, Tabriz (2002)
Mobile Robot Control: Advanced Controllers
92. Fetter Lages, W., Hemerly, E.M.: Adaptive liberalizing control
of mobile robots. In: Kopacek, P., Pereira, C.E. (eds.) Intelligent
74. Isidori, A.: Nonlinear Control Systems: an Introduction. Springer, Manufacturing Systems-a Volume from IFAC Workshop, pp. 23–
Berlin/New York (1985) 29. Pergamon Press, Gramado (1998)
75. Stotine, J.J., Li, W.: Applied Nonlinear Control. Prentice Hall, 93. Landau, Y.D.: Adaptive Control: The Model Reference Approach.
Engewood Cliffs (1991) Marcel Dekker, New York (1979)
76. Brockett, R.W.: Asymptotic Stability and Feedback Stabilization: 94. Zhang, Y., Hong, D., Chung, J.H., Velinsky, S.: Dynamic
Differential Geometric Control Theory. Birkhauser, Boston model-based robust tracking control of differentially steered
(1983) wheeled mobile robot. In: Proceedings of the American Control
77. Yun, X., Yamamoto, Y.: On Feedback Linearization of Mobile Conference, pp. 850–855, Philadelphia (1988)
Robots Technical Report (CIS). University of Pennsylvania, 95. Yang, J.M., Kim, J.K.: Sliding mode control for trajectory tracking
Department of Computer and Information Science (1992) of nonholonomic wheeled mobile robots. IEEE Trans. Robot.
78. Yang, E., Gu, D., Mita, T., Hu, H.: Nonlinear tracking control Autom. 15(3), 578–587 (1999)
of a car-like mobile robot via dynamic feedback linearization. 96. Chwa, D.: Sliding-mode tracking control of nonholonomic
In: Proceedings of Control 2004. University of Bath, UK, [paper wheeled mobile robots in polar coordinates. IEEE Trans. Control
1D-218] (2004) Syst. Technol. 12(4), 637–644 (2004)
79. DeVon, D., Brett, T.: Kinematic and dynamic control of a
wheeled mobile robot. In: Proceedings of IEEE/RSJ International
Conference on Intelligent Robots and Systems, pp. 2125–2126, Mobile Robot Fuzzy and Neural Control
San Diego (2007)
80. Astolfi, A.: Exponential stabilization of a wheeled mobile robot 97. Zadeh, L.A.: Fuzzy sets. Inf. Control 8, 338–353 (1965)
via discontinuous control. J. Dyn. Syst. Meas. Control 121, 121– 98. Kosko, B.: Neural Networks and Fuzzy Systems: a Dynamical
126 (1999) System Approach to Machine Intelligence. Prentice Hall, Engle-
81. Reyhanoglu, M.: On the stabilization of a class of nonholonomic wood Cliffs (1992)
systems using invariant manifold technique. In: Proceedings of the 99. Haykin, S.: Neural Network: a Comprehensive Foundation.
34th IEEE Conference on Decision and Control, pp. 2125–2126, Macmillan College Publishing. Upper Saddle River, New Jersey
New Orlean (1995) (1994)
82. Tayebi, A., Tadijne, M., Rachid, A.: Invariant manifold approach 100. Tzafestas, S.G. (ed.): Soft Computing and Control Technology.
for the stabilization of nonholonomic chained systems: application World Scientific Publishers, Singapore/London (1997)
to a mobile robot. Nonlinear Dyn. 24, 167–181 (2001) 101. Omatu, S., Khalid, M., Yusof, R.: Neuro-Control and Its
83. Watanabe, K., Yamamoto, K., Izumi, K., Maeyama, S.: Underac- Applications. Springer, London/Berlin (1996)
tuated control for nonholonomic mobile robots by using double 102. Das, T., Narayan, K.I.: Design and implementation of a adaptive
integrator model and invariant manifold theory. In: Proceedings fuzzy logic-based controller for wheeled mobile robots. IEEE
of IEEE/RSJ International Conference on Intelligent Robots and Trans. Control Syst. Technol. 14(3), 501–510 (2006)
Systems, pp. 2862–2867, Taipei (2010) 103. Castillo, O., Aguilar, L.T., Cardenas, S.: Fuzzy logic tracking
84. Peng, Y., Liu, M., Tang, Z., Xie, S., Luo, J.: Geometry stabilizing control for unicycle mobile robots. Eng. Lett. 13(2), EL.13-2-
control of the extended nonholonomic double integrator. In: 4:73-77 (2006)
Proceedings of IEEE International Conference on Robotics and 104. Driesen, B.J., Feddema, J.T., Kwok, K.S.: Decentralized fuzzy
Biometrics, pp. 926–931. Tianijn (2010) control of multiple nonholonomic vehicles. J. Intell. Robot. Syst.
