Deep-q-Networks-Based Adaptive Dual-Mode Energy-Efficient Routing in Rechargeable Wireless Sensor Networks

In order to enhance the sustainability of rechargeable wireless sensor networks (RWSN), a deep-q-networks (DQN)-based adaptive dual-mode energy-efficient routing is proposed in this paper. Firstly, the life expectancy of each node is calculated based on multiple related factors, and a multi-hop routing based on forward transmission principle is proposed by using the indicator. Then according to the relationship between the life expectancy of a single node and the average life expectancy of the whole network, an adaptive dual-mode energy-efficient routing is proposed, which combines the multi-hop routing and the direct upload routing. Finally, for reducing the requirement of the single node for the network state information in the process of routing mode selection, a reinforcement learning framework based on DQN is designed, enabling the nodes to learn to judge the above relationship of life expectancy based on partial state information of its local network. Simulation results show that dynamic adjustment of the routing mode enables our algorithm to effectively optimize energy efficiency, so that the network lifetime increases obviously. Based on the limited information, the correct rate of routing mode selection can reach 95%, which ensures the applicability of our algorithm.

Automated summary

In this paper, a DQN-based adaptive dual-mode energy-efficient routing algorithm is proposed to enhance the sustainability of rechargeable wireless sensor networks (RWSN). By calculating the life expectancy of each node and adjusting the routing mode dynamically based on the average life expectancy of the network, energy efficiency is effectively optimized. The reinforcement learning framework reduces the requirement of network state information for individual nodes.