Dynamic Obstacle Avoidance Planning for Robots in Unknown Environments Based on Trajectory Prediction

Abstract

In dynamic environments, robots are required to generate collision-free trajectories and effectively avoid dynamic obstacles. Previous dynamic obstacle avoidance planning algorithms suffer from issues such as long online solving times and poor realtime performance. We propose an optimized algorithm for dynamic obstacle avoidance trajectory generation based on trajectory prediction. The algorithm identifies dynamic obstacles in inflated point cloud information through a clustering algorithm and predicts their future trajectories. Then, a dynamic obstacle avoidance constraint is constructed, and the trajectory generation optimization problem is solved to obtain the planning trajectory for obstacle avoidance. Simulation and real-world experiments demonstrate that the algorithm can generate effective dynamic obstacle avoidance trajectories, achieving real-time, safe, and efficient navigation in dynamic environments. Compared with traditional methods, the method proposed has a higher planning success rate and a shorter online solving time.