85. Izumi, K., Watanabe, K.: Switching manifold control for an 26, 65–78 (1999)
extended nonholonomic double integrator. In: Proceedings of 105. Moustris, G., Tzafestas, S.G.: A robust fuzzy logic path tracker
International Conference on Control and Automation Systems, for non-holonomic mobile robots. J. Artif. Intell. Tools 14(6),
pp. 896–899. Kintex, Gyeonggi-do (2010) 935–965 (2005)
106. Tzafestas, S.G., Deliparashos, K.M., Moustris, G.P.: Fuzzy logic
path tracking control for autonomous non-holonomic robots: design
Mobile Robot Adaptive and Robust Control of system on a chip. Robot. Auton. Syst. 58, 1017–1027 (2010)
107. Moustris, G.P., Deliparashos, K.M., Tzafestas, S.G.: Feedback
86. Marino, R., Tomei, P.: Nonlinear Control Design: Geometric, equivalence and control of mobile robots through a scalable FPGA
Adaptive and Robust. Prentice Hall, Upper River (1995) architecture. In: Veleninov Topalov, A. (ed.) Recent Advances in
87. Marino, R.: Adaptive control of nonlinear systems: basic results Mobile Robotics. InTech (2011). www.interchopen.com/books
and applications. IFAC-Rev. Control 21, 55–66 (1997) 108. Rigatos, G.G., Tzafestas, C.S., Tzafestas, S.G.: Mobile robot
88. Ashoorizad, M., Barzarnimi, R., Afshar, A., Jouzdani, J.: Model motion control in partially unknown environments using a sliding-
reference adaptive path following for wheeled mobile robots. mode fuzzy-logic controller. Robot. Auton. Syst. 33, 1–11 (2000)
56 J Intell Robot Syst (2018) 91:35–58

109. Moustris, G., Tzafestas, S.G.: Switching fuzzy tracking control 126. Okamoto, J. Jr., Grassi, V. Jr.: Visual servo control of a
for the Dubins car. Control Eng. Prac. 19(1), 45–53 (2011) mobile robot using omnidirectional vision. In: Proceedings of
110. Rigatos, G.G., Tzafestas, S.G., Evangelidis, G.J.: Reactive Mechatronics Conference 2002, pp. 413–422. University of
parking control of nonholonomic vehicles via a fuzzy learning Twente (2002)
automation. IEE Proc. Control Theory 148(2), 169–179 (2001) 127. Gong, X.: Omnidirectional Vision for an Autonomous Surface
111. Lewis, F.L., Campos, J., Selmic, R.: Neuro-Fuzzy Control Vehicle [Ph.D. dissertation]. Virginia Polytechnic Institute and
of Industrial Systems with Actuator Nonlinearities. SIAM, State University (2008)
Philadelphia (2002) 128. Russel, S., Norwig, P.: Artificial Intelligence. Prentice Hall,
112. Fierro, R., Lewis, F.L.: Control of a nonholonomic mobile robot: Upper Saddle River (2003)
Backstepping kinematics into dynamics. J. Robot Syst. 14(3),
149–163 (1997) Mobile Robot Path and Motion Planning

Mobile Robot Vision-Based Control 129. Kumar, V., Zefran, M., Ostrowski, J.: Motion planning and
control of robots. In: Nof, S. (ed.) Handbook of Industrial
113. Haralik, R.M., Shapiro, L.G.: Computer and Robot Vision. Robotics, pp. 295–315. New York, Wiley (1999)
Addison Wesley, Reading (1993) 130. Lozano-Perez, T.: Spatial planning: a configuration space
114. Corke, P.: Visual control of robot manipulators: a review. In: approach. IEEE Trans. Comput. 32(2), 108–120 (1983)
Hashimoto, K. (ed.) Visual Serving, pp. 1–31. Word Scientific, 131. Erdmann, M., Lozano-Perez, T.: On multiple moving obstacles.
