Design Of The Multilevel Cooperation Vision System For Industry Multi-Robot Systems

Since the single robot technology can not meet with the demanding of the increasingly complicated tasks, a new technology based on multi-robot systems comes into focus. In order to make full use of the multi-robot's incomparable advantage of team-work which is irrealizable in a single robot system, the key problem lies on how to make every robot member get enough outside and interactive information. Nowadays, machine vision has been largely used in the information sensor because of its richness in information. But most of the vision systems recently are the ones based on single robot or self-governed single level vision system. So to multi-robot cooperation systems, there is no any excellent solution scheme.This paper roots in the fifteen 211 national key laboratory building project--the platform of industry multi-robot cooperation system building. And the platform mainly consists of two swing mobile robots and one ground mobile robot with the accessory assembling and machining in industry as the simulation object. This paper is to design and set up one solution scheme suitable for multi-robot cooperation systems.One solution scheme of multilevel cooperation vision system simulating the human's visual process of searching the objects is proposed in this paper, which can resolve the conflict between the FOV (field of view) and the precision normally seen in robot vision. This vision system is mainly comprised of:1) The tiptop level is an active global trinocular system for stereo positioning. Vision of this level consists of three active vision cells, using active to extend the FOV and flexibility of the global vision. One method of adjusting the dynamic homography of the vision cell is proposed to realize the global trinocular positioning without blind areas. The contract experiments show that the proposed method has better performance whether in positioning precision or in application rang.2) The middle level is a binocular hand-eye local stereo vision system. This level consists of three binocular stereo cameras fixed on the terminal of the robots. The restriction relationship between binocular is obtained by the two homographies and the positioning method of object relative to robot is gained by solving the relation between hand and eye. The binocular hand-eye close quarters stereo vision system has been experimentally proved to be of high precision.3) The bottom level is a monocular orientation vision system. This level uses pinhole camera fixed in the center of the machine paw, and obtain the object orientation information related to the paw through one image. The method has been proved by the experiment of its excellent orientation guidance.The preliminary frame of the cooperation in visual searching and positioning task is also proposed in this paper. This cooperation is among multilevel vision systems from tiptop level large-range search to bottom level local precise visual feedback, which will commit to the coming multi-robots cooperation research.