Good Neighbor Alternative to Best Response and Machine Learning Based Beamforming and Power Adaptation for MIMO Ad Hoc Networks

Decentralized joint transmit power and beamforming selection for multiple antenna wireless ad hoc networks operating in a multi-user interference environment is considered. An important feature of the considered environment is that altering the transmit beamforming pattern at some node generally creates more significant changes to interference scenarios for neighboring nodes than variation of the transmit power. Based on this premise, a good neighbor algorithm is formulated in the way that at the sensing node, a new beamformer is selected only if it needs less than the given portion of the transmit power required for the current beamformer. Otherwise, it keeps the current beamformer and achieves the performance target only by means of power adaptation. Equilibrium performance and convergence behavior of the proposed algorithm compared to the best response and regret matching solutions is demonstrated by means of semi-analytic Markov chain performance analysis for small scale and simulations for large scale networks.