Neural Networks for Mobile Robot Navigation:

A Survey

An-Min Zou, Zeng-Guang Hou, Si-Yao Fu, and Min Tan

Key Laboratory of Complex Systems and Intelligence Science,

Institute of Automation, The Chinese Academy of Sciences,
P.O. Box 2728, Beijing 100080, China
{anmin.zou, zengguang.hou, siyao.fu, min.tan}@ia.ac.cn

Abstract. Nowadays, mobile robots have attracted more and more at-
tention from researchers due to their extensive applications. Mobile
robots need to have the capabilities of autonomy and intelligence, and
they pose a challenge to researchers, which is to design algorithms that
allow the robots to function autonomously in unstructured, dynamic,
partially observable, and uncertain environments [1]. Navigation is the
key to the relative technologies of mobile robots and neural networks are
widely used in the field of mobile robot navigation due to their prop-
erties such as nonlinear mapping, ability to learn from examples, good
generalization performance, massively parallel processing, and capability
to approximate an arbitrary function given sufficient number of neurons.
This paper surveys the developments in the last few years of the neural
networks with applications to mobile robot navigation.

1 Introduction
In recent years, mobile robots have attracted more and more attention from
researchers since they have been widely used in various fields, such as space
exploration, under water survey, industrial and military industries, and service
and medical applications, and so on. The robots need to have the capabilities
of autonomy and intelligence, and they force the researches to deal with key
issues such as uncertainty (in both sensing and action), reliability, and real-time
response [2]. Therefore, a key challenge in robotics is to design algorithms that
allow the robots to function autonomously in unstructured, dynamic, partially
observable, and uncertain environments [1].
The problem of mobile robot navigation, which includes four fundamental
matters of mobile robots: map building, localization, path planning, and obsta-
cle avoidance, refers to plan a path with obstacle avoidance to a specified goal
and to execute this plan based on sensor readings, and is the key to the robot
to perform some designated tasks. Neural networks, motivated by how the hu-
man brain works, are increasingly being employed in various fields, including
signal processing, pattern recognition, medicine, speech production and recogni-
tion, and business. In the last few years, neural networks including feedforward
neural network, self-organizing neural network, principal component analysis

J. Wang et al. (Eds.): ISNN 2006, LNCS 3972, pp. 1218–1226, 2006.
c Springer-Verlag Berlin Heidelberg 2006
 Neural Networks for Mobile Robot Navigation: A Survey 1219

(PCA), dynamic neural network, support vector machines (SVM), neuro-fuzzy

approach, etc., have been widely used in the field of mobile robot navigation due
to their properties such as nonlinear mapping, ability to learn from examples,
good generalization performance, massively parallel processing, and capability
to approximate an arbitrary function given sufficient number of neurons. The
objective of this paper is to present the status of the applications of neural net-
works to mobile robot navigation. The rest of this paper is organized as follows:
The methods of neural networks and hybrid approaches for mobile robot navi-
gation are described in Sections 2 and 3, respectively, and conclusions are given
in Section 4.

