A digital twin-based platform for testing and optimization of path planning algorithms

Abstract

With the rapid development of autonomous driving technology, there is an increasing demand for safety, reliability, and optimization efficiency in path planning algorithms. However, traditional physical testing is often costly, time-consuming, and subject to environmental uncertainties, making it difficult to efficiently verify and optimize these algorithms. To address this issue, this paper proposes a high-fidelity digital twin-based platform for testing and optimizing path planning algorithms. By constructing a simulation environment that mirrors the physical world, the platform minimizes the gap between simulation and real-world scenarios, enhancing the safety and stability of path planning. The platform integrates global path planning using the A* algorithm, local path planning with the Timed Elastic Band method, and optimization using Bézier curves to improve the smoothness, feasibility, and safety of the path. Additionally, it incorporates the vehicle’s physical characteristics — such as velocity, steering angle, and drive mode — into the parameter optimization process, ensuring consistency between the simulation and the real-world environment. Experiments were conducted by deploying identical path planning algorithms in both the simulation and the physical environments. The results demonstrate that algorithms optimized through the digital twin platform can be reliably transferred to real-world scenarios, improving obstacle avoidance and overall path planning safety. The planned paths in the physical environment closely matched those in simulation, confirming the effectiveness of the digital twin approach for path planning testing and optimization. This research provides new insights into environmental adaptability, safety assurance, and engineering deployment of path planning in autonomous driving.