Virtual Validation Method of Automated on-Sight Driving Systems for Shunting Operations

EasyChair Preprint 9828

Abstract

Virtual scenario-based testing is considered best practice in the development and validation of highly automated driving functions. A scenario describes a holistic and coherent test setup in a use-case oriented environment in order to make a claim about the validity of derived test results. Railway kinematics are characterized by the system immanent track guiding allowing narrow clearance to infrastructure. High masses and low coefficients of adhesion result in long stopping distances. In addition, shunting is performed on sight whereas train line operation is performed on signaling. Ensuring a sufficient safety distance to obstacles, high sensor performances in range and resolution is required for real-time sensor-based object detection. A stable object classification and subsequent obstacle warning is mandatory for trouble-free automatic train operation (ATO). To validate the perception-based decision making, the following target braking or acceleration, the driver model and actuation systems, a huge number of situations has to be considered. Regarding time, cost and feasibility, real world operation cannot provide a complete set of necessary tests, thus, virtual scenario-based testing, also used for validation of advanced driver assistance systems in automotive engineering, is adapted. Therefore, a photorealistic railway environment designed along physical properties is developed. According to the layout and design of the system under test (SUT), the physical sensor positions and characteristics are simulated. Raw sensor data is emulated and looped out to the SUT respectively, where the driving decision is made. The decisions are wrapped and looped back to the vehicle model, an integrated module as a part of the simulation environment. Scheduled scenario set-ups including static and dynamic elements as well as expected system behaviors are defined as concrete test cases.

Keyphrases: ATO, Digital Engineering, Validation, simulated sensor, virtual testing

@booklet{EasyChair:9828,
  author    = {Steffen Schäfer and Lucas Greiner-Fuchs and Tobias Hofmeier and Philipp Koch and Martin Cichon},
  title     = {Virtual Validation Method of Automated on-Sight Driving Systems for Shunting Operations},
  howpublished = {EasyChair Preprint 9828},
  year      = {EasyChair, 2023}}