2 Neural Networks for Mobile Robot Navigation

2.1 Neural Networks for Interpretation of the Sensor Data
Sensors are necessary for a robot to know where it is or how it got there, or to be
able to reason about where it has gone. The sensors can be flexible and mobile
to measure the distance that wheels have traveled along the ground, to measure
inertial changes and external structure in the environment. The sensors may be
roughly divided into two classes: internal state sensors, such as accelerometers,
gyroscope, which provide the internal information about the robot’s movements,
and external state sensors, such as laser, infrared sensors, sonar, and visual
sensors, which provide the external information about the environment. The
data from internal state sensors may provide position estimates of the robot in
a 2D space; however, cumulative error is inevitable.The data from external state
sensors may be used to directly recognize a place or a situation, or be converted to
information in a map of the environment. The laser, infrared, and sonar sensors
can provide distant and directional information about an object. Visual sensors
can obtain rich-information of the environment that can be very expensive to
process. In most cases, the sensor readings are imprecise and unreliable due to the
noises. Therefore, it is important for the mobile robot navigation to process the
sensor data with noises. Since neural networks have many processing nodes, each
with primarily local connections, they may provide some degree of robustness or
fault tolerance for interpretation of the sensor data.
Feedforward multi-layer perception neural network, which is trained by the
back-propagation algorithm, has been used for classification, recognition and
function approximation. Kim et al. proposed an approach to build environmental
map where the nonlinearity error of range data from infrared sensors was cor-
rected by using a feedforward neural network [3]. Thrun has used a feedforward
neural network to “translate” the sonar readings of sonar sensors into occupancy
values of each grid cell for building metric maps [4]. Meng and Kak presented
a NEURO-NAV system for mobile robot navigation [5]. In the NEURO-NAV, a
feedforward neural network, which is driven by the cells of the Hough transfor-
mation of the corridor guidelines in the camera image, is employed to obtain the
approximate relative angles between the heading direction of the robot and the
orientation of the hallway in order to drive the robot to move in the middle of
1220 A.-M. Zou et al.

the hallway [5]. In [6], a neural network based camera calibration method was
presented for the global localization of mobile robots using monocular vision.
Since every type of sensors have their own limitations for collecting the environ-
mental information of a robot, sensor fusion is necessary for the mobile robot
navigation. A sonar and infrared sensors fusion algorithm based on a feedforward
neural network to obtain reliable data is studied in [7].
The self-organizing Kohonen neural network is known for its ability to per-
form classification, recognition, data compression and association in an unsu-
pervised manner [8]. The self-organizing Kohonen neural networks are employed
to recognition the landmarks using the measurements from laser sensors in or-
der to provide coordinates of the landmarks for triangulation in [9]. Janet et al.
proposed a global localization algorithm using self-organizing Kohonen neural
networks [10]. By using the self-organizing Kohonen neural networks, the robot
can determine the room where it is according to the sonar sensor data.
PCA, which has been applied to data compression, pattern recognition, and
so on, is a statistical technique and is well known as one of the effective methods
to extract the principal features from high-dimension data and decrease the
dimension of the data. Crowley et al. presented an approach to estimate position
of a mobile robot based on PCA of laser ranger sensor data [11]. Vlassis et al.
proposed an approach for mobile robot localization where PCA was employed
to decrease the dimensions of sonar sensor data [12]. PCA has been used to
extract features of images for mobile robot localization in [13], [14]. Though
PCA is an appropriate model for data generated by a Gaussian distribution, or
data best described by a second order correlation; however, the distribution of
natural images is highly non-Gaussian. Therefore, kernel PCA is used to extract
features from image for mobile robot localization [15]. In the work reported in
[15], kernel PCA has a higher localization rate than that of conventional PCA,
whereas, the conventional PCA is faster than the kernel PCA.
SVM, which was proposed by Vapnik, is based on the statistical learning
theory [16]. In [17], seat numbers were used as landmarks for mobile robot local-
ization because the seat number could be employed to distinguish the landmarks,
and SVM was adopted to segment number regions from images.
Hopfield neural network can be used as associative memory or to solve
optimization problems [8]. In [18], [19], an improved neural network based on
Lagrange programming method was presented for hierarchical optimization of
nonlinear large-scale systems. Djekoune and Achour proposed a localization al-
gorithm using stereo vision where the correspondence problem for a set of seg-
ments extracted from a pair of stereo images is formulated as minimization of a
cost function that is performed by means of a two-dimensional Hopfield neural
network [20].

