Scenarios under Control

Event based Testing of Advanced Driver Assistance Systems with CarMaker

Driver assistance systems provide an enormous potential for the improvement of active safety. However, the multitude of possible scenarios makes high demands on the system development. This article will present solutions to identify the most important scenarios and to reconstruct them with the simulation environment of CarMaker. This enables the user to systematically analyze and optimize security and performance of advanced driver assistance systems.

Unfortunately, we could not find the current Flash player on your computer. You can download it here: http://get.adobe.com/flashplayer: Simulation of the complex control loop of a driver, vehicle, environment and traffic

The development of advanced driver assistance systems requires a seamless method and tool chain for simulation and test driving. Using it, the functions of the systems have to be evaluated and optimized in all stages of the development process in as realistic as possible test scenarios. The greatest challenge for this is to manage the huge number of possible scenarios arising from the interactions of driver, vehicle, traffic situation and environment. The most important scenarios have to be identified as test cases and reproduced in simulation.

Unfortunately, we could not find the current Flash player on your computer. You can download it here: http://get.adobe.com/flashplayer

The vehicle dynamics software CarMaker offers a suitable solution for the systematic test of driver assistance systems such as ACC, braking assistant, pre-crash system, lane change assistant or parking assistant.The traffic module IPGTraffic simulates configurable traffic situations with an arbitrary number of moving objects (e.g. traffic, pedestrians) and stationary obstacles (e.g. parked cars, crash-barriers, traffic signs and buildings) surrounding a test vehicle equipped with sensors.

The control systems and the dedicated sensors can be integrated easily and quickly into the virtual test vehicle within Simulink or optionally on a C code level. Depending on the development stage and the test objective they are tested by model-, software- or hardware-in-the-loop in different driving situations and traffic scenarios.

Maneuver and Event based Testing

The working method to reconstruct the test cases within CarMaker can be described by the term ?event based testing?. This method builds on the maneuver based testing which is used as a standard method for the test drive on the track as well as in simulation with CarMaker. For the maneuver based testing, driving maneuvers such as µ-split braking, ISO lane changes or the steady state circle are split into single mini-maneuvers which are exactly reconstructed in simulation and test.

Unfortunately, we could not find the current Flash player on your computer. You can download it here: http://get.adobe.com/flashplayer: µ-split braking maneuver

CarMaker supports this method by providing a flexible maneuver control, which allows the user to arbitrarily combine open-loop and closed-loop maneuvers of lateral and longitudinal dynamics to a test sequence.

Mini-Maneuver	Driving / Action instruction	Longitudinal dynamics	Lateral dynamics
0	Acceleration to 100 kph and approaching the µ-split area	Closed Loop IPGDriver	Closed Loop IPGDriver
1	At route point „S1“ Free rolling	Open Loop	Closed Loop IPGDriver
2	At route point „S2"? Full braking and holding the steering wheel to find out stability in the window of 1s (remark: drag torque control)	Open Loop	Open Loop
3	After 1 s Steering correction to the desired course to find out the steering angle demand	Open Loop	Closed Loop IPGDriver
4	Approaching 50 kph Action: Wheel speed defect caused by DVA signal intervention (remark: fail-safe algorithm)	Open Loop	Closed Loop IPGDriver
5	Braking to stop	Open Loop	Closed Loop IPGDriver
6	Automatic readout of the failure memory

Reconstruction of test scenarios

This structured testing method has been enhanced to the event based testing of driver assistance systems. This enables the easy and rapid reconstruction of comprehensive scenarios consisting of specific constellations of traffic situation and driving maneuvers.

Analogous to the maneuver control of the test vehicle the user assigns a freely definable list of mini-maneuvers to each object of the traffic control. They contain e.g. velocities, accelerations or sinus waves in longitudinal or in the lateral direction. The initial conditions for the objects such as the starting position on the track, initial velocities or their appearance can be defined freely.