Singapore (1993) Algorithmica 2(4), 477–521 (1987)
132. Fugimura, K.: Motion Planning in Dynamic Environments.
115. Tsakiris, D., Samson, C., Rives, P.: Vision-based time–varying
Springer, Berlin/Tokyo (1991)
stabilization of a mobile manipulator. In: Proceedings of Fourth
133. Khatib, O.: Real-time obstacle avoidance for manipulators and
International Conference on Control, Automation, Robotics and
mobile robots. Int. J. Robot. Res. 5(1), 90–98 (1986)
Vision (ICARV’96). Westin Stamford, Singapore (1996)
134. Sheu, P.C.Y., Xue, Q.: Intelligent Robotic Systems. World
116. Cherubini, A., Chaumette, F., Oriolo, G.: A position-based
Scientific Publishers, Singapore/London (1993)
visual servoing scheme for following paths with nonholonomic 135. Gallina, P., Gasparetto, A.: A technique to analytically formulate
robots. In: Proceedings of IEEE/RSJ International Conference on and solve the 2-dimensional constrained trajectory planning for a
Intelligent Robots and Systems (IROS 2008), pp. 1648–1654, Nice mobile robot. J. Intell. Robot. Syst. 27(3), 237–262 (2000)
(2008) 136. Hatzivasiliou, F.V., Tzafestas, S.G.: A path planning method
117. Burshka, D., Hager, G.: Vision-based control of mobile robots. for mobile robots in a structured environment. In: Tzafestas,
In: Proceedings of 2001 IEEE International Conference on S.G. (ed.) Robotic Systems: Advanced Techniques and Applica-
Robotics and Automation, pp. 1707–1713, Seoul (2001) tions. Kluwer, Dordrecht/Boston (1992)
118. Cherubini, A., Chaumette, F., Oriolo, G.: An image-based visual 137. Garcia, E., De Santos, P.G.: Mobile-robot navigation with
servoing scheme for following paths with nonholonomic mobile complete coverage of unstructured environment. Robot. Auton.
robots. In: Proceedings of International Conference on Control Syst. 46, 195–204 (2004)
Automation, Robotics and Vision, pp. 108–113, Hanoi (2008) 138. Safadi, H.: Local path planning using virtual potential field.
119. Das, A.K., Fierro, R., Kumar, V., Southall, B., Spletzer, J., Report COMP 765: Spatial Representation and Mobile Robotics-
Taylor, C.J.: Real-time vision-based control of a nonholonomic Project. School of Computer Science, McGill University, Canada
mobile robot. In: Proceedings of 2001 IEEE International Confe- (2007)
rence on Robotics and Automation, pp. 1714–1719, Seoul (2001) 139. Ding, F.G., Jiang, P., Bian, X.Q., Wang, H.J.: AUV local path
120. Carelli, R., Soria, C.M., Morrales, B.: Vision-based tracking planning based on virtual potential field. In: Proceedings of IEEE
control for mobile robots. In: Proceedings of Twelfth International International Conference on Mechatronics and Automation, vol. 4,
Conference on Advanced Robotics (ICAR’05), pp. 148–152, pp. 1711–1716, Niagara Falls (2005)
Seattle (2005) 140. Koren, Y., Borenstein, J.: Potential field methods and their
121. Maya-Mendez, M., Morin, P., Samson, C.: Control of a inherent liminations for mobile robot navigation. In: Proceedings
nonholonomic mobile robot via sensor-based target tracking of IEEE Conference on Robotics and Automation, pp. 1398–1404,
and pose estimation. In: Proceedings of IEEE/RSJ International Sacramento (2005)
Conference on Intelligent Robots and Systems, pp. 5612–5618, 141. Borenstein, J., Koren, Y.: The vector field histogram: fast
Beijing (2009) obstacle avoidance for mobile robots. IEEE J. Robot. Autom. 7(3),
122. Abdelkader, H.H., Mezouar, Y., Andreff, N., Martinet, P.: 278–288 (1991)
Image-based control of mobile robot with central catadioptric 142. Wang, L.C., Yong, L.S., Ang Jr. M.R.: Hybrid of global path
cameras. In: Proceedings of 2005 IEEE International Conference planning and local navigation implemented on a mobile robot
on Robotics and Automation, pp. 3533–3538, Barcelona (2005) in indoor environment. In: Proceedings of IEEE International
123. Qian, J., Su, J.: Online estimation of image Jacobian matrix Symposium on Intelligent Control, pp. 821–826 (2002)
by Kalman-Bucy filter for uncalibrated stereo vision feedback. 143. Garrido, S., Moreno, L., Blanco, D., Jurewicz, P.: Path planning
In: Proceedings of International Conference on Robotics and for mobile robot navigation using Voronoi diagram and fast
Automation (ICRA 2002), pp. 562–567 (2002) marching. Int. J. Robot. Autom. 2(1), 42–64 (2011)
124. Gans, N.R., Hutchinson, S.A.: A stable vision-based control 144. Garrido, S., Moreno, L., Blanco, D.: Exploration of a cluttered
scheme for nonholonomic vehicles to keep a landmark in the field environment using Voronoi transform and fast marching. Robot.