2.2 Neural Networks for Obstacle Avoidance

In the environment, there are always static and non-static obstacles. Therefore,
the robots need to autonomously navigate themselves in the environment with
obstacle avoidance. The neural networks for obstacle avoidance of mobile robots
 Neural Networks for Mobile Robot Navigation: A Survey 1221

should take the sensor data from the environment as the inputs, and output the
direction for the robot to proceed. Fujii et al. proposed a multilayered model for
collision avoidance of a mobile robot through reinforcement learning [21]. Silva
et al. presented the MONODA (modular network for obstacle detection and
avoidance) architecture for obstacle detection and avoidance of a mobile robot
in an unknown environment [22]. This model consists of four modules that are
three-layered feedforward neural networks (each detects the probability of ob-
stacle in one direction of the robot). Ishii et al. developed an obstacle avoidance
method for underwater vehicle based on self-organizing Kohonen neural net-
works [23]. Gaudiano and Chang studied an approach to avoid obstacle using a
neural network model of classical and operant conditioning based on Grossberg’s
conditioning circuit [24] [25].

2.3 Neural Networks for Path Planning

The path planning problem, which may consist of two subproblems, path gen-
eration and path tracking, refers to determining a path between an initial pose
of the robot and a final pose such that the robot does not collide with any ob-
stacles in the environments and that the planned motion is consistent with the
kinematic constraints of the vehicle. The existing path planning methods include
A* algorithm [26], potential fields [27], BUG2 algorithm [28], and methods using
intelligent control technique. The A* algorithm assumes that paths are made of
a series of points in the free space. Each segments given a value that is the cost
of that particular portion of the path. The drawback of the A* algorithm is that
the generated path is made of a series of connected straight lines, which makes
its curvature radius discontinuous, resulting in wheel slippage. Potential field
methods were first introduced by Khatib [27]. The drawbacks of this approach
are that convergence is not guaranteed and it requires very heavy calculation
power. The A* algorithm and potential fields are employed in the static envi-
ronment and assume that the map of the environment is known a prior. BUG2
algorithm is one of behavior based techniques, which divides the goal-seeking
task into several dependent subtasks. Though it is simple and efficient, it does
not always generate the optimal path. The methods using intelligent control such
as neural networks, neuro-fuzzy, do not plan a global path for mobile robots and
can be employed in an unknown environment. The input pattern of the neural
network for path planning of mobile robots should consider the following data:
robot’s actual position and velocities; robot’s previous positions and velocities;
target position and sensor data, and then output the commands to drive the
robot to follow a path towards the target with obstacle avoidance according to
these data.
Kozakiewicz and Ejiri have used a human expert to train a feedforward neural
network that reads inputs from a camera and outputs the appropriate com-
mands actuators [29]. In [30], Sfeir et al. developed a path generation technique
for mobile robot using memory neuron network proposed by Sastry et al. [31].
The memory neuron network is a feedforward neural network that uses mem-
ory neurons. A memory neuron is a combination of a classic perception and unit
1222 A.-M. Zou et al.

delays, which gives it memory abilities. Glasius, Komoda, and Gielen proposed a
Hopfield-type neural network for dynamic trajectory formation without learning
[32]. Fierro and Lewis studied a control structure that integrated a kinematic
controller and a feedforward neural network computed-torque controller for non-
holonomic mobile robot, and the neural network weights are tuned on-line, with
no “off-line learning phase” needed [33], [34], [35]. Yang and Meng proposed
a biologically inspired neural network approach for motion planning of mobile
robots [36], [37], [38]. This model is inspired by Hodgkin and Huxley’s mem-
brane model [39] for a biological neural system and Grossberg’s shutting model
[40]. The proposed model for motion planning of mobile robots has the following
properties: without any prior knowledge of the environment, without explicitly
searching over the free workspace or the collision path, and without any learning
procedure.
However, neural networks have also some drawbacks. For instance, a neural
network can not explain its results explicitly and its learning is usually time-
consuming. Further, the learning algorithm may not be able to guarantee the
convergence to an optimal solution [41].