: Event based Testing

The maneuvers of the test vehicle and the traffic objects are controlled by a central control unit. It establishes a connection between the maneuvers of the traffic and the test vehicle which enables the triggering of the mini-maneuvers of the traffic objects dependent on the maneuvers of the test vehicle. A further advantage is that the central control unit allows the user to set time, distance, route and event based triggers which can be used to start mini-maneuver sequences or mini-maneuver jumps. A central state observer monitors simulation variables selected by the user and evaluates them synchronously to the simulation program with regard to defined conditions. This allows the user to choose e.g. events such as the approaching of a route point, the distance between two vehicles or the difference velocity between the test vehicle and leading vehicle as a trigger for mini-maneuvers.

These mechanisms let the user generate scenarios that are always identical relative to the test vehicle. This feature differentiates the presented solution clearly from ordinary pre-processing approaches where the simulation procedure is calculated in advance. Using a pre-processing approach test scenarios can only be defined after numerous pre-simulations for the determination of the starting positions and velocities of the traffic objects and the test vehicle. Maneuvers defined in this manner would only be reproducible for one single model parameterization. If vehicle parameters, velocities, roads or set-ups of the controller were changed, the pre-tests would have to be performed once again. In contrast, thanks to this chosen approach the trigger has to be defined only once and the scenario can be reproduced accurately under the variation of the vehicle and controller parameters, velocities or road conditions.

TestWare/ADAS

In order to give the user a quick start in the model based testing of driver assistance systems, the upgrade package TestWare for Advanced Driver Assistance Sytsems, TestWare/ADAS, was integrated into CarMaker. It contains test cases, test automation and post-processing routines to comprehensively test driver assistance systems by a mouse click.

A test and evaluation catalogue that was developed by TÜV SÜD Automotive from real driving tests provided the basis for the identification of the most important test cases of TestWare/ADAS. It contains realistic scenarios which have been developed by means of findings from the risk analysis and the functional safety analysis. This basis test catalogue was enlarged by numerous scenarios that are hard to realize on the test track or in public traffic. These are, for example, security-sensitive driving situations or tests that could not be realized under reproducible conditions due to the complex interdependencies between the test vehicle and the traffic.

Using the models and functions described above, the developed maneuver catalogue has been fully implemented in CarMaker.

Short abstract of implemented test scenarios

: Column driving acceleration

: Column driving sheer out

: Column driving deceleration

: Stop and Go

: Column driving sheer in

: Oncoming cars

A sensor module enables the positioning of an arbitrary number of sensors at arbitrary spots of the virtual test vehicle in order to simulate e.g. close-range, long-range or parking sensors. The sensor conditions such as beam width, reach and clock rate of the sensor rays can be specified freely. Once activated, the sensor model scans the sensor area starting from a reference point at the vehicle for objects and is also visible in the 3D animation, IPGMovie.

The user can choose and configure specific test scenarios via a graphical user interface within CarMaker and can perform the tests fully automatically. The user specifies e.g.:

Different velocity and acceleration profiles for the test vehicle and the traffic objects
The lane changing dynamics of the test vehicle and the traffic
Setting of the maneuver trigger
Setting of the control system
Road configurations

: Graphical user interface of TestWare/ADAS

The simulation results are collected in automatically generated test protocols. They show important signal flows, such as distances, velocities and accelerations as a function of time and distance in tables and graphs.

Due to the high number of calculated values the relative vehicle momentums (relative position, relative velocity, relative acceleration) are additionally clustered to qualifiable characteristics with an automated post-processing routine, based on experiences from vehicle dynamics. They reflect concisely the system characteristics and make the comparison of the results easier. In addition to maxima and minima, the following provide pertinent characteristics: window evaluations, particularly, reaction and response times, vehicle distances at system reaction and over-/undershooting.

Conclusion

The presented simulation environment allows for the development and testing of advanced driver assistance systems under realistic conditions. Real driving situations are reconstructed easily and quickly with CarMaker. This enables the testing if driving situations are recognized reliably and if the optimal control strategy is used. The TestWare/ADAS contains important test scenarios for simulation. Using it, tests can be performed by a mouse click without having to worry about the programming of traffic scenarios. The results are provided in an automatically generated test protocol after the simulation. The presented solution can be used continuously for model-, software- and hardware-in-the-loop testing.

Login