An efficient multicast multi-rate reinforcement learning based opportunistic routing algorithm

Multicasting through device-to-device communication (MD2D) is a promising solution for handling the heavy load caused by the extraordinary high traffic in 5G cellular networks. One of the most important challenges of effective multicast in wireless networks is routing. Traditional multicast routing is not a good option in wireless environments due to the limited battery and energy of the users and users’ mobility, leading to entering and leaving multicast groups. Opportunistic Routing (OR) was introduced to meet these challenges in wireless networks. In OR, the forwarding nodes are selected along the way and in each hop from the packet receivers. Although multicast opportunistic routing solves these problems to some extent, it poses challenges such as suitable forwarder set selection, forwarder nodes’ coordination, and the crying baby problem. Multi-rate Multicast Reinforcement Learning based Opportunistic Routing (2MRLOR) is proposed in this paper to deal with these problems. In this algorithm, the transmission ranges of the nodes are altered by changing their transmission rate. Therefore, the neighboring nodes will also vary, and more candidates may be available for the packet forwarding. In 2MRLOR, we introduce a routing metric called EMD (Expected Multicast Delay) to determine the best forwarders along the packets’ way in a multi-rate condition. In this algorithm, the suitable transmission rate of each node is calculated based on the network conditions. Reinforcement learning is also used to reduce the amount of information exchanged between the nodes in the network. Furthermore, network coding is used to facilitate the forwarders’ transmission and eliminate the need for forwarders’ coordination. Based on the simulation results, the proposed algorithm leads to an increase in network throughput and a reduction in end-to-end delay in the network compared to the benchmark algorithms.