3 Hybrid Approaches for Mobile Robot Navigation

Though neural networks have some properties that are important for the mobile
robot navigation, knowledge representation and extraction are difficult. Fuzzy
systems are able to treat uncertain and imprecise information; they make use of
knowledge in form of linguistic rules. Their main drawback is lack of systematic
methodology for their design. The technology that combines or fuses the neural
network with the fuzzy reasoning is being watched some very interesting archi-
tectures [42]. Several fuzzy neural networks have been presented and used for
mobile robot navigation successfully [43-45, 47, 49-52].
Godjevac and Steele proposed a neuro-fuzzy controller based on Takagi-
Sugeno design and a radial basis function for obstacle avoidance and wall fol-
lowing of a mobile robot [43]. Marichal et al. presented a neuro-fuzzy approach
to guide a mobile robot with obstacle avoidance [44]. The proposed neuro-fuzzy
strategy, which consists of a three-layer neural network along with a competi-
tive learning algorithm, is able to extract the fuzzy rules and the membership
functions according to the information provided by a set of trajectories that are
obtained from a human guidance. Er and Deng studied a hybrid learning ap-
proach for obstacle avoidance of a mobile robot [45]. In [45], firstly, a neuro-fuzzy
controller is developed from a pre-wired or innate controller based on supervised
learning in a simulation environment. The fuzzy inference system has been con-
structed based on the Generalized Dynamic Fuzzy Neural Networks learning
algorithm proposed by Wu and Er et al. [46], whereby structure identification
and parameters estimation are performed automatically and simultaneously. Sec-
ondly, the controller is implemented on a real robot after the learning phase.
A reinforcement learning algorithm based on the Fuzzy Actor-critic learning
 Neural Networks for Mobile Robot Navigation: A Survey 1223

algorithm is employed so that the system can re-adapt to a new environment

without human intervention.
Fuzzy Adaptive Resonance Theory (ART) and fuzzy ARTMAP were pro-
posed by Carpenter and Grossberg et al. [47], [48]. Fuzzy ART is capable of
learning stable recognition categories in response to both analog and binary in-
put patterns, and fuzzy ARTMAP can rapidly learn stable categorical mapping
between analog or binary input and output vectors. Araujo has used fuzzy ART
neural network for on line map building from actual sensor data [49]. Later,
this work has been extended. Prune-able Fuzzy ART neural network, which in-
cluded the ability to selectively remove recognition categories, was introduced to
build map of mobile robot in unknown environments [50]. The fuzzy ART based
approach for map building of mobile robots has the following characteristics
[50]: (1) Self-organization from perceived sensor data; (2) Multi-functionality for
map building, motion planning; (3) Updatability: incremental, on-line update
by learning separately each sensor data point, thus make the model available as
soon as possible; (4) Compact geometric representation with small data require-
ments; (5) Low computational costs; (6) Possible application to higher dimen-
sional space without adversely impacting on data size and complexity. Streilein
et al. presented an approach to sonar-based object recognition using a fuzzy
ARTMAP neural network for the mobile robot localization [51]. Azouaoui et
al. proposed an approach for obstacle avoidance of mobile robot using fuzzy
ARTMAP neural network [52]. This approach can provide robots with capabil-
ity, after learning based on the supervised fast stable learning: Simplified fuzzy
ARTMAP, to determine and use the rule allowing the robots to avoid collision.

4 Conclusions
In this paper, we have given a brief discussion on mobile robot navigation using
neural networks. Although a great deal of progress has been made in the field
of mobile robot navigation using neural networks, we have to go a long way
to make the robot to have the capabilities of intelligence and autonomy truly,
which will be possible when the neural hardware evolves and we get a better
understanding of how the human brain works. In the mean time, perhaps the
best approach is the hybrid approach that combines neural networks with other
artificial intelligent algorithms such as fuzzy logic, knowledge-based systems and
genetic algorithms.

Acknowledgements
This research has been supported in part by the National Natural Science Foun-
dation of China (Grant Nos. 60205004, 50475179 and 60334020), the National
Basic Research Program (973) of China (Grant No. 2002CB312200), the Hi-Tech
R&D Program (863) of China (Grant Nos. 2002AA423160 and 2005AA420040),
and the Science and Technology New Star Program of Beijing (Grant No. H02082
0780130).
1224 A.-M. Zou et al.

