System Design And Key Technology Research Of Chess-Playing Robot

With the tremendous advances in robotics and chess algorithms,there are already chess robots on the market that combine the two.But for now,most chess-playing robots are only on display in science museums for a number of reasons.The main reasons can be divided into three aspects.First,most chess-playing robots are industrial robots,which are large in size,high in cost and relatively complex in control.Second,in the way of perceives the position of board pieces,special circuit board is mostly used for positioning,which makes the identification and positioning ability of other different types of board and pieces poor,and the utility of special board pieces poor.Third,most robots are designed for a single game and cannot be extended to other board games.All these make it difficult for chessplaying robots to be promoted and developed among ordinary players.Therefore,this paper studies the above problems,mainly for chess robot chess chess recognition and positioning,game algorithm,mechanical movement three key technology research and design,and finally design and development of chess robot system platform.The main contents of this paper are as follows:Research and design chess robot piece positioning recognition.Chessman positioning recognition replaces the use of special chessboard chessmen by visual perception;It can be divided into three parts: checkerboard image preprocessing,checkerboard positioning and chess positioning recognition.Obtain complete checkerboard and chess pieces images with hd camera,and verify the images;Process the image and locate the checkerboard;Adopt efficient and accurate template matching method to identify and locate the pieces of backgammon.In order to cope with the diversity and rotation of chess pieces,the convolutional neural network model with the SE module of deform convolution and attention mechanism was used to classify and identify Chinese chess pieces,and the recognition accuracy reached 99.8%.Research and design chess robot game algorithm.Alpha Zero,a universal chess game algorithm based on reinforcement learning,was selected as the core algorithm.This paper designs an algorithm for the game of backgammon with no forbidden hands.By simplifying the network structure and using lightweight Shuffle Net network element,the volume of the algorithm model is reduced and the running speed is increased.For the Chinese chess game algorithm design,using dynamic adjustment learning rate to speed up training,increasing supervised learning to improve game ability,and using multi-process to improve processing speed.Study and design the mechanical movement of chess robot.The robot arm with small volume and low cost is used as the actuator.In order to guide the robot arm to the specified checkerboard position through two-dimensional images,vision and mechanical movement are combined.Hand-eye calibration tests were carried out by four algorithms,and comparative analysis was performed.Kinematic analysis and simulation experiments are carried out for the manipulator.Finally,the mechanical arm chess experiment is carried out to verify its feasibility.Design and develop chess robot system platform.The software part of the system is designed and developed by integrating chess position recognition,game algorithm and mechanical movement part.Choose low cost,good expansion of hardware to build chess robot hardware platform;Through comprehensive experiments of software and hardware,a chess-playing robot system platform with high efficiency,accuracy,richness and friendliness is developed.Finally,the feasibility of the system is verified by experiments.