Human Error Recovery by a Maintainable Human/Robot Parts Conveyance System

In the paper, special attention is paid to human errors in a human/robot coexistence system for their collaborative tasks in which the robot partially assisted the power of a human operator. The possibility of applying maintainability concept for maintainance of tasks is explored for human error detection and recovery. After setting a problem of inferring a goal intended to reach by the operator, we propose to provide the robot with a capability of inferring an operator-intended path pattern from a halfway sequence of detected trajectory data using HMM whose algorithm is modified for our target. We verify, through experiments of the robot operation for rapid parts conveyance tasks, that efficient human error detection and recovery can be achieved with the aid of the proposed operator-intended path pattern inference scheme and demonstrate its usefulness. Interviews from the subjects in the experiments reveal that all human error types of slips, lapses, and mistakes are included in which our proposal is effective against human lapses and mistakes.