A Distributed Active-Vision Network-Space Approach for the Navigation of a Car-Like Wheeled Robot

Abstract

The navigation for a car-like wheeled robot (CLWR) in a distributed active-vision network-space system (DAVNSS) subject to three fuzzy variable-structure decentralized controls (FVSDCs) is developed. The scheme also includes trajectory tracking and obstacle avoidance. Two distributed wireless charge-coupled-device (CCD) cameras individually driven by two stepping motors, viz., active CCD1 and active CCD2 (or simply ACCD1 and ACCD2), are constructed to capture the dynamic pose of the CLWR and the obstacle. The proposed control system includes quad processors with multiple sampling rates. First, a personal computer (PC) is employed to receive the image of the CLWR or obstacle from ACCD1 or ACCD2 by a wireless transmitter and then to plan three reference commands for CLWR, ACCD1, and ACCD2. Next, a six-step image-processing routine and the calibration between the world coordinate and the image plane coordinate using multilayer perceptrons (MLPs) are established. Finally, experiments are performed to validate the proposed control system.