In context
In 2023, industrial robots increasingly handled contact tasks such as grinding, wiping, and material placement, but conventional time-dependent motion planning methods were vulnerable to disturbances. When operators needed to inspect or repair tasks online, robots had to stop and be reprogrammed, reducing efficiency. This paper from Wuhan University addressed the need for contact task planning with strong anti-disturbance capability.
What was reported
Researchers proposed a motion planning and control method based on a diffeomorphic mapping dynamical system (DS). Using kinesthetic teaching, they collected position and force data from human operators performing carbon fiber fabric placement. A time-invariant DS was built via diffeomorphic mapping, avoiding explicit time dependence. The contact task was decomposed into three states: executing the contact task, moving under human disturbance, and moving in free space. Force and orientation adjustment terms were designed for each state.
Experiments compared the method with offline control. Without disturbance, repeated trials showed maximum position and orientation errors below 0.008 m and 1.01° relative to the demonstration trajectory, with an average force RMSE of 0.86 N—slightly higher than offline control. The robot could start from any free-space position and enter contact space to perform tasks. When human disturbance occurred, the robot followed the operator's intent and resumed the task after disturbance, with force RMSE similar to undisturbed operation.
Why it mattered
This work demonstrated a practical approach for contact tasks requiring both high precision and robustness to human interaction. By making motion planning time-invariant, robots could adapt to disturbances and recover autonomously, reducing downtime and enabling more flexible human-robot collaboration in manufacturing settings.
When a disturbance occurs during the operation process, the robot will follow the instructions of the operator and resume the tasks after the disturbance with a force RMSE similar to the result without human disturbances.
Source: 《机器人》期刊 (robot.sia.cn) · Published 2023-07-08 · “基于微分同胚映射动态系统的接触任务运动规划与控制”