A Superquadrics-Based Steering Angle Obstacle Avoidance Method of DMPs

In this paper, we propose a new method for using Dynamic Movement Primitives (DMPs) to enable robots to avoid volumetric obstacles when learning skills from human demonstrations. We can utilize coupling terms in DMPs to modify the planned trajectory of the end-effector. Our method combines superquadrics and a steering angle scheme to construct new coupling terms for obstacle avoidance, without relying on potential fields. This guarantees the convergence of the dynamic system. Additionally, we modify the original center point steering angle formula to improve the ability to avoid obstacles in a wider range of angles. Our proposed method is applicable to most types of superquadrics, including those with flat boundaries, and we validate it through simulations.