A hybrid visual servo control system for the autonomous mobile robot

Abstract

The eye-in-hand and eye-to-hand visual servo configuration both have its own merit and limitation. Eye-in-hand configuration provides more precision but less sight of scene covered. Eye-to-hand configuration is on the contrary. PBVS (Position Based Visual Servo) control suffers from its high sensitivity to model inconsistencies and task conditioning. IBVS (Image Based Visual Servo) control suffers from it lead to highly nonlinear control algorithms. This paper proposed a hybrid visual servo control system integrating both configurations (eye-in-hand and eye-to-hand) and both schemes (PBVS and IBVS) to overcome the above limitations. A wide-angle camera mounted on ceiling (eye-to-hand, calibration required) and PTZ (Pan-Tilt-Zoom, no calibration required) camera mounted on mobile robot (eye-in-hand) is adopted. Various camera calibration methods are evaluated. Sensing, perception, mapping, localization, path planning and motion control are implemented. The experiment result show an accumulated error in robot trajectory (80 mm – 99 mm). The visibility graph in path planning and Dijktra's shortest path searching algorithm offers the optimal performance. The proposed hybrid system shows the efficiency and effectiveness. The proposed hybrid system provides high field of view coverage and local detailed for the objects invisible to the global system.