Autonomous Driving (AD) and Advanced Driver Assistance Systems (ADAS) are being actively developed to prevent traffic accidents. As the complexity of AD/ADAS increases, the number of test scenarios increases as well. An efficient development process that meets AD/ADAS quality and performance specifications is thus required. The European New Car Assessment Programme (Euro NCAP®1) and the Japan Automobile Manufacturers Association (JAMA®2) have both defined test scenarios, but some of these scenarios are difficult to carry out with real-vehicle testing due to the risk of harm to human participants. Due to the challenge of covering various scenarios and situations with only real-vehicle testing, we utilize simulation-based testing in this work. Specifically, we construct a Model-in-the-Loop Simulation (MILS) environment for virtual testing of AD/ADAS control logic.When the collision scenarios defined by Euro NCAP in detail with collision condition, are utilized in simulation-based testing, problems arise due to the difficulty of manually generating scenarios in which the collision position and collision timing need to be adjusted.In this paper, we therefore propose an automatic scenario generation method that satisfies the intended collision condition, by means of inverse calculation that adjusts the collision position and timing automatically. This process consists of two steps. First trajectories are obtained by conducting simulations, utilizing the input with inputed scenario component and parameter variation based on the Euro NCAP. Second, the target’s position offset and timing offset that satisfy the ego car’s collision condition, are calculated from the trajectories. By applying position offset and timing offset to the base scenario, collision scenarios are generated. We also came up with a function that validates the Euro NCAP measurement error for thegenerated scenarios. Our automatic scenario generation method results in a collision error rate of about 0.2 cm for 98.02% of the Euro NCAP scenarios we tested. By utilizing the proposed scenario generation and error validation, the labor hours is expected to be reduced by up to 1/6 compared to the manual process. Our method is also applicable to custom scenarios that require precise adjusting of collision positions and timing.In the EuroNCAP test scenarios, the target’s speed and acceleration are both constant, but for AD/ADAS robustness testing, irregular behavior scenarios are required. We therefore added disturbances to the collision scenario by traffic flow simulation, where multiple vehicles are generated as disturbances and the target’s irregular behavior under the influence of multiple vehicles is generated.In this work, EuroNCAP and JAMA test scenarios became able to generate automaticaly, in addition collision target's irregular behavior can be considered for AD/ADAS robustness testing.


    Access

    Check access

    Check availability in my library

    Order at Subito €


    Export, share and cite



    Title :

    Automatic Scenario Generation for Simulation-Based Testing of AD/ADAS


    Additional title:

    Sae Technical Papers


    Contributors:

    Conference:

    WCX SAE World Congress Experience ; 2023



    Publication date :

    2023-04-11




    Type of media :

    Conference paper


    Type of material :

    Print


    Language :

    English




    Automatic Generation Method of Test Scenario for ADAS Based on Complexity

    Yang, Cai / Gao, Feng / Duan, Jianli et al. | SAE Technical Papers | 2017


    Causal Models to Support Scenario-Based Testing of ADAS

    Maier, Robert / Grabinger, Lisa / Urlhart, David et al. | IEEE | 2024

    Free access

    AV/ADAS Safety-Critical Testing Scenario Generation from Vehicle Crash Data

    Chen, Bo-Shian / Kibalama, Dennis / Tulpule, Punit | SAE Technical Papers | 2022


    AV/ADAS Safety-Critical Testing Scenario Generation from Vehicle Crash Data

    Kibalama, Dennis / Tulpule, Punit / Chen, Bo-Shian | British Library Conference Proceedings | 2022


    AV/ADAS Safety-Critical Testing Scenario Generation from Vehicle Crash Data

    Kibalama, Dennis / Tulpule, Punit / Chen, Bo-Shian | British Library Conference Proceedings | 2022