Quantitative assessment at task-level for performance of robotic configurations and task plans

Abstract

Given a robotic manipulation task, decision on which robotic configuration (robotic system and necessary peripherals, including assistant tools, sensor systems, and so on.) to use and evaluating performance of the solution remains as an open, challenging but a meaningful problem. This work attempts to address this problem by defining the concept of task-level performance and developing an approach for systematic assessment based on a proposed task representation for performance quantification of different task plans and robotic configurations. Starting from productivity which is one of the main concern for manufacturers, this work introduces a methodology for quantitative assessment of any given robotic configuration and task handling. Such a method is useful for comparing the productivity of different robotic configurations and evaluating the worthiness of updated solution of a sub-task through observing improvements in productivity. Implemented and tested on a peg-in-hole task with the different level of difficulty through single-arm and dual-arm manipulation on a dual-arm robot, the methodology has demonstrated encouraging results. Finally, the connection of performance assessment from task-level to economy-level is presented.