TNO report Pre-crash sensing for increasing active

01.OR.BV.013.0/PM and passive safety

literature survey towards state-of-the-art

TNO Automotive
developments

Schoemakerstraat 97 Date
P.O. Box 6033
2600 JA Delft
23 April 2001
The Netherlands
Author(s)
Phone + 31 15 269 69 00
Fax +31 15 262 07 66 P.L.J. Morsink

Client/Sponsor

Reg.nr. I 065
Passive Safety Network

Approved by
P.P.M. Lemmen
(Programme Manager)

Also seen by
J.S.H.M. Wismans
(manager R&D)

Project code
009.00311/01.03.04

Research period
week 01 - 15
UK
All rights reserved.
No part of this publication may be
reproduced and/or published by print,
photoprint, microfilm or any other means
without the previous written consent of
TNO.

In case this report was drafted on

instructions, the rights and obligations of
contracting parties are subject to either Number of pages
the Standard Conditions for Research
Instructions given to TNO or the relevant
46
agreement concluded between the
contracting parties. Number of appendices
Submitting the report for inspection to
parties who have a direct interest is 3
permitted.
Number of figures

© 2001 TNO 23

Number of tables
0

TNO Automotive provides R&D services and tools

in the field of vehicles, components and systems.
Our primary areas of expertise are Vehicle Dynamics,
Crash Safety, Combustion Engines, Advanced Transport Netherlands Organization for
Systems and Homologation. Applied Scientific Research (TNO)
TNO report

01.OR.BV.013.0/PM 23 April 2001 2 of46

Abstract

This report gives an overview of the state-of-the-art and future trends for pre-crash
sensing by order of the European Vehicle Passive Safety Network. Pre-crash
sensing overlaps both active and passive safety. It comprises all systems that
monitor crash circumstances just before and in the initial phase of the crash in
order to avoid or influence the crash. With sensors in the pre-crash phase valuable
time can be gained for anticipating to the upcoming crash event. It provides a way
to adapt safety systems to individual occupants features and real life crash
circumstances. A combination of systems for collision avoidance and damage
mitigation will lead to an integrated automotive safety system that is very
promising for increasing future traffic safety.

Most important systems for collision avoidance are advanced driver assistance
systems. Actions can be performed by the driver triggered by a warning or
autonomously by the car system. For damage mitigation mainly individual
occupant sensing systems were found that are used to trigger smart restraint
systems like airbags and belt tensioners.

The relatively unexplored area of advanced protection systems that adapt to

specific crash modes, like adjustable interior and exterior car structures (e.g. lifting
hood for pedestrian impact, self installing bumpers and head restraints, smart
materials) can very well be addressed by the Passive Safety Network. Although
current sensor technology can be used to further develop pre-crash sensing,
attention has to be paid to object classification functionality, cost efficiency and
typical automotive requirements like robustness and reliability. Furthermore
attention has to be paid to societal and legislative matters like driver acceptance
and responsibility for the operation of the systems.
TNO report

01.OR.BV.013.0/PM 23 April 2001 3 of46

Contents

1 Introduction........................................................................................4

2 Collision avoidance zone.....................................................................8

2.1 Normal driving state .............................................................8
2.2 Warning state .....................................................................11
2.2.1 Categories of driver assistance systems for collision
warning..............................................................................12
2.2.2 Obstacle detection...............................................................13
2.2.3 Tasks related to obstacle detection .......................................14
2.2.4 Internal vehicle and driver monitoring..................................16
2.3 Collision avoidable state .....................................................16

3 Damage mitigation zone....................................................................19

3.1 Collision unavoidable state .................................................19
3.1.1 Crash sensors .....................................................................20
3.1.2 First initiatives in intelligent occupant protection systems......22
3.1.3 Occupant classification sensing............................................24
3.1.4 Advanced protection actions ................................................30
3.1.5 Modelling...........................................................................31
3.1.6 Other sensors .....................................................................32
3.2 Post Event State .................................................................32

4 Conclusions and future needs ............................................................34

5 References .......................................................................................37

Appendices

A Parallel EC projects

B List of abbreviations

C Specifications of sensor types

TNO report

01.OR.BV.013.0/PM 23 April 2001 4 of46

1 Introduction

Over the past decades a considerable effort is put in improving and optimising road
vehicle safety. Next to structural improvements of the car, attention is turning
increasingly to the opportunities offered by advances in sensor technology and
information technology. In the field of active safety there has been a lot of activity
resulting in a wide range of safety devices that assist the driver in driving
comfortably and safely. Many of them belong to the advanced driver assistance
systems (ADAS) and navigation systems categories. Also in the field of passive
safety there are many opportunities. The consequences of accidents can be further
mitigated by monitoring information of the upcoming crash. Smart safety systems
can take real life crash circumstances and individual occupant features into
account.
Combined, these systems add a new dimension to vehicle safety: Intelligent
Transportation Systems (ITS). ITS comprises all applications of advanced
information processing, communications, remote sensing and control technologies
in the field of transportation engineering. They are all directed towards a common
goal: increasing transportation efficiency and road safety for all kinds of drivers
(and other road users) for all kinds of driving situations and for all kinds of crash
scenarios.

The focus of this literature study is on pre-crash sensing. Pre-crash sensing

typically deals with safety issues that are in the overlap area of passive and active
safety. It deals with sensing parameters that are directly involved in the crash. The
position of pre-crash sensing in the holistic safety approach defined by EUCAR is
illustrated in figure 1. The figure shows the level of critical safety situations that
increases as a function of time from the detection of a dangerous situation to the
actual crash.

Based on this figure the following definition of pre-crash sensing is proposed:

Pre-crash sensing =
• Monitoring crash circumstances,
• before the crash
• and in the initial phase of the crash,
• in order to avoid or influence the crash.

In this definition crash circumstance include the crash scene, characteristics of the
crash partner and individual occupant features.

The goal of this study is to give an overview of existing systems and future trends
related to pre-crash sensing in the automotive working field by order of the
European Vehicle Passive Safety Network.
TNO report

01.OR.BV.013.0/PM 23 April 2001 5 of46

Different sources have been used for the literature survey ranging from scientific
reports to popular scientific magazine articles and internet information.

Figure 1. Pre-crash sensing in the holistic safety approach defined by EUCAR

(presented on 22-11-2000 at the EUCAR conference in Brussels)

It is noted that currently research on pre-crash sensing a.o. is done in parallel EC

projects CHAMELEON and RADARNET. The goal of CHAMELEON [17] is to
support, to guide and to validate the development of a pre-crash sensorial system
necessary for near impending crash detection all around the vehicle and in all
scenarios. The goal of RADARNET is to develop a new type of low cost radar
network for automotive applications and to prepare the realisation of first
applications. A short overview of the activities in these projects in given in
appendix A.

For realising the benefits of pre-crash sensing one should go beyond the
conventional distinction of passive and active safety. The vehicle should be
considered as an integrated personal safety system. As a framework for this way of
thinking, five different vehicle driving scenarios have been identified that can
effectively be used for addressing existing or future safety systems [4]. They are
shown in figure 2 and described in more detail below.
TNO report

01.OR.BV.013.0/PM 23 April 2001 6 of46

Figure 2. Integrated safety system state diagram [4]

The collision avoidance zone concerns all events and subsequent actions that
apply to driving before a collision takes place. The damage mitigation zone is
directed towards minimising harmful effects of a collision. It deals with situations
in which the impact will definitely take place, is taking place or has already taken
place.

1. Normal driving state

No increased alertness is required, the driver can just use comfort and
convenience features supplied by modern automotive electronics.

2. Warning state
Full awareness of the driving situation is required in order to detect and prevent
potential crash situations.

3. Crash avoidable state

This is the last opportunity to avoid an accident and return to the normal
driving state. Important issues are reaction time, vehicle stability and control.
This state overlaps both zones: it also applies to preventing a secondary impact
after the primary impact has already taken place.

4. Crash unavoidable state

Immediately before and immediately after the crash all actions have to be done
to reduce the effects of the accident. All measurements that are taken to
prevent the collision may also be helpful for reducing the effects of the
accident (e.g. automatic breaking)
TNO report

01.OR.BV.013.0/PM 23 April 2001 7 of46

5. Post event state

After the collision the severity of the event with respect to driver and vehicle is
assessed and rescue warning can take place.

