AVEC ’12

Integral Simulation Toolset for

Design and Evaluation of ADA Systems
Martijn Tideman (TASS), Roy Bours (TASS)
Tino Schulze (dSPACE), Takahito Nakano (DENSO)

Corresponding author: Martijn Tideman

2301 Sino Life Tower, 707 Zhangyang Road
200120 Shanghai, CHINA
Phone: +86 21 5835 8036
Fax: +86 21 6085 3094
E-mail: [email protected]

This paper outlines the use of two complementary software packages that together provide the
ability to cover all critical aspects of advanced driver assistance (ADA) system design. One of the
software packages (PreScan) focuses on the design of the sensing and control systems, as well as
the evaluation for a wide range of traffic and weather scenarios. The other software package
(ASM) takes care of the detailed vehicle dynamics and vehicle control. By means of a real-world
application example of a Lane Keeping Assistance system, the combined use of these software
packages to design and evaluate ADA systems is demonstrated.

Topics: Active & Passive Safety Systems, Vehicle Dynamics, Modeling and Simulation Sensors and Actuators

1. INTRODUCTION roads. Prototypes are expensive and unavailable early in

the design process when changes can be made relatively
The automotive industry is developing advanced easily and at low cost.
driver assistance (ADA) systems and active safety Virtual testing with simulation software provides an
systems that combine pre-crash sensing, driver warning efficient and safe environment to design and evaluate
(acoustic, haptic or visual) and vehicle dynamics control ADA systems. Moreover, simulated scenarios are
(braking, steering). The typical ADA system completely quantifiable, controllable and reproducible.
architecture is illustrated in Figure 1. Virtual testing is already commonly used in the design
and evaluation of vehicle dynamics as well as crash
safety systems. Recently, in the development process of
ADA and active safety systems, simulation is becoming
more popular as well. This paper describes how
simulation software can be usefully applied within the
design and evaluation of ADA systems.
In Section 2, general requirements for ADA system
simulation software are described. Also, two
Fig. 1 Typical ADA system architecture commercially available software tools are introduced
that jointly provide the required simulation capabilities
Designers of ADA systems face many challenges. to cover the critical aspects of ADA system design.
The involvement of different engineering disciplines – Section 3 illustrates how these software packages can be
including sensing, electronics, control, vehicle dynamics, combined to design and evaluate ADA systems. An
and safety – results in a complicated design process. application example for Lane Keeping Assistance is
Another challenge is that ADA systems need to function presented in section 4. The results of this example are
properly for the vast amount of different traffic described in section 5. Section 6, finally, contains
scenarios which can occur on today’s roads including concluding remarks.
effects of weather circumstances, road conditions and
traffic density. This not only puts very stringent 2. ADAS & ACTIVE SAFETY DEVELOPMENT
demands on system performance and reliability but also
makes testing and verification complicated and 2.1 Traditional testing methods
time-consuming. Another complicating factor in the Testing methods for passive safety systems are
verification of ADA system designs is the need to test in extensively defined and well-established. Regulations,
reproducible conditions without the risk of endangering such as FMVSS and ECE, and consumer programs,
people. This factor often rules out field tests on public such as NCAP and IIHS, prescribe testing protocols in
 AVEC ’12

