A Novel Robust Shared Teleoperation Framework Based on Robot Learning and Dynamic Virtual Barrier

The shared teleoperation integrates the operational expertise of human tutors with the planning precision of autonomous agents, thereby assisting operators in accomplishing complex tasks with greater ease. However, teleoperating a robot arm with traditional guidance or forbidden virtual fixtures (VFs) presents inherent challenges due to the task’s high dimensionality, complexity of human action intention and the dynamic obstacles due to limited visibility or attentional constraints. In addition, relying on outdated tutor’s experience may not adequately adapt to novel or changing environments. To address these challenges, a novel robot learning based shared teleoperation framework is proposed, which utilizes guidance flow (GF) and dynamic virtual barrier (DVB) to provide a more user-friendly haptic interaction interface. The agent acquires operational skills through multiple demonstrations or updating data, enabling the generation of shared assistive tools. Inspired by the principles of fluid mechanics and time series prediction, the designed methods can reduce the workload and stress experienced by novice operators. Comparative experiments have demonstrated that the proposed method can simultaneously improve the accuracy of operators’ task performance and ensure the safety of the task execution process.