Development of an Autonomous, Explainable, Robust Robotic System for Electric Vehicle Battery Disassembly

Abstract

The vigorous growth of the electric vehicle industry calls for efficient disassembly of used electric vehicle batteries (EVBs). Screw disassembly by robots remains a challenge due to the uncertainties in this task. In this paper, we designed an architecture of NeuroSymbolic task and motion planning, which uses neural predicates to map the sensor into a quasi-symbolic state and schedules action primitives autonomously based on current state and goal state. This architecture guarantees autonomy and explainability which is important in human-robot hybrid disassembly pipeline. In primitive implementation, a customized end-effector, accurate vision-based and force-based pose estimation are enabled to ensure the robustness of the system. The experiment shows that the proposed system can achieve 100% success rate in lab environment. We will deploy and evaluate it in the real factory environment in the future.