According to the previous definition, pre-crash sensing mainly deals with events
that happen in the warning, collision avoidable and collision unavoidable states.
However a clear trend is observed to integrate all safety features in one co-
operative system. Therefore in this study attention is paid to systems that belong to
all driving states.

In chapter 2 an overview is given of developments in the collision avoidance zone,

the upper ellipsoid of figure 2. In chapter 3 developments in the damage mitigation
zone, the lower ellipsoid in figure 2, are discussed. Conclusions with respect to the
current state of the art systems and future perspectives are given in chapter 4.
TNO report

01.OR.BV.013.0/PM 23 April 2001 8 of46

2 Collision avoidance zone

The collision avoidance zone deals with all aspects of safe driving before a
collision takes place. In this chapter an impression is given of collision avoidance
systems that are already available on the market or currently under development.
Comments of the automotive industry and government agencies have been
included to illustrate current and future visions and attitudes. In the normal driving
state mainly long distance communication systems between vehicles and the
environment are found. In the warning state there mainly is a combination of short
distance communication between vehicles and the environment and in-vehicle
communication systems. The collision avoidable state merely deals with short
distance communication systems. In appendix C a listing of different sensor
specifications is given.

2.1 Normal driving state

Most of the systems that act in this state support the driver in performing routine
tasks and provide increasing comfort to the driver. They are not directly involved
in an eventual upcoming crash event. Still it is worthwhile to pay attention to these
systems, mainly because of the future trend to have fully integrated vehicle safety
concepts. This means as less hardware as possible for as many applications as
possible. Systems that have been developed for application in this driving state
may very well be upgraded to perform direct safety tasks.

Adaptive Cruise Control (ACC):

The ACC function is able to maintain both the required cruising speed and the
headway between vehicles (safe distance from the proceeding car) by acting on the
accelerator and the braking system. It is a well-known function that is already
applied by car manufacturers [4], [25], [41].

ACC only needs a single forward looking radar, but at lower speeds people have a
tendency to cut in much closer than they would at higher speeds and additional
sensors may be needed to detect vehicles cutting in very close. These sensors can
be combined with a parking aid system. Eventually all these things will start to
work together as an integrated system [25]. Therefore it is expected that the next
generation of ACC will be able to handle low-speed stop & go traffic situations as
well [4].

Stop & Go:

The Stop & Go function maintains longitudinal car control at low speeds, e.g. when
queuing on a motorway or waiting at a traffic light [41]. It is a further enhancement
of ACC: it will bring the vehicle to a complete stop and hold the brake. It only
takes pushing a button to resume [27]. It is not an emergency brake system: the
driver keeps in control. It facilitates automatic acceleration and deceleration during
driving in busy urban areas as indicated in figure 3.
TNO report

01.OR.BV.013.0/PM 23 April 2001 9 of46

Figure 3. Stop-and-go system: facilitates driving in town [25]

relevant numbers: 1. Close range sensor 3. Radar 11. Steering angle sensor 12.
Jaw angle and lateral acceleration sensor.

GPS (Global Positioning System):

GPS that is originally developed for military purposes, is increasingly applied for
navigation purposes. The general feeling is that with a resolution less than 1 m the
application of many automotive safety features by GPS becomes interesting. E.g.
direct feedback of driver control actions to the position of the vehicle may be
given.

Other potentials of car navigation systems like GPS can be illustrated by the
Intelligent Speed Adaptation (ISA) project that was initiated by authorities in the
Netherlands to prevent speeding [28]. The maximum speed of cars is influenced in
specific zones by intelligent speed control. The GPS system determines the zone in
which the car is. The car that is further equipped with an on-board computer that
acts on the throttle to control speed, is actively slowed down when it drives into a
zone with a higher than locally prescribed maximum speed. Driver acceptance was
surprisingly high.

Convoying or automatic vehicle guiding (AVG):

AVG requires advanced environment sensing (e.g. radar) with vehicle-to-vehicle
communication. It can do what a good driver normally does: keep an eye on the
vehicle immediately in front and at the same time communicate with the car at the
front of the queue. Applications are in the field of safety, comfort and traffic
management [39]. If the lead car touches its brakes, all cars/drivers would know
about it and adjust their speed accordingly. This may be done by an automatic
action of the vehicle system but also a warning can be issued to the driver to take
an appropriate action. Technologically it may be achieved by upgrading ACC
TNO report

01.OR.BV.013.0/PM 23 April 2001 10 of46

systems (with more sensors) combined with car-to-car communication tools (e.g.
infrared) [39] and advanced GPS applications. There is some doubt about the car
drivers' acceptance concerning automatic vehicle control actions. In [27] it is
stated that convoying is efficient but that it will not happen soon, because it is not
what the customer wants.

Vehicle and infrastructure communication:

In addition or next to vehicle-vehicle communication remote control functions can
be used to allow two-way transmission between vehicle and infrastructure: signals
are received from electronic poles and combined with speed data (permitted or
recommended) for longitudinal car control [41].

Vehicle dynamics/stability control:

A wide range of vehicle dynamics sensors is applied in vehicles to facilitate the
electronic stability programme (ESP), that is e.g. applied in the Ford Focus [27],
[29]. See figure 4.
Fiat a.o. has developed a dynamic stability control tool [41]. A demonstration car is
equipped with active, integrally-managed systems (active suspension, steering,
four-wheel drive and ABS) to ensure a high standard of comfort and safety while
driving. The increase in active safety is particularly appreciated over slippery
ground and during emergency manoeuvres such as overtaking or avoidance of
obstacles that appear suddenly in the car's path.

Figure 4. Vehicle dynamics sensors [4]

Personal car settings :

Tailor made safety systems that already apply to the normal driving state or even
before driving can be realised by personal car settings. The following examples
were found:
Belt usage reminder system: a safety usage sensor located in the belt buckle will
determine if the driver is buckled up. An appropriate action will be taken even
before driving.
Keyless entree into a car: the car recognises the owner's key card and can perform
subsequent action (setting variables to a personalised value e.g.) [23].
TNO report

01.OR.BV.013.0/PM 23 April 2001 11 of46

2.2 Warning state

The goal of warning systems in general is to provide alerts to drivers to assist them
in avoiding or reducing the severity of crashes. The system monitors the scene and
evaluates potential threats. It should be noted that the transition between systems
that are mainly used in the normal driving or warning state is not very strict.

This can be illustrated by the Ford technology road map shown in figure 5 [7]. A
key role is identified for ACC and parking and reversing aid systems (PRA). By
integration of longitudinal control, lateral control and obstacle detection as key
building blocks, an upgraded ACC system may act as a collision warning or even
avoidance system. Collision warning is a natural extension of ACC using capable
all weather forward looking sensors. An ACC system can be seen as a first step in
longitudinal control and PRA is a first application of a high value and affordable
obstacle detection system. The next steps on the path to collision avoidance are
lateral vehicle control and frontal collision warning. Lane control could first be
introduced as lane keeping/departure warning.

Figure 5. Technology Roadmap for Collision Avoidance [7]

Another example to illustrate the overlap between current normal driving

assistance systems and warning systems is the possible upgrading of the GPS
system. Nissan is doing research towards predicting driver behaviour (with MIT)
[23]. Sensors monitor movement of eyes, feet and hands in order to predict what a
person is going to do within a split second. If this can be combined with technology
that tells the car where it is in relation to other cars and its surroundings, using e.g.
TNO report

01.OR.BV.013.0/PM 23 April 2001 12 of46

GPS, the car could assess whether the move the driver was going to make was
unsafe. Subsequently warning and control actions can be performed.

2.2.1 Categories of driver assistance systems for collision warning

In [14] three possible application areas of warning devices have been identified in
order of sophistication (see also figure 6):

1. Warning functions and functions increasing the comfort of driving without

overtaking active control of vehicle movement (less critical with respect to
safety demands). Alerts should be provided in time to help drivers avoid most
common crashes (e.g. by braking only), while also minimising false alarms [1]
in order to improve driver acceptance.