References

1. Sukhatme, G. S., Mataric, M. J.: Robots: Intelligence, Versatility, Adaptivity. Com-

munications of the ACM 45 (3) (2002) 30-32
2. Kortenkamp, D., Bonasso, R. P., Murphy, R. (Eds): Artificial Intelligence and
Mobile Robots: Case Studies of Successful Robot Systems, AAAI press (1998)
3. Kim, H. H., Ha, Y. S., Jin, G. G.: A Study on the Environmental Map Building
for a Mobile Robot Using Infrared Ranger-finder Sensors. Proceedings of the IEEE
International Conference on Intelligent Robots and Systems, 1 (2003) 711-716
4. Thrun, S. B.: Exploration and Model Building in Mobile Robot Domains. Proceed-
ings of the IEEE International Conference on Neural Networks, 1 (1993) 175-180
5. Meng, M., Kak, A. C.: Fast Vision-Guided Mobile Robot Navigation Using Neural
Networks. Proceedings of the IEEE International Conference on Systems, Man and
Cybernetics, 1 (1992) 111-116
6. Zou, A. , Hou, Z.G., Zhang, L., Tan, M.: A Neural Network-based Camera Calibra-
tion Method for Mobile Robot Localization Problems. Proceeding of the Second
International Symposium on Neural Networks, Chongqing, China, Springer Lecture
Notes in Computer Sciences (LNCS), 3498 (2005) 277-284
7. Barbera, H. M., Skarmeta, A. G., Izquierdo, M. Z., Blaya, J. B.: Neural Networks
for Sonar and Infrared Sensors Fusion. Proceedings of the Third International
Conference on Information Fusion, 2 (2000) WEB4/18 - WEB4/25
8. Lippman, R.: An Introduction to Computing with Neural Nets. IEEE ASSP Mag-
azine 4 (1987) 4-22
9. Hu, H. S., Gu, D. B.: Landmark-Based Navigation of Mobile Robot in Manufactur-
ing. Proceedings of the IEEE International Conference on Emerging Technologies
and Factory Automation, 1 (1999) 114-121
10. Janet, J. A., Gutierrez-Osuna, R., Chase, T. A., White, M., Luo, R. C.: Global
Self-localization for Autonomous Mobile Robots Using Self-organizing Kohonen
Neural Networks. Proceedings of the IEEE International Conference on Intelligent
Robotics and Systems, 3 (1995) 504-509
11. Crowley, J. L., Wallner, F., Schiele, B.: Position Estimation Using Principal Com-
ponents of Range Data. Proceedings of the IEEE International Conference on Ro-
botics and Automation, 4 (1998) 3121-3128
12. Vlassis, N., Motomura, Y., Krose, B.: Supervised Dimension Reduction of Intrin-
sically Low-dimensional Data. Neural Computation 14 (1) (2002) 191-215
13. Nayar, S. K., Murase, H., Nene, S. A.: Learning, Positioning, and Tracking Visual
Appearance. Proceedings of the IEEE International Conference on Robotics and
Automation, 4 (1994) 3237-3244
14. Artac, M., Jogan, M., Leonardis, A.: Mobile Robot Localization Using an Incre-
mental Eigenspace Model. Proceedings of the IEEE International Conference on
Robotics and Automation, 1 (2002) 1025-1030
15. Tamimi, H., Zell, A.: Global Visual Localization of Mobile Robots Using Kernel
Principal Component Analysis. Proceedings of the IEEE International Conference
on Intelligent Robots and Systems, 2 (2004) 1896-1901
16. Vapnik, V.: Statistical Learning Theory. John Willey & Sons (1998)
17. Zou, A. , Hou, Z.G., Tan, M.: Support Vector Machines (SVM) for Color Image Seg-
mentation with Applications to Mobile Robot Localization Problems. Proceeding
of the International Conference on Intelligent Computing, Hefei, China, Springer
Lecture Notes in Computer Sciences (LNCS), 3645 (2005) 443-452
 Neural Networks for Mobile Robot Navigation: A Survey 1225