of view. In: Proceedings of 2007 International Conference on Auton. Syst. 56(12), 1069–1081 (2008)
Robotics and Automation, pp. 2196–2200, Rome (2007) 145. Arney, T.: An efficient solution to autonomous path planning by
125. Lopez-Nicolas, G., Gans, N.R., Bhattacharya, S., Sagues, approximate cell decomposition. In: Proceedings of International
C., Guerrero, J.J., Hutchinson, S.: Homography-based control Conference on Information and Automation for Sustainability
scheme for mobile robots with nonholonomic and field-of-view (ICIAFS 07), pp. 88–93, Colombo (2007)
constraints. IEEE Trans. Syst. Man Cybern. Cybern 40(4), 1115– 146. Katevas, N.I., Tzafestas, S.G., Pnevmatikatos, C.G.: The
1127 (2010) approximate cell decomposition with local node refinement
J Intell Robot Syst (2018) 91:35–58 57

global path planning method: path nodes refinement and curve 168. Crisan, D., Doucet, A.: A survey of convergence results on
parametric interpolation. J. Intell. Robot. Syst. 22, 289–314 (1998) particle filtering methods for practitioners. IEEE Trans. Signal
147. Olunloyo, V.O.S., Ayomoh, M.K.O.: Autonomous mobile robot Process. 50(3), 736–746 (2002)
navigation using hybrid virtual force field concept. Eur. J. Sci. Res. 169. Rituerto, A., Puig, L., Guerrero, J.J.: Visual SLAM with an
31(2), 204–228 (2009) omnidirectional camera. In: Proceedings of Twentieth Interna-
148. Katevas, N.I., Tzafestas, S.G., Matia, F.: Global and local tional Conference on Pattern Recognition (ICPR’10), pp. 348–
strategies for mobile robot navigation. In: Katevas, N. (ed.) Mobile 351, Istanbul (2010)
Robotics in Healthcare. IOS Press, Amsterdam (2001) 170. Rigatos, G.G., Tzafestas, S.G.: Extended Kalman filtering for
149. Katevas, N.I., Tzafestas, S.G.: The active kinematic histogram fuzzy modeling and multisensory fusion. Math. Comput. Model.
method for path planning of non-point non-holonomically Dyn. Syst. 13(3), 251–266 (2007)
constrained mobile robots. Adv. Robot. 12(4), 375–395 (1998) 171. Kim, J.M., Chung, M.J.: SLAM with omnidirectional stereo
150. Zelinsky, A., Jarvis, R.A., Byrne, J.C., Yuta, S.: Planning paths vision sensor. In: Proceedings of 2003 IEEE/RSJ International
of complete coverage of an unstructured environment by a mobile Conference on Intelligent Robots and Systems, pp. 442–447, Las
robot. In: Proceedings of International Symposium on Advanced Vegas (2003)
Robotics, Tokyo (1993)
151. Zelinsky, A., Yuta, S.: A unified approach to planning, sensing Mobile Robot Intelligent Control and Software Architectures
and navigation for mobile robots. In: Proceedings of International
Symposium on Experimental Robotics, Tokyo (1993)