2. Intervention into lateral and longitudinal control of vehicle movement (e.g.

intelligent cruise control for longitudinal and heading control for lateral
steering). This in general concerns more time-critical interventions.

3. Automated hands-off driving: lateral and longitudinal control is entirely

automated.

Most systems discussed in this section belong to the first category. These systems
only provide alerts, no attempt to control the vehicle to avoid an impending crash is
made.

Figure 6. Categories of Driver Assistance Systems [14]

For classifying the warning state systems in the following a rough distinction is
made between external object detection systems and internal vehicle state and/or
driver monitoring systems.
TNO report

01.OR.BV.013.0/PM 23 April 2001 13 of46

2.2.2 Obstacle detection

Obstacle detection is a major functionality area for collision warning and

avoidance systems. In case of an impending collision with another vehicle or
object, information about the collision partner is necessary for issuing the
appropriate warning or action. Among the information that is needed from sensors
there is:

• time to impact
• distance to the object
• object classification (including size and shape)
• trajectory of the obstacle
• closing velocity
• impact direction
• acceleration of the incoming object
• mass and stiffness of the incoming object
• point of impact

In the warning state the focus will be on information with respect to the distance,
the time to impact and object classification. An overview of different warning
functions related to obstacle detection is given in figure 7.

Figure 7. Radar system network for different warning applications [38]

Many of the systems found in this category use a combination of radar, laser,
acoustic, infrared, video camera and microwave sensors to provide short and long-
range data of obstacles present in the car's path [40], [41]. Another example is
given by Mitsubishi [10] that developed a compact scan laser radar sensor for
TNO report

01.OR.BV.013.0/PM 23 April 2001 14 of46

detecting distances. Laser radar is already in commercial use as warning device in

trucks, buses and upscale cars. To use it in a wider range it should be compact,
lower priced, lightweight, and easy to install.

An example of vehicle environment sensing by video sensors is given by Bosch

[11]. The system comprises a high luminance dynamic range video camera,
dedicated evaluation hardware based on commercially available PC technology and
algorithms for the interpretation of video scenes in video real time.

2.2.3 Tasks related to obstacle detection

Obstacle detection can be used for a wide range of driving tasks, that are already
applied in the current car fleet to some extent. For object detection sensors the
following additional tasks exist:

• Lane, roadway departure warning: Fiat presents a system that issues an

advance warning if the car inadvertently crosses the lane boundaries [41]. In
case of this example the road in front of the car is filmed by a micro TV
camera. The driver may select between tactile or acoustic alert indicators.
A scanning laser mounted on the right rear corner of the car is presented in [2].
It will scan a full 360 degrees. This laser is able to generate trajectories of
passing cars, and categorise them according to their length and width. The
scanning laser can serve as a stand-in Collision Avoidance System:
information from the laser plus other vehicle sensor inputs is processed by a
digital signal processor (DPS) and it is decided if a warning shall be presented
to the driver.
• Overtaking warning
• Parking warning or assistance

Figure 8. Radar based near distance sensing (Siemens)

• Road sign recognition

• Blind spot warning: example given by Fiat [41]: The blind spot coverage
sensor gives a warning of the presence of vehicles that are about to enter the
side area not covered by the door mirrors. The sensor turns on a warning light,
TNO report

01.OR.BV.013.0/PM 23 April 2001 15 of46

which can be used in conjunction with the door mirror. The warning signal
becomes acoustic when the turn signal is activated.

Figure 9. Blind spot warning

• Rear end collision warning

• Intersection collision avoidance (ICA) warning: in [26] an ICA system is
introduced that includes three 'of-the-shelf' headway radar systems integrated
with a global positioning system (GPS), a customised geographical information
system (GIS), and a head-up display (HUD). Both audible and visual warnings
inform the driver of a potential intersection danger. Collection and analysis of
data takes place from three directions.

Path prediction is a problem that occurs in detecting objects at longer ranges.

Current systems can only address situations that are observable by line-of-sight
sensing. If it is not possible to accurately project the host vehicle's path, it will be
impossible to determine which objects in the field of view represent a threat to the
host vehicle. This situation is depicted in figure 10.

Figure 10. Complex Roadway Scenario [4]

TNO report

01.OR.BV.013.0/PM 23 April 2001 16 of46

Furthermore false alarms should be minimised in order to increase the level of

functionality and to increase driver acceptance [4].

2.2.4 Internal vehicle and driver monitoring

Information about the conditions of the vehicle and the driver is valuable for
issuing a warning action. Examples of internal vehicle state warnings:

• Low tire pressure warning

• Impending rollover warning back up assistance or warning
• Tire-pressure sensors with display integrated in the rearward-mirror (JCI
product, announced to be on the market in January 2001).
• Temperature and road resistance sensors (combined to road condition).

Examples of driver monitoring are the following:

• Driver alertness and anti-dozing systems based on driver's eye blink rate (Ford)
or biological sensors monitoring the physical conditions of the driver [4]. In
[15] an application of a heavy vehicle drowsy driver detection system is
described. The system comprises an automated monitor using CCD (Charge
Coupled Device) cameras and data processing devices/software/algorithms.
• Driver alertness can also be monitored by checking how the driver is sitting
and/or operating the pedals and steering wheel [39]. This information can be
used to give an assessment of the safety situation.

2.3 Collision avoidable state

In the warning state with the various sensors' input the appropriate warning is to be
delivered to the driver with a minimum of false alarms. After the warning however
there must be an action (by the driver or automatically) to avoid the collision. In
the collision avoidable state reaction time, vehicle stability and control are
important. The systems described in this section mainly belong to category 2 of
section 2.2.1: intervention into lateral and longitudinal control.

The following three modes can be distinguished in direct collision avoidance [4]:
1. Driver initiated
2. Vehicle initiated
3. A blend of both

The same suit of sensors and sensor fusion algorithms as in the warning state can
be used to provide information about the state of the vehicle and its surroundings.
This is supplemented with a suit of X-by-wire products (steering, braking, throttle
and suspension) to de-couple the actuation from the mechanical input by the driver.
TNO report

01.OR.BV.013.0/PM 23 April 2001 17 of46

The information input for a collision avoidance system for forward, rear and side
applications is depicted in figure 11.

Figure 11. Information Flow Scheme for a Collision Avoidance System [4]

Most systems discussed until now do not act autonomously on vehicle control.
However there are exceptions. The Anti Collision Assist (ACA) system presented
by Fiat [41] helps to prevent collisions by sending the driver an acoustic or visual
signal to warn for the presence of standing or slow-moving obstacles in the car's
path. In addition the system acts on the car's brakes when necessary. Renault also
announced developments in this area [25]. Another example of active warning is
proposed by the electronic crumple zone of Daimler Chrysler [23]: a car is slowed
down if it gets too close to another vehicle and the driver fails to react.

In [5] a rear-end collision warning/avoidance (CW/CA) system algorithm is

presented that meets the following criteria based on acceptance by drivers:

• System warnings should result in a minimum load on driver attention

• Automatic control of the brakes should not interfere with normal driving
operation
• The system should perform well in a variety of driving conditions

The systems use distance, relative velocity and vehicle velocity information to
warn the driver or control the vehicle. A warning critical distance is defined as a
function of vehicle velocity and relative velocity. A warning is given to the driver
when the vehicle spacing is less than this warning critical distance. A braking
critical distance can be defined similarly. The system applies the brakes when the
spacing is less than the braking critical distance. Different definitions of the critical
warning and braking distance have been derived by Mazda and Honda [5]. Most
TNO report

01.OR.BV.013.0/PM 23 April 2001 18 of46

existing CW/CA systems in which two vehicles are involved use a similar
algorithm.

However not everybody appears to be convinced of vehicle initiated collision

avoidance. In [24] it is stated that the introduction of active collision avoidance,
where the vehicle would take some active control, will take considerable time.
Integration with all other systems on the vehicle is needed. At this moment the
driver still has to be in control because of reliability of the systems and processing
speed of the data involved. On the other hand collision warning systems, that do
not take active control, may be only five years away.

