Real-time Visual Servo Tracking And Muti-agent Coordination Control

Multi-agent system is a complex system which could achieve information sharing and cooperative work through a series of communication and cooperative control algorithm. It could solve a lot of problems which single agent is unable to cope with, owning incomparable advantage to single agent, so it is widely used in many fields nowadays. In multi-agent system, the most important problem we need to solve is multi-agent cooperative control. And as we know, the reason that a class of mechanical equipment was intelligent is that they hold sensors to sense external environment and the algorithm to judge and process.In this paper, I design an agent firstly according to the experimental requirements with self-localization unit and cooperative positioning unit based on driver layer, sensing layer, control layer and communication system. And then I study the real-time visual detection algorithms based on the servo camera of self-localization unit, including image segmentation algorithm, target detection algorithm and image localization algorithm, achieved target real-time localization in camera image. And then I study the control of visual servo system in self-localization unit, firstly the PID control algorithm without modeling is discussed in this section and then I explore the controller design of speed-loop and position-loop in double-closed-loop control system, identified the math model through the method of least square and Hankel matrix based on Markov coefficient, designed the speed-loop controller with the principle of output feedback and internal model, designed the position-loop controller with the principle of PID, tested the performance of controller with MATLAB/Simulink simulation, verified the effect of controller with the visual servo system. Finally I study the method of design of cooperative control for multi-agent system, discussed the Lyapunov method and decoupling control method for trajectory tracking and l- ? control method for Leader-follow, tested the effect of them with MATLAB Simulation, and then combined the method of Leader-follow with virtual structure, I proposed the method of cooperative control based on self-localization, and verified the effect with MATLAB Simulation and cooperative experiment.