Improving Strategy in Robot Soccer Game by Sequence Extraction
Václav Svatoň, Jan Martinovič, Kateřina Slaninová, Tomáš Bureš · Procedia Computer Science · 2014
Robot Soccer is a very attractive platform in terms of research. It contains a number of challenges in the areas of robot control, artificial intelligence and image analysis. This article presents a look at the overall architecture of the game and describes some results of our experiments in analysis and optimization of strategies using sequence extraction. We have extracted sequences of game situations from the log of a game played in our simulator, as they occurred during the game. Afterwards, these sequences were compared by methods LCS, LCSS and T-WLCS, which are usually used for sequence comparison in the sequence alignment area. Using these methods, we are able to visualize the relations between the sequences of game situations and clusters of similar game situations in a graph. In conclusion, a possible description improvement of these game situations is introduced. Therefore, a possible strategy improvement to ensure a smoother and faster performing of actions defined by these situations is described.