This is supported in [23] by stating that not a lot of detailed output is needed to
generate a collision warning. However in contrary to collision warning, for
collision avoidance a high level of input is needed like: steering angle position,
accelerometer information etc. More data requires more processing power. With
current processing power the driver has to stay in control. Also the reliability of the
systems is connected to the control issue.

A government based opinion is given in [18]. Here it is stated that with the advent
of Advanced Driver Assistance Systems (ADAS) the responsibility for the
operation of these safety functions is moving from the driver to the vehicle, or
perhaps more correctly, the vehicle manufacturer. ADAS have the disadvantage
that only a limited number of sensors can be applied. Furthermore these sensors
have a limited action range. Besides the information that is obtained can only be
processed with predefined strategies. Thus human failure can not be compensated
in all cases. Also it should be noted that an intelligent car may make the driver less
alert.
TNO report

01.OR.BV.013.0/PM 23 April 2001 19 of46

3 Damage mitigation zone

The collision avoidance zone concerns all events and subsequent actions that apply
to driving before a collision takes place. The damage mitigation zone deals with
situations in which the impact will definitely take place, is taking place or has
already taken place. Minimising harmful effects of the impact is the goal.

As indicated in the ellipsoid of figure 2 the collision avoidable state is in the

overlap area of the collision avoidance zone and the damage mitigation zone.
Collision avoidance is relevant for damage mitigation, for instance in situations
where there already was an accident but the vehicle should be prevented from
getting involved in a secondary impact. However the same sensor technology as in
the warning state and in the avoidance state discussed in the previous chapter are
used to mitigate further damage. Therefore in this chapter no further attention will
be paid to the collision avoidable state.

In the collision unavoidable state, the state that comprises the working field of
‘traditional’ occupant protection and passive safety, mainly short distance and in-
vehicle communication systems were found. In the post event state both long
distance and in-vehicle communication systems are relevant. Like in the previous
chapter the developments addressed here will be supplemented with visions in the
automotive working field.

In appendix C a listing of specifications of different sensors is given.

3.1 Collision unavoidable state

This state is most time critical. All actions should be done immediately before as
well as immediately after the crash to reduce the effects of the accident. Four
subsequent actions can be distinguished:

1. Monitoring of all relevant information

2. Processing all monitored information

3. Making the right decision for the most appropriate action

4. Carrying out the mitigation action

All information that is monitored to prevent the collision is also helpful for
reducing the effects of the accident. E.g. all object detection parameters that are
mentioned in section 2.2.2 are relevant in this state. However sensing of parameters
that will directly be involved in the crash is a step ahead. It implicates a maybe
small but very significant gain of time. This extra time can be used for optimising
post-impact countermeasures (such as variable stage airbags and seatbelt
TNO report

01.OR.BV.013.0/PM 23 April 2001 20 of46

tensioners) but also for reversible, pre-impact countermeasures (such as adaptable

interior and exterior structures and pedestrian protection). It opens a way to direct
crash safety more towards real world factors like crash type and direction, severity,
seat belt usage, occupant position and classification.

The future benefits of pre-crash sensing are expressed as follows [30]: `Pre-crash
sensing is most promising for the future of increasing safety. If a collision can
absolutely not be avoided anymore, the airbag can already be activated when the
first deceleration occurs, so much quicker than without warning by the forward
looking sensors. The crash pulse is still necessary because e.g. a radar sensor can
not distinguish an empty box or concrete pillar. A similar thing can be done with
the belt pretensioner: when a pre-crash alert signal is detected, the belt can be
tightened. When there is no severe deceleration detected, the belt can be released
into its comfort position again’.

Furthermore it is stated in [30]: 'In the period 2000-2005 new cars will be
introduced to the market with pre-crash sensor systems. They will mostly be used
to support existing systems. Especially in side impact there are big opportunities.
At the moment for the detection of side impact only 6 milliseconds are available.
Every additional information will be helpful to e.g. activate the airbag quicker,
making the deployment slower and therefore less aggressive, thus increasing
overall safety'.

In the following sections different ways of acquisition of crash information will be

addressed. First crash sensor systems that are used to detect the actual crash pulse
are discussed. These systems are logically extended with pre-crash capabilities.
Subsequently intelligent occupant protection systems are discussed that are
enhanced with occupant sensing technology.

3.1.1 Crash sensors

The goal of crash sensors is to make an early determination of the crash forces
transmitted to the occupant compartment. This type of crash severity sensors has
already been used for a long time e.g. for triggering airbags and belt tensioners.
They have mainly been focussed on assessing the severity of the crash by
monitoring the crash pulse in the earliest state of the crash. Examples like inertia
switches a.o. are given in [8].

Both single point and multi-point sensing systems are used. In [37] it is explained
that state-of-the-art single point sensing systems use sensors mounted inside the
passenger compartment on the tunnel. This way of sensing is clearly limited in
accurate determination of crash conditions. For an earlier and more accurate
determination of crash conditions, information from sensors outside of the
passenger compartment becomes necessary. For this goal up-front acceleration
sensors and pre-crash sensors are suggested.
TNO report

01.OR.BV.013.0/PM 23 April 2001 21 of46

Peripheral sensors in the crushable zone (figure 12) of the vehicle can give an
improved separation of low, mid and high severity crashes. They can also better
distinguish different crash modes. Therefore in case of e.g. air bags more
sophisticated deployment strategies are possible.

Figure 12. Up-front peripheral crash sensors (UPS) in the vehicle crush zone [37]

However the problem remains that there is not unlimited time to detect crash
velocity and severity and to make a decision upon the level of action. Monitoring
crash parameters prior to the crash can be used to gain valuable time in decision
making and to perform smoother safety actions. For this purpose anticipatory or
pre-crash sensing should be done. Especially short and medium range detectors
will be interesting. A natural extension of sensor systems that have been discussed
in the previous chapters will be effective from both a functional and cost point of
view. Figure 13 gives an overview of a possible configuration of close range pre-
crash sensors at all sides of the vehicle.

Figure 13. Pre-crash sensors at all sides of the vehicle [4]

TNO report

01.OR.BV.013.0/PM 23 April 2001 22 of46

It should be noted that sensing of only the velocity of colliding objects prior to the
crash without information about relevant masses and stiffnesses of crash partners is
limited [37].

3.1.2 First initiatives in intelligent occupant protection systems

In [13] functional areas of intelligent occupant protection systems (IOPS) are

defined. It is stated that IOPS need to fulfil and improve the occupant protection as
defined by crash test standards as well as offering an enhanced occupant
protection. The enhancement of occupant protection should first be directed
towards the 50th , 5th and 95th percentile in frontal impact. Subsequently protection
at critical crash conditions for all 3 standard dummy types should be addressed
(e.g. airbag bottoming out). The adaptive load limiter and the multi-stage airbag are
identified as the two main components of IOPS.

In the NTHSA report [21] examples are presented of vehicle manufacturers that are
beginning to install advanced air bag technology in an increasing variety of
vehicles. Honda introduced a dual stage inflator in its MY 1999 Acura. The MY
2000 Ford Taurus and Honda Accord, which are among the highest selling models
in the US, have dual-stage air bags. In [35] the double action airbag of the Renault
Laguna is shown: two airbags in one package that operate in conjunction with the
seatbelt to offer different levels of restraint according to the speed of the impact.
Mercedes and BMW also have dual-stage air bags in some of their MY 2000 cars.

The increasing role of sensor technology in restraint systems is expressed in [24]:

'Systems that reduce side impact damage to occupants, such as curtain airbags and
head bags are where we are likely to see some of the most immediate
developments. These will be combined with smart sensor technology and
integrated with vehicle systems to react in even more flexible ways'. In addition in
[24] it is stated: ' The ability to determine occupant weight and proximity prior to a
collision is a significant advance in the development of smart airbag systems.'

In [34] it is predicted that cars in the year 2006 will contain about 30 sensors that
are related to safety restraint systems. The increase of the number of sensors and
specific examples is indicated in figure 14.
TNO report

01.OR.BV.013.0/PM 23 April 2001 23 of46

Figure 14. Predicted number of sensors as a function of time [34]

Figure 15 gives a graphical overview of the positioning of different sensors in the

future car.

