Issues in validating sensor functionality
Currently, many vehicles are equipped with sensors for ADAS functions, and it is expected that the number of sensors will continue to increase in the future to realize automated or autonomous driving functions. We believe that functional validation of these sensors will require a complex test environment for the vehicle as a whole, not just the control target as in the past for basic control. There are several challenges, especially in real-world driving tests.
First, the real world is very complex, and reproducing traffic participants, natural environments, etc. is very complicated, and it is not possible to have exactly the same timing of traffic participants coming and going, and at the same time have the same sunlight, cloud cover, etc. as in the previous test. Second, the real world is unpredictable. It is sometimes necessary as a validation, but it is not reproducible and can be very risky for the test driver. For example, falling rocks are difficult to artificially generate in the real world, even when attempting to test them, and they put the test driver at risk. Finally, real-world driving tests in the real world are fraught with various hazards, and there will be some test items that cannot be conducted to avoid all of them. Even if there is a child jumping item, we cannot actually let children jump.
Thus, there are many challenges in trying to establish a real-world test environment. Therefore, dSPACE collaborated with Ono Sokki Co., Ltd. to investigate the feasibility of constructing a sensor simulation environment (VIL: Vehicle in the Loop) using a real vehicle. In this way, the inspection items of the actual driving test could be performed as much as possible in the simulation to reduce risks and ensure the reproducibility of the scenario. In addition to reducing risk, we also investigated the feasibility of reproducing the scenario.
Sensor Simulation
There are several methods for sensor simulation. The figure shows our idea of sensor processing. First (from left to right) is sensing, followed by perception, fusion, and application processing to move actuators. This would be the basic flow of a sensor based function. However, since the simulation is based on assuming that it is equivalent to an actual driving test, it was decided to build a "front end" simulated environment using a full vehicle. Because the vehicle prepared for the VIL environment construction is equipped with a front camera and front radar, a monitor was installed in front of the vehicle windshield and DARTS (dSPACE Automotive Radar Test Systems ) was installed in front of the front-end to create a sensor simulation environment.
VIL configuration
For the VIL environment considered in this study, the environment corresponds to the normal vehicle environment up to the point at which the vehicle is connected to the test bench.In this case, the results of the AURELION camera sensor simulation for the front camera of the entire vehicle were displayed on a monitor . For the front radar, the external environment was simulated with DARTS (dSPACE Automotive Radar Test System). These signals are generated based on the simulation results of the ASM (Automotive Simulation Model, vehicle simulation model). A real-time closed-loop environment was created by using SCALEXIO (real-time simulator) to link with the test bench. In addition, the latitude and longitude information that can be generated by ASM can now be used to simulate GNSS signals. This means that car navigation systems can be simulated at the same time.
For maneuverability, the tie-rod ends of the front wheels were removed to free the movable parts at the ends of the steering rack to allow steering operation while the test vehicle was held in place, and a wire encoder was used to measure the amount of steering rack movement and provide feedback to the vehicle simulation model. The wire encoder measures the movement of the steering rack and provides feedback to the vehicle simulation model. An environment was created with no restrictions on the angle and angular velocity of the steering angle, an environment was created that is fully adaptable to steering maneuvers during right and left turns and sudden steering maneuvers for emergency avoidance.
Furthermore, the creation of a driving actuator allows for a quantitative and automated testing environment, addressing real-world evaluation challenges. The driving actuator can be linked to the ASM driver model and can also be operated remotely via SCALEXIO.
VIL validation results
The video is showing the results of the VIL environment construction. This time, a commercial vehicle was used for the validation with no detailed sensor specifications or sensor outputs available. Therefore, it was possible to check the OBD monitor and the vehicle's instrument panel information, and to confirm the collision mitigation braking, lane departure detection, and lane keeping assist. This proved that not only vehicles under development, but also finished vehicles after coming off the line could be tested in this environment.
We are convinced that we were able to build a "practical" VIL environment for automated driving by providing a sensor simulation environment that can simulate external appearance, an environment that allows unrestricted steering operation necessary for AD/ADAS, and an environment that allows quantitative operation and automation by driving actuators. However we are not going to stop here. Future enhancements will focus on adding functions requested by the market, such as reproduction of steering reaction force, which is currently not reproduced, and support for corner radar sensors.
Author of the paper
• Akiyuki Kozuki
Business Prototyping Group, CX Technology Department, dSPACE Japan Corporation
Hiroki Hanaoka
Business Development Group, CX Technology Department, dSPACE Japan Corporation
Shota Abe
HIL1 Group, Application Engineering Department, dSPACE Japan Corporation
Koichi Matsumoto
East Sales Engineer Group, Technical Sales Department, dSPACE Japan Corporation
Takuma Nagashio
MBD Promotion Group, Model Based Design Block, Development and Design Division, Ono Sokki Co.
Kosuke Maesai
MBD Promotion Group, Model Based Design Block, Development and Design Division, Ono Sokki Co.
