A Compact and Low-cost Gobang Robot

This paper proposed a compact and low-cost Gobang robot, which can realize combat between a robot and a real person. Three modules, which includes the vision module, the chess playing strategy module, and the human-robot module, have been accomplished and integrated. Several key problems and challenges in the procedure of human-robot interaction have been solved: 1) robust and accurate detection of the chessboard corners; 2) efficient mapping between the logical and physical boards; 3) robust recognition, classification and grasping of chess pieces; 4) determination of the best drop position of the robot. The workflow of playing chess is as follows. At first, the visual module captures the current situation on the chessboard, and returns the results to the background Gobang program in real time. Then the program judges the current playing status. If it is the robot's turn to play, the robot control module works. The best position for the next move was returned. The robot goes grab a chess piece and then drops down it in the specified position. Hundreds of actual human-robot combats validate the performance of the proposed project. Due to its advantages of low cost and harmlessness to human eyes compared with computer games, the proposed Gobang robot has great potential to be a domestic robot.