great detail and these testing methods are extensively 2. Testing methods are time-consuming resulting in a
applied in the development and certification of passive limited coverage of the wide variety of traffic
safety equipment. Manufacturers have worked for many conditions.
years on efficient development processes which include 3. The reproducibility of the test scenarios in
simulation, sled testing and full system crash testing. test-track testing is limited, which makes it difficult
In contrast with passive safety, testing protocols for to compare different systems in an objective way
active safety and ADA systems are hardly regulated and 4. Many ground truth reference data cannot be
consumer protocols are not established. Only now, measured (accurately) during test drives.
organizations as Euro NCAP are starting initiatives to 5. Critical scenarios are often too dangerous to test.
award active safety systems. Euro NCAP has recently
announced Euro NCAP Advanced, a reward system and 2.3. Virtual testing of ADA systems
standardized assessment procedure for new safety Many of the drawbacks of hardware testing of ADA
technologies. Unlike Euro NCAP’s well established systems as described in the previous sections are not
assessments involving physical tests at a crash present for a virtual test environment. Virtual testing
laboratory, the new process is based entirely on the with simulation software provides an efficient and safe
assessment of scientific evidence presented by the environment to design and evaluate ADA and active
vehicle manufacturer and no test protocols are safety systems. Moreover, simulated scenarios are
prescribed. completely quantifiable, controllable and reproducible.
Several initiatives are working on developing Virtual testing is already commonly used in the design
standards describing system requirements and standard and evaluation of vehicles dynamics as well as crash
test programs. Some examples are the ISO standards [1], safety systems. Recently, in the development process of
the Crash Avoidance Metrics Partnership initiative ADA systems, simulation is becoming more popular.
(CAMP) and NHTSA confirmation test requirements. In Virtual testing of ADA systems requires software that
Europe, the Beyond NCAP group from EURO NCAP gives realistic representations of:
[2] and European projects such as PReVENT [3],  Road objects (e.g. roads, lanes, sidewalks, curbs,
Interactive and ASSESS [4] are working on light poles, traffic signs)
standardization of test programs.  Road users (e.g. pedestrians, bicyclists, cars, trucks)
In recent years, the automotive industry has  Environmental conditions (e.g. rain, snow, fog,
developed various testing procedures for ADA systems, lights, reflections, shadows)
ranging from test track protocols, pre-crash tests with  Sensing technology (e.g. camera, radar, lidar,
stationary crash-forgiving foam dummy vehicles [5], infrared, ultrasonic, antennas, GPS)
plastic vehicle shapes attached to driving cars [6] and  Vehicle dynamics (e.g. braking system, steering
the use of vehicles that are driven by robots [7,8]. Some system, suspension)
manufacturers have even developed small test villages Many perturbations of these different characteristics
as shown in Figure 2. need to be simulated in order to assess the ADA
system’s reliability and robustness for the large range of
conditions that can occur on real roads. This requires
that the simulation models are easy to create and modify.
Also, fast simulation times are required to ensure that
the vast of amount of simulations can be executed
within the limited time available.

2.3.1 ADA system modeling

PreScan is a physics-based simulation tool for the
design and evaluation of intelligent vehicle systems,
such as active safety systems, ADA systems and
intelligent transportation systems [10].
The PreScan user interface allows users to build
Fig. 2 Example of a pre-crash test in a test village [9] traffic scenarios using a database of road sections,
infrastructural components (trees, buildings, traffic
2.2. Challenges for ADA system testing signs) and road users (cars, trucks, cyclists, pedestrians).
The experimental methods for ADA system Weather conditions (rain, snow, fog) and light
development have a number of drawbacks: circumstances can be modeled as well. The vehicle
1. Experimental methods require the availability of models can be equipped with models of different sensor
system prototypes. Prototypes are expensive and types, including radar, laser, camera, ultrasonic, GPS
unavailable early in the design process when and antennas for V2X communication. Sensor design
changes can be made relatively easily and at low and benchmarking is facilitated by easy exchange and
cost. Therefore, problems are found late in the modification of sensor type and sensor characteristics.
development process and errors are expensive to An interface with MATLAB/Simulink enables users to
fix. design and verify algorithms for sensor data processing,
 AVEC ’12

sensor fusion, risk estimation, decision making and 3. PRESCAN – dSPACE/ASM INTEGRATION
control (see figure 3).
Integration of both software packages makes it
possible to investigate the robust functioning of
closed-loop ADA system during complex vehicle
maneuvers and detailed vehicle behavior.
Therefore, a unique integration has been established
between PreScan and dSPACE/ASM which allows
closed-loop simulation of ADA systems with automated
braking or steering actions. The vehicle motion is
calculated in the dSPACE/ASM model. Updated vehicle
positions and orientations are made available to the
PreScan ego vehicle. Each time step, the world model is
updated and new sensor signals are sent to
MATLAB/Simulink for signal processing and actuation
Fig. 3 Open-loop ADA system modeling using PreScan algorithm decisions (see Figure 5).

