基于嵌入式的机器人视觉伺服寻迹系统研究-watch their moves(good).pdf

Watch their Moves: Applying Probabilistic Multiple Object Tracking to Autonomous Robot Soccer Thorsten Schmitt, Michael Beetz, Robert Hanek, and Sebastian Buck Munich University of Technology, Department of Computer Science, D-80290 Muenchen fschmittt,beetzm,hanek,buckg@in.tum.de Abstract Schulz et al. 2001). Using probabilistic motion and sensing models these algorithms maintain probabilistic estimates of In many autonomous robot applications robots must be capa- ble of estimating the positions and motions of moving objects the objects’ positions and update these estimates with each in their environments. In this paper, we apply probabilistic new observation. Probabilistic tracking algorithms are at- multiple object tracking to estimating the positions of oppo- tractive because they are concise, elegant, well understood, nent players in autonomous robot soccer. We extend an exist- and remarkably robust. ing tracking algorithm to handle multiple mobile sensors with In this paper, we apply MHT, one of the probabilis- uncertain positions, discuss the specification of probabilistic tic tracking algorithms introduced above, to estimating the models needed by the algorithm, and describe the required positions of opponent players in autonomous robot soccer vision-interpretation algorithms. The multiple object track- based on image data. In robot soccer (middle-size league) ing has been successfully ap