We investigated how extended reality (XR), haptic feedback, and task-parameterized Gaussian mixture regression (TP-GMR) can be fused into a single teleoperation framework for electric vehicle (EV) battery disassembly. The work demonstrates how variable autonomy lets operators fluidly hand tasks to the robot while constraint barrier functions guarantee safe tool motion inside tightly constrained battery modules.
Overview
- XR visualization keeps operators co-present with the remote workcell, while haptic feedback communicates contact forces in real time.
- TP-GMR learns manipulation skills from demonstrations and generates corrective trajectories whenever the operator deviates from a safe path.
- Variable autonomy enables seamless switching between manual, shared, and autonomous modes without interrupting the job.
Experimental Highlights
- Implemented on an industrial KUKA manipulator handling real EV batteries.
- Constraint barrier functions preserved spatial safety envelopes even during aggressive teleoperation manoeuvres.
- Predictive replanning reduced operator response times from ~2.0 s to sub-millisecond updates, cutting task completion time by up to 48% and path deviation by 32% compared with unaided teleoperation.
Why It Matters
Designing for human-in-the-loop operation is crucial in high-risk, semi-structured environments such as battery recycling. This framework shows how learning-from-demonstration and XR interfaces can elevate operator performance while keeping safety constraints under tight control.
Media
- Paper PDF
- Journal page
- EV battery disassembly project page (context on the lab behind the experiments)