Research On Intention Recognition And Decision Design In Confrontation Environment

In the confrontation scenario,the robot's self-understanding of the enemy's intention and decision making based on the understood intention is the core technology to realize the robot's autonomous control.The ICRA RoboMaster Artificial Intelligence Challenge is the first attempt to combine artificial intelligence with robots and is of great significance for future robot intelligence.Based on the ICRA RoboMaster Artificial Intelligence Challenge,this paper studies the intentional understanding of enemy robots in the confrontation scene and the design of our robot decision algorithm.The main work of the thesis is as follows:First,the definition of the coordinate system and the conversion relationship between the coordinate systems are given.The kinematics model of the wheeled robot is established,and the mathematical description of the intent understanding and decision problem is given.Secondly,based on the data-driven method,the framework of intent understanding problem is constructed,and the data collected in the simulation environment and experimental environment is used to construct the hierarchical data set.Thirdly,the enemy intention abstraction based on heat map is proposed.Two kinds of intent understanding algorithms based on long-term and short-term memory and sequence similarity analysis are designed.The simulation results show the effectiveness of the intent understanding algorithm.Then,hierarchical decision machines based on intentional understanding,and reinforcement learning are proposed.The simulation results show the effectiveness and stability of the algorithm.Finally,the RoboMaster AI experimental platform is built.The design experiments verified the validity of the proposed intent understanding algorithm and decision algorithm,and analyzed the experimental results.