18. Hou, Z.G.: A Hierarchical Optimization Neural Network for Large-scale Dynamic
Systems. Automatica 37 (12) (2001) 1931-1940
19. Hou, Z.G.,Wu, C.P., Bao.: A Neural Network for Hierarchical Optimization of
Nonlinear Large-scale Systems. International Journal of Systems Science 29 (2)
(1998) 159-166
20. Djekoune, O., Achour, K.: Vision-guided Mobile Robot Navigation Using Neural
Network. Proceedings of the 2nd International Symposium on Image and Signal
Processing and Analysis, (2001) 355-361
21. Fujii, T., Arai, Y., Asama, H., Endo, I.: Multilayered Reinforcement Learning for
Complicated Collision Avoidance Problems. Proceedings of the IEEE International
Conference on Robotics and Automation, 3 (1998) 2186-2191
22. Silva, C., Crisostomo, M., Ribeiro, B.: MONODA: A Neural Modular Architecture
for Obstacle Avoidance without Knowledge of the Environment. Proceedings of the
IEEE-INNS-ENNS International Joint Conference on Neural Networks, 6 (2000)
334-339
23. Ishii, K., Nishida, S., Watanabe, K., Ura, T.: A Collision Avoidance System based
on Self-organizing Map and its Application to an Underwater Vehicle. 7th Interna-
tional Conference on Control, Automation, Robotics and Vision, 2 (2002) 602-607
24. Gaudiano, P., Chang, C.: Adaptive Obstacle Avoidance with a Neural Network for
Operant Conditioning: Experiments with Real Robots. Proceedings of the IEEE
International Symposium on Computational Intelligence in Robotics and Automa-
tion, (1997) 13-18
25. Grossberg, S., Levine, D. S.: Neural Dynamics of Attentionally Modulated Pavlov-
ian Conditioning: Blocking, Inter-stimulus Interval, and Secondary Reinforcement.
Applied Optics 26 (1987) 5015-5030
26. Nilsson, N. J.: Principles of Artificial Intelligence. Tioga Publishing Co., Palo Alto,
CA (1980)
27. Khatib, O.: Real-time Obstacle Avoidance for Manipulators and Mobile Robots.
The International Journal of Robotics Research 5 (1) (1986) 90-98
28. Prusky, A.: Robotique Mobile. La Planification de Trajectoire, Hermes (1996)
29. Kozakiewicz, C., Ejiri, M.: Neural Network Approach to Path Planning for Two
Dimensional Robot Motion. International Workshop on Intelligent Robots and Sys-
tems, 2 (1991) 818-823
30. Sfeir, J., Kanaan, H. Y., Saad, M.: A Neural Network Based Path Generation
Technique for Mobile Robots. Proceedings of the IEEE International Conference
on Mechatronics, (2004) 176-181
31. Sastry, P. S., Santharam, G., Unnikrishnan, K. P.: Memory Neuron Networks for
Identification and Control of Dynamic Systems. IEEE Transactions on Neural Net-
works, 5 (2) (1994) 306-319
32. Glasius, R., Komoda, A., Gielen, S. C. A. M.: Neural Network Dynamics for Path
Planning and Obstacle Avoidance. Neural Networks 8 (1) (1995) 125-133
33. Fierro, R., Lewis, F. L.: Control of a Nonholonomic Mobile Robot Using Neural
Networks. Proceedings of the IEEE International Symposium on Intelligent Con-
trol, (1995) 415-421
34. Fierro, R., Lewis, F. L.: Practical Point Stabilization of a Nonholonomic Mobile
Robot Using Neural Networks. Proceedings of the 35th IEEE on Decision and
Control, 2 (1996) 1722-1727
35. Fierro, R., Lewis, F. L.: Control of a Nonholonomic Mobile Robot Using Neural
Networks. IEEE Transactions on Neural Networks 9 (4) (1998) 589-600
36. Yang, S. X., Meng, M.: An Efficient Neural Network Approach to Dynamic Robot
Motion Planning. Neural Networks 13 (2) (2000) 143-148
1226 A.-M. Zou et al.