Figure 15. Sensors in a Distributed Systems Interface (DSI) architecture with plug-
and-play functionality by TRW [33]

In the NTHSA paper [21] it is further stated that proposed US requirements appear
to be achievable. The number of failures however illustrates that many vehicles
will need to be redesigned in a short period of time to meet a highly complex set of
new requirements. In many cases, manufacturers will be introducing several new
technologies simultaneously: dual level inflators, seat belt sensors, weight pattern
TNO report

01.OR.BV.013.0/PM 23 April 2001 24 of46

seat sensors, seat track position sensors, more complex algorithms, etc. That these
new requirements are taken seriously is indicated by the development of occupant
classification anthropometric test devices (OCATD) for evaluation of weight
pattern sensors [22].

3.1.3 Occupant classification sensing

Occupant classification systems have already been implemented in vehicles to

some extent. The MY 2000 Cadillac Seville e.g. has weight and pattern sensors in
the passenger seat that work together to turn off the passenger air bag when
children are present. The MY 1999 Hyundai Sonata has a weight sensor designed
to prevent the passenger air bag from deploying unless a weight of more than 66
pounds is detected on the passenger seat [21].

GM publicly announced in February 1999 that it will introduce technology in MY

2000 that is designed to detect the presence of a small child in the front passenger
seat and suppress the deployment of the passenger frontal air bag in the event of a
frontal crash [21]. GM stated that weight-based sensors, coupled with pattern
recognition technology, will distinguish between a child and a small adult female
whose weight may be similar to a large child restrained in a child safety seat. If the
front passenger seat is occupied by a small child, whether in a child safety seat or
not, GM said that the air bag will not deploy. GM further stated that it will
introduce this technology on the Cadillac Seville in the 2000 calendar year, and
that it has a roll-out plan to extend this technology throughout its product line.

In the following sections existing or occupant classification systems that are under
development are discussed. A distinction is made between occupant weight,
position (also e.g. position of the head) and motion sensors.

occupant weight sensing:

Goals [20]:
1. measure the size of an occupant by measuring forces on the seat
2. measure the distribution of the occupant on the seat to improve the ability to
classify occupants and their location on the seat.
Recent developments: measure the pressure distribution on the seat or deflection of
the seat and use this pattern to identify the occupancy of the seat (also regarding
the presence of a child restraint), the size of the occupant, and whether the
occupant is sitting back in the seat or up on the front edge.
Current technology:
1. Resistive strain gauges or load cells, typically near the base of the seat, that
indirectly lead to a measurement of weight.
2. Bladder system within the seat cushion that measures pressure.
3. Weight sensing electronic mat installed in the seat cushion. An array of
conductive polymeric sensors changes resistance under load.
Examples:
TNO report

01.OR.BV.013.0/PM 23 April 2001 25 of46

Tailorable occupant protection system (TOPS), Breed and Siemens joint venture
product, based on individual occupant data and crash severity [24]. The system
uses a weight classification system (WCS), with 4 strain gauge based force sensors
in or beneath the seat.

Figure 16. Occupant weight sensors TOPS system (internet site Siemens)

The data from these sensors go to an electronic control unit (ECU) that is based on
seat frame integrated occupant sensing technology. It also takes into account the
weight of the legs. The weight sensor collaborates with the seat position sensor to
determine the occupant's position with respect to the airbag. The outcomes are
classified in three position zones that are coupled to deployment characteristics of
the airbag. Figure 17 gives a graphical overview of the three possible positions.
TNO report

01.OR.BV.013.0/PM 23 April 2001 26 of46

Figure 17. Occupant position relative to the airbag detected by OOP sensors [31]

Passive Occupant Detection System (PODS, Delphi) [32]: a weight-based system

to enable the smart deployment or suppression of passenger airbags. The PODS
technology consists of bladder-based weight-sensing technology mounted under
the passenger-seat cushion. It concerns a silicone fluid filled bladder system, with a
pressure sensor under the seat cushion and an electronic control unit for sensor data
processing. Variable deploying or suppression of the airbag is achieved by
occupant classification algorithms and extensive signal processing. Delphi is
integrating PODS with another supplier's ultrasonic sensing system to provide
additional passenger positioning measurements. The ultrasonic sensors constantly
monitor the presence and position of the front passenger, the PODS detects the
approximate weight of the occupant. PODS can also be integrated with a seatbelt-
tension sensor to provide an airbag suppression system to meet the new federal US
guidelines.
TNO report

01.OR.BV.013.0/PM 23 April 2001 27 of46

Figure 18. Bladder based weight sensing (PODS) [32]

Like the previous example, a system developed by Bosch [33] uses a combination
of pressure sensitive seat sensors and an ultrasonic OOP sensor placed in the roof
lighting panel. This combined sensing system determines the occupancy of the
seats, the weight distribution of the occupants, and the position of the occupant
(sitting forward or well back).
Next to that pre-crash sensors are implemented that use radar to predict the
distance to the impact object, relative speed and impact angle. No precaution
measures are taken yet by the controller although in the near future it will be used
to assess the activation of seatbelt tensioners.

Figure 19. Integrated car sensor system, Bosch [33].

TNO report

01.OR.BV.013.0/PM 23 April 2001 28 of46

Occupant proximity and motion sensing

As has been shown in the two last examples, not only weight is a good parameter
to classify the occupant. Other parameters are useful as well, like the exact position
of the occupant determined by e.g. the sitting height but also the position of body
parts like the head.

Honda has developed a sensing device that determines the size of the passenger via
six sitting height sensors built into the back of the passenger seat [24]. While a
sensor determines whether the passenger's head is in the deployment path of the
airbag, the airbags operation can be cut off if necessary.

In some cases manufacturers have used the option to combine different

technologies for a co-operative occupant sensing system. On the Siemens internet
site a smart airbag system is announced with precise airbag inflation (front for
driver and front passenger/side for driver and front passenger/ side for rear
passenger). It concerns a multifunctional system that detects seat occupancy
(sensors and interface electronics in the seat), weight, seat position, distance of
occupants to dashboard, and the presence of an infant seat (using infrared sensors).

Going one step further than only determining the position, one of the latest
developments concerns dynamic occupant detection during the crash [31]. It can be
used for further optimising damage mitigation actions in various stages of the
crash.

In [9] the following types of dynamic occupant proximity and motion sensors that
are currently developed for the market are identified:

1. An 'electronic curtain' that is placed at a specified distance in front of the

airbag module. If an object passes through the curtain, a signal is returned.
2. Continuous target tracking: continuous position data is returned (e.g.
ultrasonic).

Below systems of type 2 will be discussed in more detail.

In [20] it is stated that suppliers and manufacturers are working towards a dynamic
system updating information every 10 ms to make a decision (which is quite coarse
compared to the previously mentioned 6 ms that are available for side impact
measurements). Such a dynamic system theoretically can determine that an
occupant has moved too close to the air bag (OOP), either through pre-impact
braking or through the movement caused by more minor initial impacts in a
multiple impact crash. Subsequently it should be quick enough to turn off the air
bag or determine that a low-risk deployment is appropriate.
The following types of technology have been identified, remarking that advanced
systems combine more than one sensing system for dynamical updating [20]:
1. Capacitive (electric field): the dielectric loading of an oscillation electric field
set up between sets of electrodes (e.g. placed in seat cushion, seat back,
TNO report

01.OR.BV.013.0/PM 23 April 2001 29 of46

instrument panel, head liner) is sensed. Electrical characteristics are used to

determine the presence and size of the occupant in the seat. Currently in
production.
2. Passive infrared systems: depending on detection of infrared emission from
the skin and face of occupants. A coarse resolution optical system is required
to focus the seat environment onto an infrared sensing array. Infrared
emissions must be correlated with conditions of occupancy.
3. Ultrasonic Sensing, non-imagining pattern recognition: broad beam of pulsed
ultrasonic waves to illuminate the air bag deployment zone and the seat
occupancy zone. Sensing of presence and dynamic location. Based on sonar,
pulsing a brief, inaudible signal, timing its return and calculating the distance.
Multiple transducers can be placed in instrument panel, overhead console, trim
around A-pillar, B-pillar or side roof rail. Recognise and track occupant
movement.
4. Ranging system: relying on bouncing a beam of waves (acoustic, optical,
infrared , radar) off an object and measuring their transit time from source to
target to detector. Measuring proximity of objects to the air bag. Narrow, less
than 10 degrees beams, intercept limited portion of the object.
5. Imaging systems: two dimensional maps of some reflective feature of the
vehicle interior (visual optical or infrared), to be interpreted by a computer.