172. Saridis, G.N.: Toward the realization of intelligent controls. Proc.
152. Choi, Y.H., Lee, T.K., Baek, S.-H., Oh, S.-Y.: Online complete
IEEE 8, 1115–1133 (1979)
coverage path planning for mobile robots based on linked
173. Saridis, G.N.: Foundations of intelligent controls. In: Proceedings
spiral paths using constrained inverse distance transform. In:
of IEEE Workshop on Intelligent Control, pp. 23–28, Troy (1985)
Proceedings of IEEE/RSJ International Conference on Intelligent
174. Meystel, A.M.: Architectures of intelligent control: The science
Robots and Systems, pp. 5688–5712, St. Louis (2009)
of autonomous intelligence. In: Proceedings of IEEE International
Symposium on Intelligent Control, pp. 42–48, Chicago (1993)
Mobile Robot Localization and Mapping 175. Meystel, A.M.: Multidirectional hierarchical decision support
systems. IEEE Trans. Syst. Man Cybern.-Part C SMR-AR 33, 86–
101 (2003)
153. Papoulis, A.: Probability, Random Variables and Stochastic 176. Albus, J.S., Quintero, R.: Towards a reference model architecture
Processes. McGraw-Hill, New York (1965) for real-time intelligent control systems. In: Robotics and
154. Meditch, J.S.: Stochastic Optimal Linear Estimation and Control. Manufacturing, vol. 3. ASME, New York (1990)
McGraw-Hill, New York (1969) 177. Albus, J.M.: Outline for a theory of intelligence. IEEE Trans.
155. Anderson, B.D.O., Moore, J.B.: Optimal Filtering. Englewood Syst. Man Cybern. SMC-21(3), 473–509 (1991)
Cliffs, Prentice Hall (1979) 178. Brooks, R.A.: A robust layered control system for a mobile robot.
156. Borenstein, J., Everett, H.R., Feng, L.: Navigating Mobile Ro- IEEE J. Robot. Autom. RA-2, 14–23 (1986)
bots: Sensors and Techniques. A.K. Peters Ltd, Wellesley (1999) 179. Brooks, R.A.: Intelligence Without Reason. AI Memo, No. 1293,
157. Adams, M.D.: Sensor Modeling Design and Data Processing for AI Laboratory, MIT (1991)
Automation Navigation. World Scientific, Singapore (1999) 180. Arkin, R.C.: Motor schema-based mobile robot navigation. Int.
158. Davis, E.R.: Machine Vision: Theory, Algorithms, Practicalities. J. Robot Res. 8(4), 92–112 (1989)
Morgan Kaufmann, San Francisco (2005) 181. Arkin, R.C.: Cooperation without communication: multi-agent
159. Kleeman, L.: Advanced sonar and odometry error modeling for schema based robot navigation. J. Robot. Syst. 9(2), 351–364
simultaneous localization and map building. In: Proceedings of (1992)
the 2004 IEEE/RSJ International Conference on Intelligent Robots 182. Simmons, R.G.: Concurrent planning and execution for
and Systems, pp. 1866–1871, Sendai (2004) autonomous robots. IEEE Control Syst. Mag. 12(1), 46–50 (1992)
160. Betke, M., Gurvis, L.: Mobile robot localization using land- 183. Simmons, R.G.: Monitoring and error recovery for autonomous
marks. IEEE Trans. Robot. Autom. 13(2), 251–263 (1997) walking. In: Proceedings of IEEE International Conference on
161. Hu, H., Gu, D.: Landmark-based navigation of industrial mobile Intelligent Robots and Systems (IROS’92). Raleigh (1992)
robots. Int. J. Indust. Robot. 27(6), 458–467 (2000) 184. Coste-Maniere, E., Simmons, R.: Architecture: The backbone
162. Andersen, C.S., Concalves, J.G.M.: Determining the Pose of a of robotic systems. In: Proceedings of IEEE International
Mobile Robot Using Triangulation: a Vision Based Approach. Conference on Robotics and Automation (ICRA’2000), pp. 67–
Technical Report No. 1. European Union Joint Research Center 72, San Francisco (2000)
(1995) 185. Xie, W., Ma, J., Yang, M.: Research on classification of
163. Castellanos, J.A., Tardos, J.D.: Mobile Robot Localization and intelligent robotic architecture. J. Comput. 7(2), 450–457 (2012)
Map Building: a Multisensory Fusion Approach. Springer, Berlin 186. Saridis, G.N.: Analytical formulation of the principle of
(1999) increasing precision and decreasing intelligence for intelligent
164. Guivant, J.E., Nebot, E.M.: Optimization of the simultaneous machines. Automatica 25(3), 461–467 (1989)
localization and map-building algorithm for real-time implemen- 187. Hayes-Roth, B.: A blackboard architecture for control. Artif.
