Adaptive Decision-Making in Attack-Defense Games With Bayesian Inference of Rationality Level
Abstract
This article investigates two-player attack-defense (AD) games involving players with bounded rationality, where the defender aims to intercept the attacker, while the attacker aims to invade the protected area and avoid interception. We first set path planning optimization problems in a receding horizon fashion for each player and formulate the AD game. Then, using the level-k model of behavioral game theory, we specify the decision mechanisms for players with bounded rationality. We propose an adaptive path planning strategy, coupled with the Bayesian learning method, for the defender to counter the attacker with an unknown reasoning level of the decision mechanism. The Bayesian inference algorithm, which combines current observation information and historical receding horizon prediction trajectories to form the belief on the attacker's reasoning level, allows the defender to generate an adaptive interception trajectory with the multimodel strategy. Finally, both numerical simulations and experiments confirm the effectiveness of the proposed algorithm.
