High-Fidelity Joint Self-Driving Simulation Platform for Unmanned All-Terrain Vehicles

Self-driving simulators play a key role in the development and deployment of unmanned vehicle algorithms. While the ad-hoc usage of simulation in self-driving and robotics research is commonplace, none of the current simulators can fulfil the high-accuracy dynamics and vision simulation demands of all-terrain specialized vehicles. This paper addresses the limitations of current simulation platforms; to simulate our unmanned active articulated all-terrain specialized vehicles, called A3TV, we introduce a joint simulation platform, integrating Gazebo for high-fidelity dynamic simulation and Unreal Engine 5 (UE5) for high-fidelity visual simulation. Gazebo simulates links-and-joints architecture, including a novel approach to modeling locomotion mechanisms, while UE5's advanced graphics engine handles dynamic real-time lighting. The simulator supports flexible sensor configurations, providing rich signal data for training machine learning models. Encompassing diverse off-road scenarios and weather conditions, it significantly contributes to mechanical design and navigation algorithm testing, accelerating A3TV development and enhancing overall efficiency.