자율주행 검증을 위한 혼합현실 시뮬레이션 시험

초록

Autonomous driving requires safe and efficient navigation in complex environments. This has led to an increased demand for reliable test methods, where driving simulators are widely used to simulate sensors in autonomous vehicles. In addition, the concept of digital twins, which combines real and simulated environments, is being applied to autonomous vehicles for efficient and realistic testing. However, inaccuracies in vehicle modeling in simulators can lead to cumulative position errors, especially during sharp maneuvers, undermining the reliability of test results. This paper presents a method to correct cumulative errors in a driving simulator environment by synchronizing the vehicle position and orientation using navigation data. The system periodically adjusts the vehicle dynamics in the simulator to reflect real-world dynamics, eliminating cumulative position errors and narrowing the gap between the simulation and the real-world environments. This approach is particularly effective for interaction-based tests such as Autonomous Emergency Braking (AEB), where position accuracy is crucial. To demonstrate the feasibility of this mixed-reality system architecture, experiments were conducted in accordance with the Euro-NCAP AEB protocol. The results illustrate the benefits of the proposed method in mitigating simulator errors and enhancing the overall reliability of autonomous driving testing.

키워드