Distributed Spectrum Resource Allocation via Stochastic Learning in Potential Games

Aiming for closer-to-optimal solutions to the distributed opportunistic heterogeneous spectrum access (OHSA) problem, we study from the perspective of game theory and propose a memory and regret based learning algorithm (MRLA). Firstly, aiming for the optimization of the total throughput in a cognitive radio network, we build a weighted congestion game (WCG) by considering each cognitive user as a game player and introducing a differentiating coefficient. Afterward, we propose the MRLA in which each player makes choices from the action set based on its identity and memory. Thirdly, we prove that the MRLA converges to a Nash equilibrium solution in a distributed manner within a finite number of coordination steps. Finally, comparative simulations validate MRLA's advantages in terms of solution accuracy and convergence speed.