Visual Positioning Technology And Coordinate Formation Control For Multi-agent System

Due to the complexity, diversity and universality of the task in the practical application environment, the multi-agent system has been widely used in many fields because of its distributivity in the space, concurrency in the mission, scalability of the system and strong fault tolerance, so the research for multi-agent system has important theory value and practical significance. However, there is a certain gap between the practical application and theory research of multi-agent system, especially the positioning and coordination control issues. Present research always focus on the theoretical simulation, the theoretical study can not be directly extended to practical application, in order to make theory research results of the multi-agent system used in practical application, this paper carried out research on the following several aspects:Firstly, an overall framework of multi-agent system is designed, and a multi-agent experiment system that consists of visual positioning module, communication module, and intelligent mobile robot SRV-1 is built. The system provides verification platform and experimental basis for the follow-up positioning and coordination of multi-agent formation control strategy research.Secondly, a multi-agent positioning algorithm based on Camshift and video stitching is designed, which improved a lot in the aspects of anti-interference and positioning range limitation. Tracking and positioning of agents based on Camshift algorithm can greatly reduce the errors caused by light interference, and using SIFT video stitching algorithm can expand the positioning scope, making up for limited scope of locating method based on visual positioning.Finally, a coordinated formation algorithm based on behavior and leader-follower is designed. Conbining the advantage of behavior and leader-follower formation control algorithm, the whole formation process is divided into four parts: formation generation, formation transform, obstacle avoidance and exception handle, and also these behavior subalgorithms based on the SRV-1 muti-agent platform are also designed. Through the behavior selection module, SRV-1 robot can choose the behavior which is active, finallythese behaviors have carried on the simulation and experimental validation.