37. Yang, S. X., Meng, M.: An Efficient Neural Network Method for Real-time Motion
Planning with Safety Consideration. Robotics and Autonomous Systems 32 (2-3)
(2000) 115-128
38. Yang, S. X., Meng, M. Q.-H.: Real-time Collision-free Motion Planning of a Mobile
Robot Using a Neural Dynamic-based Approach. IEEE Transactions on Neural
Networks 14 (6) (2003) 1541-1552
39. Hodgkin A. L., Huxley, A. F.: A Quantitative Description of Membrane Current
and its Application to Conduction and Excitation in Nerve. Journey of Physiology
(London) 117 (1952) 500-544
40. Grossberg, S.: Nonlinear Neural Networks: Principal, Mechanisms, and Architec-
ture. Neural Networks 1 (1988) 17-61
41. Huang, S. H.: Artificial Neural Networks and its Manufacturing Applica-
tion: Part 1. Online slides are available at : www.min.uc.edu/icams/resourses/
ANN/ANNMaul.ppt
42. Jang, J.-S. R.: ANFIS: Adaptive-network-based Fuzzy Inference System. IEEE
Transactions on Systems, Man, and Cybernetics 23 (3)(1993) 665-685
43. Godjevac, J., Steele, N.: Neuro-fuzzy Control of a Mobile Robot. Neurocomputing
28 (1999) 127-143
44. Marichal, G. N., Acosta, L., Moreno, L., Mendez, J. A., Rodrigo, J. J., Sigut, M.:
Obstacle Avoidance for a Mobile Robot: A Neuro-fuzzy Approach. Fuzzy Sets and
Systems 124 (2) (2001) 171-179
45. Er, M. J., Deng, C.: Obstacle Avoidance of a Mobile Robot Using Hybrid Learning
Approach. IEEE Transactions on Industrial Electronics 52 (3) (2005) 898-905
46. Wu, S., Er, M. J., Gao, Y.: A Fast Approach for Automatic Generation of Fuzzy
Rules by Generalized Dynamic Fuzzy Neural Networks. IEEE Transactions on
Fuzzy Systems 9 (4) (2001) 578-594
47. Carpenter, G. A., Grossberg, S., Rosen, D. B.: Fuzzy ART: Fast Stable Learning
and Categorization of Analog Patterns by an Adaptive Resonance System. Neural
Networks 4 (1991) 759-771
48. Carpenter, G. A., Grossberg, S., Markuzon, N., Reynolds, J. H., Rosen, D. B.:
Fuzzy ARTMAP: A Neural Network Architecture for Incremental Supervised
Learning of Analog Multidimensional Maps. IEEE Transactions on Neural Net-
works 3 (1992) 698-713
49. Araujo, R., de Almeida, A. T.: Learning Sensor-based Navigation of a Real Mobile
Robot in Unknown Worlds. IEEE Transactions on Systems, Man and Cybernetics,
Part B 29 (2) (1999) 164-178
50. Araujo, R., Lourenco, D., Ferreira, G.: Integrating Laser and Infrared Sensors
for Learning Dynamic Self-organizing World Maps. International Conference on
Multisensor Fusion and Integration for Intelligent Systems, (2001) 293-298
51. Streilein, W. W., Gaudiano, P., Carpenter, G. A.: A Neural Network for Ob-
ject Recognition through Sonar on a Mobile Robot. Proceedings of the IEEE
ISIC/CIRA/ISAS Joint Conference, (1998) 271-276
52. Azouaoui, O., Ouaaz, M., Chohra, A., Farah, A., Achour, K.: Fuzzy ARTMap
Neural Network Based Collision Avoidance Approach for Autonomous Robots Sys-
tems. Proceedings of the Second International Workshop on Robot Motion and
Control, (2001) 285-290