Siemens has developed a 3-D object recognition sensor based entirely on

semiconductor components [33]. It is used to control the triggering and rate of
airbag deployment by monitoring the dynamic position of occupants in the
passenger compartment. Near-infrared, low-power laser pulses are directed through
the passenger compartment. Processing speed and memory are critical as data
being used must be as close to real time as possible. An example of the detected
occupant is given in figure 20.
TNO report

01.OR.BV.013.0/PM 23 April 2001 30 of46

Figure 20. Occupant surface monitoring [33]

With the technology described in this section the future smart airbag system will be
a multifunctional system that couples precise airbag inflation to detection of seat
occupancy, occupant weight, seat position, distance of occupants to dashboard,
and infant seat presence a.o.

3.1.4 Advanced protection actions

It should be noted that in this literature survey almost no information was found
about smart anticipative protection actions apart from smart restraint systems.
However a wide range of advanced protection methods triggered by by pre-crash
sensor input can be thought of as indicated below:

§ adaptive bumper position

§ automatically lifting hood (pedestrian protection)
§ automatically installing rollover bar
§ smart material properties (adapt material properties of relevant car structural
parts to characteristics of the collision partner)
§ self installing head restraint
TNO report

01.OR.BV.013.0/PM 23 April 2001 31 of46

3.1.5 Modelling
To support the development of intelligent protection systems, integrated CAE
modelling is an effective tool. In [9] a method is described to simulate the
continuous occupant tracking system based on ultrasonic position sensing. In this
study two sensors, a primary and secondary (figure 21), have been placed in the
cabin. Software has been used that allows: modelling of occupant and vehicle
dynamics, sensors, environmental factors, application of algorithms, controlling the
restraint system. This requires a combination of previously unrelated technology;
the modelling could not be done in one single software package yet. An overview
of the biomechanical algorithm that is responsible for making a decision of how to
control the airbag and the seatbelt depending on the occupant size, weight, position
and the crash pulse is depicted in figure 22.

Figure 21. Ultrasonic sensor locations [9]

Figure 22. Biomechanical algorithm [9]

TNO report

01.OR.BV.013.0/PM 23 April 2001 32 of46

3.1.6 Other sensors

Examples of other types of sensors that are relevant in the damage mitigation zone
are the following:

• safety belt use status sensors

A warning light and/or buzzer can be activated if the driver is not using the
safety belt. More common for the driver (even required [20]) than the
passenger.

• seat position sensors :

Information of these sensors give an indication of the position of the driver or
passenger. They are a surrogate for a more direct measurement of driver size
(the driver seat pulled all the way forward indicates a small person) [20].

• stability sensors:
Electronic sensors trigger an alarm and prepare the vehicle for accidents, e.g. a
rollover bar is raised in critical driving situations. This type of functionality is
incorporated in the Electronic Stability Program (ESP) that was announced by
Mercedes in 1997 [24].

3.2 Post Event State

The next generation of cars is predicted to include a wide range communication

services. Among many others in [23] emergency call services are mentioned
prominently. For post crash emergency communication information with respect to
the conditions of the occupant and the vehicle is needed.

On board biosensors can be used to assess injury levels in case the occupant can
not communicate. Also the number of occupants should be known. Subsequently or
in parallel emergency services may be called automatically.

For an effective and safe rescue operation the current state of the vehicle has to be
assessed accurately. Answers have to be given to questions like: Did all airbags
deploy? Did the vehicle roll over? What is the rest position of the vehicle? Is there
a fuel leak? Is there a fire? On board sensors and video cameras can assist for this
goal. Furthermore automatic detecting and extinguishing of fires, releasing
seatbelts, unlocking car doors, disconnecting unnecessary electrical power, shutting
off the fuel pump and activating warning lights will be helpful for rescuers.
TNO report

01.OR.BV.013.0/PM 23 April 2001 33 of46

Figure 23. Cooperation crash sensors and telematics [31]

TNO report

01.OR.BV.013.0/PM 23 April 2001 34 of46

4 Conclusions and future needs

In this report an overview of developments in the area of pre-crash sensing has

been presented. The systems have been organised in driving state: normal driving,
warning, collision avoidable, collision unavoidable and post event state. These
states belong to the collision avoidance and damage mitigation zone. A clear trend
is observed to integrate all safety features in one co-operative system. Therefore in
this survey systems that belong to all driving states have been addressed, although
pre-crash sensing mainly concerns the information that is monitored from the
warning state to the initial phase of the actual collision (dealing with the warning,
collision avoidable and collision unavoidable state).

Benefits of pre-crash sensing

• Pre-crash sensing increases safety by:

• Collision avoidance: issue a warning to the driver that can interpret the
situation and act according to it or automatic intervention on vehicle control
by actuators in the car (e.g. brake, throttle, steer).

• Collision mitigation: with a significant gain of time, critical parameters can

be monitored or calculated to adapt safety systems to the occupant and to
the crash mode. The side impact case is illustrative; without pre-crash
sensing only 6 ms are available to detect the crash and to perform safety
actions.
It can be applied to existing systems like smart air bags and seat belt
tensioners. It is also very promising for the development of advanced
protection methods like a lifting hood for pedestrian impact, automatically
installed bumpers or rollover bars and/or smart material properties.

• Providing a way to adapt safety systems to individual occupant features and

real life crash circumstances.

• Pre-crash sensing is building a bridge between active and passive safety: for
optimal pre-crash sensing performances one should combine the conventional
working fields of active and passive safety. Monitoring and processing of all
relevant information in an efficient way and making the right decision for
collision avoidance or damage mitigation will lead to an integrated automotive
safety system, which is very promising for increasing future traffic safety. E.g.
by integrating long-range cruise control and collision warning sensors with
short-range and in-vehicle sensors, pre-crash sensing systems can perform
object detection and tracking up until the actual time of impact and in the
initial phase of the impact. With a significant gain of time, critical parameters
can be calculated to assess crash type and to deploy the appropriate protection
actions.
TNO report

01.OR.BV.013.0/PM 23 April 2001 35 of46

• Pre-crash sensing can be based on current sensor technology: sensors for

collision avoidance and damage mitigation basically employ the same
technologies, that will also be suitable for future applications. Probably no
revolutionary new technique will be introduced to the market on short terms.

Technical & economical restrictions

• Cost efficiency: the automotive industry demands high cost efficiency for pre-
crash sensing technology. Even low-cost pre-crash sensor design bears too
little cost efficiency if there is no multi-functional use envisioned. A modular
approach can reduce the amount of sensors for the required integrated safety of
the vehicle.

• Sensor technology:

• For pre-crash sensing object classification functionality (providing

information about e.g. stiffness and mass of the opponent) is lacking.

• Most of the current technology has not been developed in the automotive
working field. It therefore has to be further matured to meet automotive
requirements like robustness, high volume production, ease of installation
and man-machine interfacing.

• The increase of scene information from combined applications will require

enhanced control and processing speed of data to make the right decisions.
Neural networks and fuzzy logic may become important in this respect.

• Reliability: especially in the case of autonomous vehicle intervention the

reliability of the system should be extremely high. Misjudgement of a situation
by the system followed by an inappropriate action should be prevented. No
clear indications were found yet about how to deal with fault tolerances that is
acceptable for the many parties involved.

Societal & legislative restrictions

• Driver's acceptance: in case of automatic intervention for collision avoidance

the driver's acceptance should be evaluated well. Is the driver willing to release
some of his control and authority over the vehicle, or does he only want to get
support in routine tasks? Furthermore the amount of false alarms should be
minimised to increase driver's acceptance.

• An intelligent car may make the driver less alert. This is not allowed because
the driver should always be able to take over control when the system fails.

