Simulation-Based Testing Framework for Autonomous Driving Development

This paper presents a simulation-based testing and validation framework for ADAS (Advanced Driver Assistant System) and AD (Autonomous Driving) technology development. One of the main challenges in ADAS development is validating the perception, planning, and control algorithms in a closed-loop fashion, where both vehicle dynamics configuration and a wide variety of traffic scenarios are taken into account. This requires extensive testing efforts during the development cycle. Moreover, the designed algorithms should obtain optimal performance toward safety, comfort, and time/fuel optimality. Our contribution is twofold. First, we demonstrate a cosimulation platform for high fidelity vehicle dynamics, sensors and traffic environment modelling, using Siemens Simcenter Amesim and Simcenter Prescan. An interface that facilitates the testing processes and frontloads design verification during the early phases is studied. Second, we present some of our algorithm developments, i.e., deep learning, scene recognition, planning, and control, dealing with different driving scenario types and requirements. The designs are mainly based on the proposed co-simulation framework. Finally, the approach is demonstrated with four different use cases: adaptive cruise control, green wave technology, autonomous valet parking, and double lane change.