This study aims to propose a virtual fixture (VF) – assisted robot teleoperation framework that modulates the generated trajectory from demonstrations to response varying obstacles in complex environment.
First, a single trajectory is learned from demonstration using dynamic movement primitives (DMP). Then, the classic DMP is improved by integrating adaptive terms and updating the radial basis function kernel weights, allowing the single-skill trajectory to alter to respond to obstacles dynamically. Finally, composite virtual forces are generated on the haptic device to enhance operational stability and prevents unintentional operations that could bring the robot into close proximity with obstacles.
The VF method can accomplish online obstacle avoidance and operation guidance. The experimental results show that the operation trajectory based on VFs is smoother compared to the operation without assistance. Moreover, the operational speed has increased by up to 44.37% compared to the demonstrations.
The proposed composite VF-based protection framework solves the problem that classical DMP method cannot dynamically avoid obstacles, and effectively improves the operational safety and efficiency.
