An Optimization-Based Magnetic Field Potential Learning Framework for Robotic Volumetric Obstacle-Avoidance

Safe and high-efficiency motion planning technology for robots is especially crucial in industrial and daily life. This article aims to enhance the obstacle avoidance planning capabilities of robots within their operating space. To improve the efficiency of obstacle handling in the process of robot motion planning, a new learning framework has been presented for generating obstacle avoidance trajectories using dynamic movement primitives (DMPs). The proposed method combines demonstration learning with a magnetic field potential to achieve volumetric obstacle avoidance of superquadratic potential, thereby eliminating the issue of local minima in the planning process. In addition, the convergence of the target state is demonstrated through an analysis based on Lyapunov stability theory and the effectiveness of the presented method has been verified and analyzed through a series of numerical simulations. The results demonstrate that the proposed approach offers a reliable solution for motion planning, enabling robots to generate smooth, fast, and collision-free trajectories using artificial magnetic fields. Furthermore, the effectiveness of the proposed method is validated through experiments conducted on Kinova Gen2 robot.