Optimal Tracking Design For Mobile Robot Based On Image Sensing Information

With the age of intellectualization coming,the real-time network control technology based on computer vision has got wide attention from many researchers,which has been widely applied on intelligent monitor,unmanned vehicle,intelligent transportation system,interactive processing,robot sensing and so on.Due to the complexity and multiplicity of the applied scenes,different computer vision algorithms are to be chosen properly and boosted.Besides,data transmission by network is likely to undermine the realtime performance in some cases,which exerts a bad effect on the control performance.Base on the multi-robot control system built up by our own lab,this article,which aims at the data transmission delay and packet loss problem,design a system model with multiplication noises in feedback loop and an H2 optimal tracking controller.In addition,base on the research of the computer vision algorithm,comparative study has been made and an proper algorithm for this system is adopted,with which a realtime multiply robot detection system is developed.The main work is as follows:1)The overall introduction of the multiple robot system,mainly on the hardware informations of the basic components and software devolpe tools,the task of these components and the signal flow around them,as well as the control block diagram of this system.2)The optimal tracking controller design base on network feedback.On the data packet loss problem,we study the H2 optimal state feedback controller design with multiplication noises in feedback loop.In this way,the robots are able to track the trajectory with higher accuracy even when the community quality is poor.Moreover,this optimal controller can be a general reference controller for systems with the same control structure,which can be a fundation study for other networked control systems.3)The design of multiple moving object detection system base on computer vision.The article firstly studies some kinds of moving object detection algorithms and compares their features,runtime and experiment outcomes,and then develops a detection system base on Codebook algorithm.The system can calculate the world coordinate of the moving objects,which makes up the position detection block of the moving robot close-loop system.Besides,a probability data association algorithm is raised to solve the problem that the object image might crack if the background substation algorithm is used.With the constraint that the object is on the ground and its centroid altitude can be ignored,a monocular camera calibration algorithm is raised,which can solve the communication delay and syschronization problem of binocular camera calibration.