• Legislation: there is a tendency to move the responsibility for the operation of

Advanced Driver Assistance Systems from the driver to the vehicle, or perhaps
TNO report

01.OR.BV.013.0/PM 23 April 2001 36 of46

more correctly, the vehicle manufacturer. How are these systems going to be
tested and what requirements do they have to meet? What party is responsible
when the system fails?

Based on these restrictions the prognosis of the automotive industry is that

advanced collision warning systems will be available in about 5 years. The market
introduction of active control systems however will take much more time. For
these systems first niche markets will be explored.

Future research needs to be addressed by PSN

• Advanced protection systems for damage mitigation: very limited information

was found about damage mitigation actions apart from intelligent occupant
restraint systems. This is a relatively open area for future research. Advanced
protection systems like adjustable interior and exterior car structures (e.g.
lifting hood for pedestrian impact, automatically installed bumpers, head
restraints or rollover bars, smart materials) can be addressed by PSN.

• Advanced simulation tools can be used for feasibility studies of these systems
and to support their early design stage. The use of biofidelic human body
models will increase the link to real world crash circumstances. PSN can play
an important role in this item.

• The developments in the area of sensor technology mainly with respect to object
classification should be followed closely by PSN and where possible a
contribution should be made to it.

• The next generation of communication services in cars will include many

telematics features (e.g. roadside assistance, emergency calling, real time
traffic congestion information, real-time parking booking, anti theft and
tracking, fleet management and wireless information). They will be integrated
with safety systems in on-board-computers. Eventually vehicles will be able to
communicate with vehicles and drivers with drivers. PSN should closely
follow these developments on the boundary of passive and active safety and
contribute where possible.

• PSN should link and co-operate with networks in the field of active safety,
collision avoidance and vehicle sensor technology like CHAMELEON and
SAFENET.
TNO report

01.OR.BV.013.0/PM 23 April 2001 37 of46

5 References

[1] 'Development and Validation of Functional Definitions and Evaluation

Procedures For Collision Warning/Avoidance Systems', NHTSA technical
report, August 1999.

[2] 'Development of Performance Specifications for Collision Avoidance

Systems for Lane Change, Merging and Backing', NHTSA technical report,
May 1997.

[3] D. Middleton, D. Jasek, R. Parker, 'Evaluation of some existing technologies

for vehicle detection', report number 1715-S, Texas Transportation Institute,
September 1999.

[4] Rohr S.N., Lind R.C., Myers R.J., Bauson A., Kosiak W.K., Yen H., 'An
Integrated Approach to Automotive Safety Systems', Delphi Automotive
Systems, SAE paper 2000-01-0346.

[5] Seiler P., Song B., Hedrick J.K., 'Development of a Collision Avoidance
System', SAE paper 980853, February 1998.

[6] Bowerman N., 'Why should the government be concerned with ADAS?',
ImechE paper S653/009/99.

[7] McNamara D.A., Tucker B.K., Kwiecinski R., 'Automotive Obstacle

Detection Systems: A Survey of Design Requirements & Vehicle Integration
Issues', paper number 98C021.

[8] Turner J.D., Austin L., 'A review of current sensor technologies and
applications within automotive and traffic control systems', TRL, Proc Instn
Mech Engrs Vol 214 Part D, page 607 - 613, ImechE 2000.

[9] Murad M., Burley E., Blackburn B., 'Integrated CAE Modeling of Intelligent
Restraint Systems', Breed Technologies Inc., SAE paper number 2000-01-
0606.

[10] Tanaka S., Nakagawa T., Akasu M., Fukushima Y., Bracken W.,
'Development of a Compact Scan Laser Radar', SAE paper number 1999-01-
1044.

[11] Huertgen B., Poechmueller W., Stiller C., 'Vehicle Environment Sensing by
Video Sensors', Robert Bosch Gmbh, SAE paper 1999-01-0932
TNO report

01.OR.BV.013.0/PM 23 April 2001 38 of46

[12] Thomae A., Schellin R., Lang M., 'A Low Cost Angular Rate Sensor in Si-
Surface Micromachining Technology for Automotive Application', Robert
Bosch Gmbh, SAE paper 1999-01-0931.

[13] Clute G., Zellmer H., Jawad S., 'An Intelligent Occupant Protection System
(IOPS); theoretical evaluation', Autoliv, University of Hertfordshire, Isata
paper number 99F025, 1999.

[14] Donges E., 'A Conceptual Framework for Active Safety in Road Traffic',
Vehicle Systems Dynamics, 32 (1999), pp.113-128, Swets & Zeitlinger,
1999.

[15] Grace R., Benjamin A., 'Application of a Heavy Vehicle Drowsy Driver
Detection System', SAE paper 1999-01-3754, 1999.

[16] Feige A., Stobart R., 'Plattformen für mehr Sicherheit; Systeme zur
Unfallvermeidung', Automobil Produktion; 13 (1999) 3 pp. 146-148, 1999.

[17] CHAMELEON Project IST - 1999 - 10108, Workpackage 2, 'System

Concept, Technical and Operational Requirements', Deliverables D02.2 -
D02.3, April 2001.

[18] Bowerman N., 'Why should the government be concerned with ADAS?',
IMechE conference 1999, paper number S653/009/99.

[19] Jansen G.R.M., Boedijn S.J., 'Doelfinancieringsprogramma Verkeer en

Waterstaat 2001', TNO internal report, TNO Verkeer en Vervoer, December
2000, Delft, the Netherlands.

[20] National Highway Traffic Safety Administration (NTHSA) - Preliminary

economic assessment SNPRM FMVSS no.208 - document 'V. Technology
Options and Analysis', Oktober 19999, NTHSA site:
www.nhtsa.dot.gov/cars/rules/rulings/AAirbagSNPRM/PEA/Index.html.

[21] National Highway Traffic Safety Administration (NTHSA) - SNPRM

document 49 CFR Parts 552, 571, 585, and 595 [Docket No. NHTSA 99-
6407; Notice 1] RIN 2127-AG70', Occupant crash protection, NTHSA site:
www.nhtsa.dot.gov/cars/rules/rulings/AAirbagSNPRM/PEA/Index.html

[22] FTSS presentation 'OCATD, Adult and Child Human Surrogates for Weight
Pattern Sensing', FTSS Seminar October 2000.

[23] Global Automotive Network magazine, page 8-13, December 2000.

[24] Automotive Engineer magazine, page 44-45, November 1999.

TNO report

01.OR.BV.013.0/PM 23 April 2001 39 of46

[25] Hauri S., 'Passiv, aktiv, interaktiv. Renault und die Sicherheit', Autotechnik,
48(1999)6, 1999.

[26] Advanced Transportation Technology News (ATTN) Highlights January

2000, BBC Inc. Publications, 25 Van Zant St., Norwalk.

[27] Automotive Engineer magazine, page 47-54, November 1999.

[28] 'De Ingenieur', Koninklijk Instituut voor Ingenieurs (KivI), Jaargang 112,
November 2000.

[29] MOT magazine 24/1999, page 18-24

[30] 'Intranet im Fahrzeug', ATZ Automobiltechnische Zeitschrift 101 (1999) 9,

page 630-631, 1999.

[31] 'Wege zum intelligenten Rundumschutz', ATZ Automobiltechnische

Zeitschrift 102 (2000) 2, page 114-117, 2000.

[32] Automotive Engineer international SAE magazine , page 102-107,

December 2000.

[33] European Automotive Design magazine, page 14-15, October 2000.

[34] Agustin, X.., 'Intelligent Restraint Systems. New trends', SEAT Technical
Centre, 2000.

[35] Automotive Engineer international SAE magazine, July/August 1999

[36] Moritz R., 'Pre-crash sensing - Its Functional Evolution based on a Platform
Radar Sensor', Robert Bosch GmbH, SAE paper 2000-01-2718, 2000.

[37] Stuetzler F.J., Century K., 'Advanced Frontal Crash Sensing with Peripheral
Sensors', Robert Bosch Corp., SAE paper 2000-01-2683, 2000.

[38] Klotz M., Rohling H., 'A high range resolution radar system network for
parking aid applications', Department of Telecommunications, Technical
University of Hamburg-Harburg, 1998/1999.

[39] Ploeg J., 'Mensor fase 2000 Ontwikkeling en demonstratie van een
applicatie' , TNO Automotive internal report, TNO Automotive New
Transport Systems department, 2000.

