Attack and defense cooperation mechanism of robot football team based on eligibility trace Q-learning algorithm

Abstract As science and technology progress rapidly, the application of AI in robotics technology has received widespread attention. Aiming at the problem of low efficiency in the current methods used for attacking and defending cooperation in robot football teams, a eligibility trace Q-learning algorithm combining eligibility trace and reinforcement learning Q-learning algorithm is proposed, and a robot football team attacking and defending cooperation mechanism based on eligibility trace Q-learning is designed. Comparing the performance of the eligibility trace Q-learning algorithm with Q-learning, SARSA and HTD algorithms, it is found that the convergence period and average computation time of the proposed algorithm are 3367 and 57.6ms, respectively, which are superior to the compared algorithms. Subsequently, the proposed robot soccer team’s attack and defense collaboration mechanism is analyzed in the 2D simulation system of the Robot Soccer World Cup. The results show that under the proposed mechanism, the passing success rate, interception success rate, and shooting accuracy of the robot soccer team are 98.08%, 97.28%, and 96.37%, respectively, and the average reward metric is 1.82, which is better than the comparison mechanism. The data presented above suggest that the introduced algorithm and robot soccer team’s attack and defense cooperation mechanism have good performance and application effects, which can effectively improve the efficiency of robot soccer team’s attack and defense cooperation. It can provide theoretical basis for the research field related to robots.