Realistic pedestrian behaviour in the CARLA simulator using VR and mocap

• DOI: 10.48550/arxiv.2309.04418
• Publication Date: 2023-01-01
• Authors: Sergio Martín Ser...; David Fernández L...; Iván García Daza; Miguel A. Vázquez
• Keywords: Overtaking; Crowd Simulation; Probabilistic roadmap; Driving simulator

Simulations are gaining increasingly significance in the field of autonomous driving due to demand for rapid prototyping and extensive testing. Employing physics-based simulation brings several benefits at an affordable cost, while mitigating potential risks prototypes, drivers, vulnerable road users. However, there exit two primary limitations. Firstly, reality gap which refers disparity between prevents simulated systems from having same performance real world. Secondly, lack empirical understanding regarding behavior agents, such as backup drivers or passengers, well other users vehicles, pedestrians, cyclists. Agent is commonly implemented through deterministic randomized probabilistic pre-programmed models, generated real-world data; but it fails accurately represent behaviors adopted by agents interacting within a specific scenario. This paper extends description our proposed framework enable real-time interaction environments, means immersive virtual human motion capture CARLA simulator driving. We have designed set usability examples that allow analysis interactions pedestrians vehicles we provide first measure user's sensation presence environment.