ABSTRACT
Autonomous vehicle technology has been advancing rapidly. Comprehensive testing is essential before road deployment. However, no test can cover every possible situation. Thus, a good simulation system is necessary. Most existing systems are restricted to limited aspects that fail to correspond with reality in complex and volatile environments. Cyber-Physical Vehicle System (CPVS) technology enables the simulation systems to integrate more closely with actual vehicles. Nevertheless, the behavior analysis and decision-making capabilities are still far from perfect. The proposed framework is based on CPVS that integrates physical sensors and mechanical components on a vehicle with a ROS2-based cyber-system receiving messages through WAN for cooperative optimization. The messages conform to IEEE and SAE standards integrated with RealSense, Lidar, IMU sensor data and mechanical components information to display the vehicle status with Unity graphics in two modes: real-time scene rendering and vehicle monitoring mode for vehicle monitoring, and scene restoring mode for error analysis. We emphasize the cyber-physical interaction in which the physical system serves feedback data for model building and scene construction while the cyber system tests the controller with various simulated scenes. A scene editor subsystem is provided to visually layout static/moving objects on a virtual plane to construct any desired scene and trigger Unity to generate simulated sensor data for evaluating the reactions of simulated or physical vehicles. The CPVS is used in the design of an autonomous ground vehicle (AGV) simulated and tested using serval situations layout with the scene editor. Compared with testing the AGV in real environments, the result is highly satisfactory.
