A new Reinforcement Learning based for Energy-efficient Multi-channel Data Gathering in Wireless Sensor Networks

Wireless sensing networks (WSNs) have attracted widespread attention in the last few years as they have become necessary in recent fields such as the Internet of Things (IoT) which have caused in countless applications. The use of multichannel technology in such kind of networks represents a challenging field due to its advantage in improving throughput and latency, while the major challenge that faces WSNs is the drain of energy. Moreover, adapting to the dynamic property of the transmission flow for channel assignment in an energy-efficient manner is considered as an NP-hard problem. Hence, a Reinforcement Learning (RL) approach is proposed to overcome this challenge. Meanwhile, the use of the RL approach requires a number of iterations to obtain the best solution which in turn creates a considerable communication overhead and collisions which leads to more energy consumption. In this paper, we propose an Energy-efficient method for Reinforcement Learning based Multi-channel MAC (ERL MMAC) that performs a hybrid channel assignment using a decentralized tree for multi-channel data gathering in WSNs. The proposal focuses on the reduction of energy consumption by using the least chosen default channel allocation in two hops rather than one hop in order to reduce as much as possible the conflict links in one side, and use of parent selection strategy rather than parent default channel selection strategy in the learning phase to avoid the redundant data messages in the other side. The results of extensive simulation experiments show the effectiveness of our approach in improving network lifetime with a rate of 97.53%.