tation. IEEE Trans. Robot. Autom. 17(3), 242–257 (2001) Intell. 26(3), 251–321 (1985)
165. Rekleitis, I., Dudek, G., Milios, E.: Probabilistic cooperative 188. Oreback, A., Christensen, H.L.: Evaluation of architectures for
localization and mapping in practice. Proc. IEEE Robot. Autom. mobile robots. Auton. Robot. 14, 33–49 (2003)
Conf. 2, 1907–1912 (2003) 189. Jawari, D., Deris, S., Mamat, R.: Software reuse for mobile
166. Bailey, T., Durrant-Whyte, H.: Simultaneous localization and robot applications through analysis patterns. Int. Arab J. Inform.
mapping (SLAM). Part I. IEE Robot. Autom. Mag. 13(2), 99–110 Technol. 4(3), 220–228 (2007)
(2006). Part II, ibid, (3) 108–117 190. Canas, J.M., Ruiz-Ayucar, J., Aguero, C., Martin, F.: Jde-neoc:
167. Rekleitis, I., Dudek, G., Milios, E.: Multirobot collaboration for component oriented software architecture for robotics. J. Phys.
robust exploration. Ann. Math. Artif. Intell. 31(1–4), 7–40 (2001) Agents 1(1), 1–6 (2007)
58 J Intell Robot Syst (2018) 91:35–58

191. Kerry, M.: Simplifying Robot Software Design Layer by Layer. 209. Akkizidis, I.S., Tzafestas, S.G.: Navigation of a mobile robot
National Instruments RTC Magazine. http://rtcmagazine.com/ using fuzzy clustering techniques. In: Proceedings of 11th
articles/view/102283 (2013) International Workshop on Robotics in Alpe-Adria-Danube
192. Katevas, N.I., Tzafestas, S.G., Koutsiouris, D.G., Pneumatikatos, Region. Balatonfared (2002)
C.G.: The SENARIO autonomous navigation system. In: Tzafes- 210. Tzafestas, C.S., Tzafestas, S.G.: Full-state modeling, motion
tas, S.G. (ed.) Proceedings 1st MobiNet Symposium on Mobile planning and control of mobile manipulators. Stud. Inf. Control
Robotics Technology for Health Care Services, pp. 87–89, Athens 10(2), 109–127 (2001)
(1997) 211. Tzafestas, S.G., Melfi, A., Krikochoritis, A.: Omnidirectional
193. Laengle, T., Lueth, T.C., Remboldt, U., Woern, H.: A distributed mobile manipulator modeling and control: analysis and simula-
control architecture for autonomous mobile robots: implementa- tion. Syst. Anal. Model. Control 40, 329–364 (2001)
tion of the Karlsruhe multi-agent robot architecture. Adv. Robot. 212. Moustris, G.P., Deliparaschos, K.M., Tzafestas, S.G.: Tracking
12(4), 411–431 (1998) control using the strip-wise affine transformation: an experimental
194. Fischer, C., Schmidt, G.: Multi-modal human-robot interface SoC design. In: Proceedings of EUCA European Control
for interaction with a remotely operating mobile service robot Conference. Budapest, Hungary, [paper MoC3.5] (2009)
(ROMAN). Adv. Robot. 12(4), 375–395 (1998)
195. Tzafestas, S.G.: Sociorobot World: A guided Tour for All.
Springer, Berlin (2016)
Spyros G. Tzafestas was born in Corfu, Greece on December 3, 1939
Point Stabilization, Path Planning/Following and Trajectory He earned a B.Sc. in Physics (1962) and a Postgraduate Diploma
in Electronics and Communications (1963) from Athens University,
Tracking Control D.I.C in Electrical Engineering-Automatic Control Systems from
Imperial College of London, M. Sc. (Eng.) in Control from London
196. De Wit, C.C., Sordalen, O.: Exponential stabilization of mobile University in 1967, and Ph.D. in Systems and Control from
robots with nonholonomic constraints. IEEE Trans. Autom. Southampton University of the United Kingdom in March 1969.
