Skip to Main Content
Frequent changes in topology and link quality in mobile ad-hoc networks (MANETs) present challenging problems in achieving optimal performance. We propose a self-learning routing protocol based on Q-learning that makes use of Quality of Service (QoS) parameters such as Signal to Interference plus Noise Ratio (SINR), delay and throughput, to make routing decisions. At the same time, a Bayesian Network (BN) is implemented to estimate neighboring network congestion level to tune the Q-learning weights. Our protocol also sends out probing packets to detect and solve the routing-loop problem which is not addressed in most Q-learning-based routing proposals. The simulation results show that the proposed system demonstrates comparatively better performance in a dense heavy-loaded scenario.