Research Of Multi-robot Co-location On Autonomous Robot Soccer System

Today, robot technology have been developed rapidly and been used inmultiple fields such as military, industrial manufacturing, medical andaerospace. Although many researchers have been in-depth studied individualrobot which makes it more functional and smarter, it’s difficult to completethe complex task because of its limitations. Therefore, the multi-robotcollaboration technology started to develop. This technology taking theadvantage of the multi-robot groups to complete the task which one singlerobot can’t do.As an intelligent robot system and the platform of multi-robot system,the research of autonomous robot soccer system rapidly improved the level ofmulti-robot collaboration technology. The highest level of autonomous robotsoccer game is ROBOCUP medium-sized robot soccer game. Firstly, thepaper introduced the hardware, software and subsystems of medium-sizedsoccer robot system. Then, to lay the foundation for multi-robot observationpositioning, paper gives the detailed description of how a single robotoperating the self-positioning and target-positioning. By introducing thedensity-based spatial clustering of application with noise algorithm(DBSCAN), paper had analysis and studied the multi-robot informationfusion system, finding out the problems of accuracy and stability of traditionalDBSCAN fusion algorithm on data processing. Therefore, the researchproposed the improved DBSCAN algorithm which the robot observationdistance threshold was introduced. This improved algorithm, combining theactual condition of less observation distance the higher accuracy ofpositioning, uses observation distance threshold to re-screening the datawhich might be deleted by traditional DBSCAN algorithm. Comparing theobservation distance with the threshold, the algorithm chose the real noise data to improve the volume and accuracy of integration data, so that we cansolve the problem of observation information error and instability of dataintegration. In the experiment, we compared the results of data integrationusing both traditional DBSACN and improved one, in the situation ofobserves the same one target by multi-robot. The experiment result shows thatcomparing the traditional algorithm the improved one gets more stable andaccurate integration data in the condition of changing EPS. At the end of thepaper, the information sharing between robots in the experiment was told, andthe problems and some points need to be improved in the study was proposed.