Measurement campaign at JARI, Japan boosts development of automated driving on motorways
From 4 to 8 December 2023, Mercedes-Benz and VTT conducted an extensive testing campaign at the Japan Automobile Research Institute (JARI) Tsukuba fog chamber. It was the third measurement campaign of the project, thus a significant step towards a crucial evaluation and validation milestone of the AI-SEE platform.
The 200-meter-long fog chamber enables the testing of environment perception devices over long distances, comparable to the range of high-beam headlights. It can produce stabilized fog and rain under various lighting conditions and hence offers excellent conditions to accurately simulate adverse weather conditions, which are typically represented as constants in simulations.
The primary objective of the testing campaign was to collect comprehensive ground truth data using a synchronized, multimodal long-range sensor array designed for SAE Levels 3-5 automated driving. This data collection focused on challenging environmental conditions such as heavy fog and rain. The main goal was to establish a benchmark for assessing the maximum capabilities of these sensors and to generate reference data crucial for developing accurate simulation models.
The test encompassed the three critical use cases identified in the AI-SEE project:
- Lost Cargo Detection: Testing sensor efficacy in identifying obstacles or hazards on the road under low visibility conditions.
- Traffic Jam Detection: Assessing the ability of the sensor array to detect and respond to congested traffic situations amid changing weather.
- Pedestrian Detection: Evaluating the sensors' precision in detecting pedestrians in adverse weather, which is crucial for ensuring safety.
In total, 13 different LiDAR systems, 2 radars, and 3 types of cameras were tested by Mercedes-Benz and VTT under the above-described realistic conditions.
The test was funded as part of the VVM project in Germany and part of the AI-SEE project in Finland, which is co-funded by Business Finland. The aim of VVM is to determine measurement methods for determining the performance of sensors and perception stacks under the influence of inclement weather conditions. AI-SEE and the VVM project are conducting the same test scenarios and are sharing the data.