Advanced VIL Architecture for Autonomous Driving: Legal, Technical, and Sensor Integration From ADAS to Full Autonomy

초록

Validation and verification (V&V) of autonomous vehicles in real-world testing are limited by time, cost, safety, and scalability. Driving simulators provide safer alternatives but often fail to reproduce actual vehicle dynamics. This paper presents a Vehicle-in-the-Loop (VIL) simulation framework that integrates a chassis dynamometer with a virtual driving simulator to emulate real vehicle responses in real time. The framework enables bidirectional communication between the autonomous driving controller and simulated sensor data. For ADAS-level sensors (e.g., radar, ultrasonic), Over-the-Air (OTA) stimulation converts virtual outputs into physical signals. However, OTA becomes impractical for high-bandwidth sensors such as LiDAR and multi-camera systems due to data complexity and transmission latency. To overcome these challenges, a data injection approach is adopted, directly feeding simulated sensor data into the controller through a custom packet interface. Scenario-based V&V experiments aligned with national traffic regulations demonstrate that the proposed framework achieves realistic, regulation-compliant testing and quantitative consistency with real-world driving, achieving reliable validation in complex scenarios.

키워드