[40] CarTALK 2000 'Safe and comfortable driving based upon inter-vehicle
communication', IST programme, Preliminary description of work, February
2001.
TNO report

01.OR.BV.013.0/PM 23 April 2001 40 of46

[41] Internet site Centro Ricerche Fiat (CRF),

http://194.244.198.166/uk/activ/activ.htm
TNO report

01.OR.BV.013.0/PM 23 April 2001 41 of46

TNO report

01.OR.BV.013.0/PM 23 April 2001 42 of46

Appendix A Parallel EC projects

EC project: Chameleon:

The project deals with assisting the driver in avoiding an accident and minimising
the consequences of a crash.[17]. The project started in 2000 and will continue
until the end of 2002. The goal of this project is to support, to guide and to validate
the development of a pre-crash sensorial system necessary for near impending
crash detection all around the vehicle and in all scenarios. A subgoal is the
improvement of already existing sensor performances, their robustness and
operative range extension. In the future it will be reasonable to have few sensors in
the car devoted to many applications. Therefore an additional objective is the
investigation of possible synergies of the pre-crash sensors with other Advanced
Driver Assistance Systems (ADAS).

In the upcoming crash event a distinction is made between reversible actions and
irreversible actions. Reversible means: the action of the safety system is reversible
and does not cause damage to the system itself (e.g. pre-tensioning of safety belts).
These systems are foreseen to be introduced to the market on short term.
Irreversible means: the safety system has to be substituted after its activation (e.g.
airbag). Market introduction will be on medium term.

For the activation of (current) safety restraint systems the following steps are
addressed:
• pre-warn the safety actuators
• reduce and optimise activation time in case of danger detection
• input from in-vehicle occupant sensors to detect e.g. OOP, presence, weight

EC project RadarNet, Multifunctional Automotive Radar Network:

The project is aiming at the development of a new type of low cost radar network
for automotive applications and to prepare the realisation of first applications:
urban collision avoidance, collision warning, stop and go functionality, airbag pre-
crash warning, and parking aid.

The radar network shall consist of 4 completely new smart near distance radar
sensors (25 m range), equidistantly distributed behind the bumper, together with a
single far distance radar sensor (180 m range) implemented in the middle of the
bumper. All radar sensors will be equipped with the newly developed MMIC
chipset and transmit/receive (T/R)-modules. Far distance (FD) sensors, developed
for the first time in 77 GHz MMIC technology will provide improved performance
and higher angular coverage. By realising very precise time synchronisation of the
individually positioned near distance (ND) sensors, their output signals will be
processed using a new sophisticated multi-lateration technique. Through these
TNO report

01.OR.BV.013.0/PM 23 April 2001 43 of46

measures a target information with high resolution in range, Doppler frequency and
azimuth angle will be achieved.
TNO report

01.OR.BV.013.0/PM 23 April 2001 44 of46

Appendix B List of abbreviations

Collision Avoidance Zone:

Normal driving:
ITS: Intelligent Transportation System
ADAS: Advanced Driver Assistance System
ACC: Adaptive Cruise Control
GIS: Geographic al Information System
GPS: Global Positioning System
AVG: Automatic Vehicle Guiding
ESP: Electronic Stability Program
PRA: Parking and Reversing Aid system
ABS: Antilock Braking System
ISA: Intelligent Speed Adaptation

Warning:
ICA: Intersection Collision Avoidance system
HUD: Head-Up Display
ACA: Anti Collision Assist system
CW/CA: Collision Warning / Collision Avoidance system
ESP: Electronic Stability Program
HMI: Human Machine Interfacing
CCD: Charge Coupled Device (video system)

Damage Mitigation Zone:

IOPS: Intelligent Occupant Protection System

ECU Electronic Control Unit
WCS: Weight Classification System
TOPS: Tailorable Occupant Protection System
PODS: Passive Occupant Detection System
OOP: Out Of Position
UPS: Up-front Peripheral crash Sensor
OCATD:Occupant Classification Anthropometric Test Device

For more abbreviations with respect to ITS and car electronics see the glossaries at
the Siemens internet site:

http://www.siemensauto.com/glossaries
TNO report

01.OR.BV.013.0/PM 23 April 2001 45 of46

Appendix C Specifications of sensor types

For active and passive applications a wide series of sensor types is available. An
overview of sensors with relevant features as identified within the scope of this
literature survey, is listed below. Relevant references a.o. are [7, 8, 11, 36].

Application range Sensor type Features

Long Radar (typically 77 • Distance monitoring
(50m<d<250 m) GHz)1 • Closing velocity monitoring
• Insensitive to low visibility conditions
• Resolution increases with decreasing scanning
range 2
• No object classification
• Small lateral resolution/angular coverage (typically
10 to 16 degrees)3
• Frontal resolution typically 1m at 100m range
• Near range cut-off (typically 5m: limitation in
dense or slow moving traffic)
Laser • Pulsed mode of operation
• Better resolution than radar (typically 0.1m at
100m range)
• Typical wave length = 1,55 µm, pulse duration =
50 ns
• Sensitive to mist, dust, rain, road spray (may be
overcome by combination with infrared)
Infrared • Better resolution than radar
• Sensitive to mist (dust, rain to some extent)
• Demonstrated applicability for vehicle-vehicle
communication
• Sensitive to environmental light intensity (sun
light)
Medium Imaging sensors • 2-D maps of reflective features of the vehicle
(20m<d<50m) (typically video • Good lateral coverage (typically 40 degrees)
optical systems or • Object classification and pattern recognition
infrared) possible with image processing software
• Object dimensions monitoring
• Monitoring of road signs and lane borders
• Current CCD technology not robust enough for
automotive applications
Laser • Scanning 360 degrees
• Typically mounted on the right rear car corner4
Radar • Typical configurations include a 360 degrees
sweep at a radius up to 20m
TNO report

01.OR.BV.013.0/PM 23 April 2001 46 of46

Short Radar (pulse echo) • Typically a 100 degrees beam angle at a radius up
(2m<d< 20m) to 14m at a frequency of 24 GHz
• To be supplemented by Doppler functionality to
measure closing velocity
Passive infrared • Detection of infrared emission from objects with
(PIR) respect to the environment
• Typical range up to 25m and field of view of 10-20
degrees
• Relatively expensive for automotive use
Imaging sensors See above
Direct vicinity of Microwave radar • Penetrates plastic coverings easily: so can be built
the vehicle (< 2m) into the car
• Can not penetrate sheet metal
• Typically 300 ps electromagnetic pulses
PIR See above
Ultrasonic • Monitor object's presence, distance, moving
direction
• Restricted to low speed manoeuvring5
• Already introduced to the market in parking aid
systems
• Relatively low costs
Capacitive • Based on monitoring dielectric properties different
from air between poles (oscillating electric field)
• Size and shape detection of objects
• Low costs
• Problem with radio frequency interference
In-vehicle (dyn. Ranging systems • Operates by bouncing a beam of waves (acoustic,
occupant sensing) optical, infrared, radar) off an object and
measuring their transit time from source to target
to detector
PIR • Detection of infrared emission from the skin and
face of occupants
• Coarse resolution is required distinguish the
occupant from the seat environment
Capacitive See above
Imaging systems

1
The propagation time and the frequency shift of the reflected pulse determine the
distance and relative velocity of the detected object.
2
This is a general specification of radar systems that applies to all application
ranges
3
Angular coverage can be improved by using a set of adjacent sensors. This
applies to the situation where the sensors are placed behind plastic car components
to make them invisible from the outside (for design reasons).
4
In [2] an example is presented that concerns a laser that is able to generate
trajectories of passing cars and categorise them according to their length and width.
TNO report

01.OR.BV.013.0/PM 23 April 2001 47 of46

5
Due to the relatively low speed of sound in air compared to the propagation speed
of e.g. radar pulses.

Future technology needs are directed towards the modification of different sensor
types for automotive application with respect to:
• Robustness
• Ease of installation
• Possibilities for high volume production
• Decrease of manufacturing/installation/maintenance/reparation costs
• Use of a limited amount of sensors for multiple applications