A Novel Shared Control Framework Based on Imitation Learning

Learning from Demonstration (LfD) stands out as a powerful tool for swiftly deploying tasks in robotics, distinguished by its capacity to effectively leverage the task expertise of experts. However, contemporary robotic intelligence has not yet attained a level of complete autonomy. The integration of shared control methodologies allows for the assimilation of offline experiences derived from experts, concomitant with the introduction of online professional guidance during task execution. This study augments the established framework by incorporating electromyographic signals to discern expert intentions and utilising the Gaussian Mixture Model (GMM) to learn experience. The proposed approach engenders a more efficient shared control method. Validation of this methodology is conducted on the Omni and Franka Emika Panda robotic platforms.