Reinforcement learning movement path for multiple mobile sinks in wireless sensor networks

Mobile sink nodes play a very active role in wireless sensor network (WSN) routing. Because hiring these nodes can decrease the energy consumption of each node, end-to-end delay, and network latency significantly. Therefore, mobile sinks can soar the network lifetime dramatically. Generally, there are three movement paths for a mobile sink, which are as follows: (1) Random/stochastic, (2) controlled, and (3) fixed/ predictable/predefined paths. In this paper, a novel movement path is introduced as a fourth category of movement paths for mobile sinks. This path is based on deep learning, so a mobile sink node can go to the appropriate region that has more data at a suitable time. Thereupon, WSN routing can improve very much in terms of end-to-end delay, network latency, network lifetime, delivery ratio, and energy efficiency. The new proposed routing suggests a reinforcement learning movement path (RLMP) for multiple mobile sinks. The network in the proposed work consists of a couple of regions; each region can be employed for a special purpose, so this method is hired for any application and any size of the network. All simulations in this paper are done by network simulator 3 (NS-3). The experimental results clearly show that the RLMP overcomes other approaches by at least 32.48% in the network lifetime benchmark.