AOR: Adaptive opportunistic routing based on reinforcement learning for planetary surface exploration

Planetary surface exploration mobile ad-hoc networks (PSEMANET) show the characteristics of routing void caused by craters and electromagnetic interference, relatively high dynamics caused by node heterogeneity, and small node capacity caused by load constraints, which is the reason for the increase of communication delay and packet loss. Adaptive opportunistic routing based on reinforcement learning (AOR) is an opportunistic routing protocol that determines the forwarding node based on the current coordinates, queue length, and the number of neighbors. The full forwarding areas are used to avoid routing void, and the dual competition mechanism is designed to avoid the node hidden problem. The Q-value obtained by reinforcement learning is used to design the dynamic delay cost (DDC) so that nodes in the forwarding areas can compete to achieve adaptive path switching to the environment. Compared with representative proactive routing OLSR, reactive routing AODV, geographic routing GPSR and opportunistic routing BLR, AOR and AOR with memory(AOR-M) have high packet delivery ratio (PDR) and low end-to-end delay in planetary surface exploration scenarios implemented by our simulation system. And the AOR-M protocol shows the lowest expected end-to-end delay and highest channel utilization in both high and low speed scenarios. The result shows that the AOR-M provides efficient and robust routing in planetary surface exploration mobile ad-hoc networks with a complex environment, highly dynamic nodes, and performance constraints.