A Q-Learning-based Power-Controlled Routing protocol in multihop wireless ad hoc network

In wireless ad hoc networks, power control has great impact on routing since transmission range is directly determined by a node's transmission power. Higher power can give higher connectivity and shorter path. However, larger transmission range causes more interference to nearby neighbors and may further impair overall network performance. We propose a Q-Learning-based Power-Controlled Routing (QLPCR) protocol which makes use of Q learning techniques for routing and power control to optimize delay performance of the whole network. A Markov chain CSMA/CA delay model is used to estimate delay of each link in order to determine the optimal power level for all possible routing options.