A Reinforcement Routing Algorithm with Access Selection in the Multi–Hop Multi–Interface Networks

In this paper, a routing algorithm is proposed for access selection in a network to find the optimal paths among intermediate nodes with multiple interfaces. Markov Decision Process is applied in each node to find optimal policy and select proper paths to the best access point in a dynamic environment. A reward function is defined as environment feedback to optimize and adapt routing behavior of nodes based on the local information. Selection metrics in each node are interface load, link quality and destination condition. It is shown, by using the proposed algorithm, there are better management in the node which decreases interference and collision and selects links with better quality toward the best possible destination. The performance of the method is exemplified and it is shown how the throughput and average delay of the network with more interface in its nodes, improved while packet loss degrades. As an example a two-interface and a one-interface network are studied. It is shown when network load is increased, interface management will improve the throughput, in the network with two-interface nodes. Also, by considering the link quality factor in the reward function, packet dropping becomes less but average delay increases.