Reliable Hybrid Multicast Routing in Mobile Ad Hoc Networks: Reinforcement Learning-based Approach
Gyanappa A. Walikar and Rajashekhar C. Biradar
Mobile Ad Hoc Networks (MANETs) are types of wireless network consist of mobile nodes are communicating with each other via wireless radio links without any underlying physical infrastructure. Due to infrastructure-less, node mobility, and unreliable radio links, reliable routing is the most challenging issue in the MANET. In this paper, we proposed a novel heuristic approach called Reinforcement Learning (RL) based Reliable Hybrid Multicast Routing (RL-RHMRP) which possesses the ability to learn the network context and makes routing decisions in the selection of neighbor nodes and route establishment process. RL-RHMRP works based on reliable intermediate node forwarding mechanism and follows Q-Learning (QL) method (one of the RL technique) with On-policy and Model-based features. Reliable Decisive Factor (RDF) is computed based on measured power level, received signal strength, mobility, and link stability of a node. Our scheme chooses the best path from the mesh of paths for data transmission by considering the computed sum of RDF value for both proactive and reactive MANETs region. Simulation evaluation has been done in NS-2 for various performance parameters like Packet Delivery Ratio, Jitter, End-to-end delay, and Overheads in comparison to the zone-based routing protocols such as ZRP (Zone Routing Protocol) and MZRP (Multicast Zone Routing Protocol). It is observed from the result and discussion section that RL-RHMRP outperforms than ZRP and MZRP.
Keywords: MANETs, Hybrid Routing, Reliability Decisive Factor, Link Stability, Zone Radius, Q-Learning