A MAB-Based Discrete Power Control Approach in Anti-jamming Relay Communication via Three-layer Stackelberg Game

In this paper, we investigate the discrete power control problem in anti-jamming relay communication networks. Based on the hierarchical competitive relationships between transmitters (user and relay) and jammer, a three-layer Stackelberg game is formulated, in which user acts as leader, relay acts as vice-leader and jammer acts as follower. From the perspective of hierarchical-game theoretic, we formulate the power optimization problem as a multi-armed bandit (MAB) problem, where user, relay and jammer act as players and each optional power strategy is considered as an arm to select. Based on MAB theory, we give the regret function to express the loss of payoff of the whole communication process. To minimize the regrets of user and relay, we propose a UCB1-based discrete power control online learning algorithm. Simulation results give the power selection rate and logarithmic incremental regrets in the proposed anti-jamming scenario. The user's and relay's utilities are also compared under different algorithms.