Control 37(11), 1791–1797 (1992) He worked at ‘Demokritos’ Nuclear Research Center as a research
197. Do, K., Jiang, Z., Pan, J.: Simultaneous tracking and stabilization leader of the Computer Control and Automation Group of the
of mobile robots: An adaptive approach. IEEE Trans. Autom. Computer Science Division from April 1969 to September 1973. From
Control 49(7), 1147–1151 (2004) October 1973 to March 1985 he served as a Professor of Control
198. Jiang, Z.-P., Nijmeijer, H.: Tracking control of mobile robots: a Systems, Director of the Automatic Control Systems Laboratory, at the
case study in backstepping. Automatica 33, 1393–1399 (1997) Electrical Engineering Department of the University of Patras, Patras,
199. Kostic, D., Adinandra, S., Caarls, J., Nijmeijer, H.: Collision-free Greece. He then gained a position at the School of Electrical and
tracking control of unicycle mobile robots. In: Proceedings of the Computer Engineering of the National Technical University of Athens
IEEE Conference on Decision and Control, pp. 5667–5672 (2009) (NTUA) where he served as Professor of Control and Robotics. He
200. Kant, K., Zuckler, S.: Toward efficient trajectory planning: the founded the Intelligent Robotics and Automation Laboratory of which
path-velocity decomposition. Int. J. Robot. Res. 5, 72–89 (1986) was the Director up to his retirement on August 31, 2006. From 1999
201. Tzafestas, S.G., Stamou, G.: A fuzzy path planning algorithm for to 2009 he was the Director of the NTUA Institute of Communication
autonomous robots moving in an unknown and uncertain environ- and Computer Systems (ICCS). He holds a D.Sc. (Southampton
ment. In: Proceedings of European Robotics and Intelligent Sys- University, 1978) and two honorary doctorates in engineering (Dr.-Ing.
tems Conference (EURISCON’94), pp. 140–149, Malaga (1994) Eh. ,TU Munich, Germany, 1997, and Docteur (Honoris Causa), EC
202. Katevas, N., Sgouros, N.-M., Tzafestas, S.G., Papakonstantinou, Lille France, 2003).
G., Beatie, G., Bishop, G., Tsanakas, P., Koutsouris, D.G.: The Over the years he worked in the following areas: 1. Distributed
autonomous mobile robot SENARIO: a sensor-aided intelligent parameter control systems, 2. Computing and micro-processor sys-
navigation system for powered wheel chairs. IEEE Robot. Autom. tems, 3. Multivariable and multidimensional systems, 4. Walsh series-
Mag. 4(4), 60–70 (1997) based modeling and control, 5. Robotic systems, 6. AI, KB, and
203. Samson, C.: Control of chained systems application to path computational intelligence control techniques, 7. Applications to
following time-varying point-stabilization of mobile robots. IEEE nuclear reactors and manufacturing systems, and 8. Roboethics/
Trans. Autom. Control 40(1), 64–77 (1995) robophilosophy. He is known for important contributions in the above
204. Aguilar, A., Hespanha, J.: Trajectory tracking and path following areas. He is a LIFE FELLOW of IEEE and a FELLOW of IET (IEE).
of underactuated autonomous vehicles with parametric modeling He served as the founding editor of the Journal of Intelligent
uncertainty. IEEE Trans. Autom. Control 52(8), 1362–1379 (2007) and Robotic Systems (1988-2006), and he is the Chief Editor of the
205. Coelho, P., Nunes, U.: Path following control of mobile robots in Springer book series on Intelligent Systems, Control and Automation.
the presence of uncertainties. IEEE Trans. Robot. 21(2), 252–261 He has edited 30 research books, 20 Conference Proceedings, and
(2005) 26 journal special issues in his fields of expertise. He has organized
206. Deliparaschos, K.M., Moustris, G.P., Tzafestas, S.G.: and/or chaired many international conferences (IEEE CDC, EUCA,
Autonomous SoC for fuzzy robot path tracking. In: Proceedings IMACS, IASTED, etc.). He has served as President of EUCA and
of EUCA European Control Conference (ECC’2007), pp. 5471– Vice President of IMACS, and has been the scientific coordinator of
5478. Kos (2007) many national and European projects in IT, CIM, Robotics, Intelligent
207. Skoundrianos, E.N., Tzafestas, S.G.: Mobile robot modeling Systems, and Control. He is the author of seven international
using local model networks. In: Proceedings of EUCA European books and seven Greek books on automation, control, robotics and
Control Conference (ECC’2003), Cambridge (2003) Artificial/Computational Intelligence. He has received many world-
208. Zavlangas, P.G., Tzafestas, S.G.: Hierarchical motion control wide scientific awards and his biography is included in more than 20
system for mobile robot path planning and navigation. In: Pro- international biographical volumes. Currently, Dr. Tzafestas continues
ceedings of 2002 Japan-USA Symposium on Flexible Automation his scientific work at NTUA as a Professor Emeritus-Senior Research
(JUSFA’2002). Tokyo (2002) Associate of the School of Electrical and Computer Engineering.