Research on Dynamic Path Planning and Real-Time Collision Detection for Bench Dragon Based on Multi-Algorithm Collaborative Optimization

Abstract

The dynamic and complex nature of traditional dragon dance performances, especially the multi-segmented "Bench Dragon" presents considerable challenges in real-time collision detection and path optimization. This study introduces an integrated optimization framework that combines an enhanced golden section search algorithm, dynamic collision detection based on the Separation Axis Theorem (SAT), and a genetic algorithm to tackle these challenges. The enhanced golden section algorithm effectively reduces computational complexity by efficiently narrowing the search space, while SAT ensures collision-free motion through real-time projection analysis. Additionally, genetic algorithms optimize the turnaround path by minimizing the total length under geometric and continuity constraints. Experimental results demonstrate the framework's effectiveness: an optimal pitch of 0.4529 m facilitates smooth coiling, with the total turnaround path length reduced to 13.6206 m without any collisions. The algorithm maintains high accuracy while ensuring computational efficiency, as validated through simulations of the dragon's movement across critical time intervals (0–300 s). This research offers a robust solution for coordinating large-scale rigid-chain movements, with potential applications in swarm robotics and automated logistics. However, limitations include idealized assumptions of rigidity and 2D motion, underscoring the need for future research into real-time dynamic modeling and multi-physics constraints.