2.3.2. Vehicle Dynamics modeling

For ADA systems which control the vehicle’s
steering, braking or throttle actions (e.g. lane keeping
assistance or emergency braking), a detailed model of
the vehicle chassis and chassis controls are essential
aspects of the system. dSPACE Automotive Simulation
Models (ASM) is a simulation toolsuite offering a
package for vehicle dynamics simulation. The package
is an open Simulink® model specially designed for
Hardware-in-the-Loop (HIL) testing of ECUs - as well
as for the controller design phase. Since the
implementation of ASM Vehicle Dynamics, as depicted
in Figure 4, is an open Simulink model, additional
components such as an interface to PreScan can be
added easily.

Fig. 5 Closed-loop ADA system modeling using PreScan and

dSPACE/ASM software (overview)

3.1 Driver modeling

The ASM vehicle model is controlled by (i) a
lateral controller resembling the steering behavior of the
driver and (ii) a longitudinal controller for the throttle
and braking behavior of the driver.
The longitudinal controller (driver model for
velocity control) is a Simulink block which translates a
target velocity into throttle, clutch and braking signals
for the vehicle model. The standard longitudinal
controller in ASM is applied, and fed with a target
velocity from the PreScan trajectory definition.
Fig. 4 Chassis system and control modeling using The lateral controller (driver model for steering
dSPACE Automotive Simulation Models actions) is a Simulink block which translates a trajectory
into a steering angle input for the vehicle dynamics
The dynamics of the vehicle under test are model. The standard lateral controller from PreScan is
described as a multi-body system which yields a set of applied. The lateral controller uses a 10-point preview
differential equations of 2nd order. The necessary and is based on the optimal preview control theory as
physical abstraction to model the relevant dynamics for published by MacAdam [11,12]. By introducing future
ADA system function development and testing requires path inputs, tracking models with preview information
10 degrees of freedom (DoFs), six for the body give much better tracking accuracy than 1-point
translation/rotation and four for the wheels. Furthermore compensation tracking model.
the relevant kinematics and compliances and an
appropriate tire modeling is necessary to simulate the
dynamic behavior of the vehicle under test with an
appropriate accuracy.
 AVEC ’12

4.3. LKA model

The simulation tool set as described in section 3
was used to model a DENSO LKA system (see Figure
7). Corresponding to the descriptions in section 4.2, the
LKA algorithm (which is modeled in Simulink)
calculates the required steering torque based on the lane
information derived from PreScan. Input signals used
are, for example, “distance to line” and “road curvature”.
The steering torque is then fed into a steering model in
ASM to prevent departure from the intended lane.
Finally, arbitration between the lateral driver model and
the LKA model is implemented to make sure that - after
the LKA model is activated – the input from the lateral
driver model is deactivated.

