SQGS: Sensing-based Quality-aware Robot Programming Guidance System for Non-experts

Abstract

Skill-based robot programming has been extensively investigated in robotic manufacturing systems. Sensor inputs are usually used to determine the correctness of skill execution. However, the effectiveness of sensor monitoring is affected by the limitation of sensor coverage (e.g., camera view), the detection algorithms' physical requirements, and the trajectory of robot motion. Without sufficient sensor coverage, the robot system may be late in capturing critical faults that drastically reduce performance. Furthermore, without proper metrics to quantify the capability of sensor monitoring, it is difficult for non-expert users to know how well the monitor system can capture fault events and their impact on actual task execution time. To address the above issues, we propose a sensing-based quality-aware robot programming guidance system to quantify the capability of sensor monitoring in terms of its execution time and camera coverage. We provide users a flexible way to specify quality specifications for user-defined sections-of-interests. Our system guides users to select proper skill parameters and add additional cameras to meet the quality requirements based on the sensing quality measures. We apply our system framework to a 6DOF robot arm for an object pick-up task.