Closed-Chain Manipulation of Large Objects by Multi-Arm Robotic Systems
This work solves a specific problem in robotics for multi-arm systems, offering incremental improvements in closed-chain manipulation tasks.
The paper tackles the problem of manipulating large objects with multiple robot arms by addressing the connectivity issues in the configuration space due to closed-chain constraints and joint limits. It introduces an 'IK-switch' regrasping move and other methods, enabling tasks like flipping a chair frame that were previously impossible with existing multi-arm planning approaches.
Closed kinematic chains are created whenever multiple robot arms concurrently manipulate a single object. The closed-chain constraint, when coupled with robot joint limits, dramatically changes the connectivity of the configuration space. We propose a regrasping move, termed "IK-switch", which allows efficiently bridging components of the configuration space that are otherwise mutually disconnected. This move, combined with several other developments, such as a method to stabilize the manipulated object using the environment, a new tree structure, and a compliant control scheme, enables us to address complex closed-chain manipulation tasks, such as flipping a chair frame, which is otherwise impossible to realize using existing multi-arm planning methods.