Fig. 6 Closed-loop ADA system modeling using PreScan and

dSPACE/ASM software (detailed)
TA

3.2 Frequency subcycling

High-fidelity vehicle dynamics simulations are DN-MWS
typically performed at 500-1000 Hz, whereas an ADA LKA
system typically runs at frequencies between 10 and 100
Hz. To facilitate this difference, PreScan and ASM are ASM
able to run at different frequencies. Within a typical
ADAS setup, the vehicle simulation in ASM runs at
1000 Hz, whereas the world simulation and sensor
simulation within PreScan run at 10 - 100 Hz. However, Fig. 7 Simulink representation of the LKA model
both the longitudinal and lateral driver models require including ASM vehicle dynamics, test automation (TA),
input from PreScan with 500-1000 Hz. Therefore, LKA algorithm and PreScan sensor inputs
PreScan still generates the target path and target
velocity with a frequency of 500 Hz. This process of Simulation of ADA systems allows fast analysis of
running the various PreScan modules at different many different scenarios using automated parameter
frequencies is called “subcycling”. variations. Two important points had to be addressed for
the LKA model:
4. LKA APPLICATION 1. The performance of the lateral driver model to
accurately control the ASM vehicle dynamics
4.1. Introduction model. The PreScan lateral controller using a
By means of a real-world application example of a 10-point preview was optimized for the specific
Lane Keeping Assistance (LKA) system, the combined vehicle under test.
use of these software packages to design and evaluate 2. An efficient method for scenario generation. The
ADA systems is demonstrated. lateral driver model was extended with an offset
function allowing a constant or runtime-changeable
4.2. LKA system description lateral offset from the defined target path.
A Lane Keeping Assistance (LKA) system detects Due to these two modifications, it was possible to
lane markers on the road, and assists the driver’s execute an automated test series of several hundred
steering to help keep the vehicle between lane markers. scenarios to verify the LKA performance.
When the system detects the vehicle straying from its
lane, it alerts the driver visually as well as with a buzzer, 5. RESULTS
while applying a slight counter-steering torque, trying to
prevent the vehicle from moving out of its lane. An Using the LKA model as described in section 4, the
LKA system consists of two key components: (1) a LKA system’s basic functionality and performance is
vision sensor, which detects shapes and positions of verified. Figure 8 depicts a scenario on a straight road,
lane markers, and (2) a steering assist electronic control in which the driver is leaving the lane. The blue line
unit (ECU), which calculates a target steering torque shows the driver’s steering input, whereas the red line
based on the data from the vision sensor. The steering shows the correction made by the LKA (thereby
assist ECU then sends this steering torque signal to an overruling the driver). It is shown that, in this scenario,
electric power steering (EPS) ECU to control the EPS the LKA works and that the vehicle gets back into its
motor. Furthermore, the steering assist ECU determines lane. It can also be seen that there is a slight deviation
if it is necessary to alert the driver about the vehicle’s between the target path as input to the driver model
deviation from the lane. These alerts are communicated (yellow line) and the actual path of the vehicle as
to the driver by a buzzer and a visual signal.
 AVEC ’12

controlled by the driver model and/or the LKA [5] G. Sala, Safety, legal issues, standards, Deliverable
controller (purple line). D5.01, European Commission, CHAMELEON Project,
Brussels, Belgium, January 28, 2003. Available online
at: www.crfproject-eu.org
[6] N. Fecher, et al., Test and Evaluation Methods for
Safety Systems on the Guidance Level,
Automatisierungstechnik 56, pp. 592-600, 2008.
[7] O. Gietelink, et al., Development of advanced driver
assistance systems with vehicle hardware-in-the-loop
simulations, International Journal of Vehicle System
Dynamics, 44(7), pp. 569–590, July 2006.
[8] H-P Schöner, et al., Testing and verification of active
safety systems with coordinated automated driving,
Proceedings of the 21st International Technical
Conference on the Enhanced Safety of Vehicles ESV,
2009.
[9] Autoliv website: www.autoliv.com

[10] PreScan website: www.tass-safe.com/prescan

[11] MacAdam, C.C. ‘An optimal Preview Control for

Fig. 8 Example results from the LKA simulation model Linear Systems, ”Journal of Dynamic Systems,
Measurement, and Control, ASME, Vol. 102, No. 3,
6. CONCLUSION Sept. 1980.

Advanced driving assistance (ADA) systems need [12] MacAdam, C.C. “Application of an Optimal
to function properly in a vast amount of different traffic Preview Control for Simulation of Closed-Loop
scenarios. This makes the development process of ADA Automobile Driving,” IEEE Transactions on Systems,
systems a complicated, dangerous and time-consuming Man, and Cybernetics, Vol. 11, June 1981.
activity. Simulation software can help to overcome
these difficulties by providing an efficient and safe
environment for designing and evaluating ADA
systems.
PreScan, dSPACE/ASM and MATLAB/Simulink
form together a simulation tool chain that is suitable for
virtual design and assessment of ADA systems such as
ACC, CMB, PCS, LDW and LKA. This tool chain is
capable of assessing the effects of the performance of
all ADA system components (such as sensors, object
detection & interpretation algorithms and vehicle
dynamics control algorithms) on the safety performance
of the complete ADA system. Unique in the
methodology is the fact that different system-elements
can be assessed using the same tool-set.
The use of the simulation tool chain was presented
based on a real-world application example: DENSO’s
LKA system. It was described how the simulation tool
chain is used to achieve a efficient and effective ADA
system development process.

REFERENCES

[1] International Organization for Standardization

website: www.iso.org
[2] Euro NCAP website: www.euroncap.com
[3] EC project PReVENT website: www.prevent-ip.org
[4] EC project ASSESS website: www.assess-project.eu