Visual Servoing for Robot Manipulators Considering Sensing and Dynamics Limitations

Abstract

This paper presents a control scheme of visual servoing. Real-time vision guidance is necessary in many desirable applications of industrial manipulators. Challenge comes from the limitations of visual sensing and robot dynamics. Typical industrial machine vision systems have low sampling rate and large latency. In addition, due to the large inertia of industrial manipulators, a proper consideration of robot dynamics is important. In particular, actuator saturation may cause undesirable response. In this paper, an adaptive tracking filter is used for sensing compensation. Based on the compensated vision feedback, a two-layer controller is formulated using multi-surface sliding control. System kinematics and dynamics are decoupled and handled by the two layers of the controller respectively. Further, a constrained optimal control approach is adopted to avoid actuator saturation. Validation is conducted using a SCARA robot.