MuJoCo-Simulation for Telemanipulation through Haptic Force Feedback Devices

Kurzfassung

In haptic teleoperation, an operator controls a robot from a distance while receiving force feedback through a haptic device. However, the controller often behaves the same way in free space and near obstacles. This means that the operator must mainly rely on visual feedback to avoid unintended contact, which can make precise manipulation difficult. This thesis investigates whether the teleoperation system can support the operator by using proximity information before contact occurs. A bilateral teleoperation setup was developed using a Novint Falcon haptic device as the leader and a simulated KUKA LWR3 robot arm in MuJoCo as the follower. The controller was implemented in MATLAB/Simulink, and distance information from the simulation was used to detect nearby objects. Two proximity-based strategies were compared with a baseline controller. The first strategy, velocity scaling, reduced the commanded robot motion when the end-effector moved close to objects. The second strategy, stiffness modulation, increased the stiffness of the haptic interaction near objects, so that the operator felt stronger resistance. A grasp-aware filter was also included to avoid treating a grasped object as an obstacle. The controllers were evaluated in a user study with twelve analysed participants. Each participant completed a pick-and-place task, an arc-following task, and a peg in-hole task with all three controllers. The results show that the effect of proximity-based adaptation depends on the task. Velocity scaling reduced unintended contacts in the sorting task compared with stiffness modulation, but it also caused more indexing and longer completion times in some tasks. Stiffness modulation did not significantly improve objective performance, but showed favourable workload trends. Overall, the thesis shows that proximity-based assistance can be useful, but its benefit depends on the task and the trade-off between precision, speed, and operator effort.