Abstract
In Wireless Sensor Networks (WSN), the energy problem plays the critical role in network performance and lifetime, because of the limited battery capacities of sensor nodes. Recently, emerging wireless charging technologies provide a promising approach to address the energy problem in WSN. Researchers construct Wireless Rechargeable Sensor Networks (WRSN), which introduce mobile chargers with high capacity batteries to charge sensor nodes. Most studies in WRSN have paid attention to charging static nodes or mobile nodes with deterministic trajectories. In this work, we explore how to charge nodes with non-deterministic mobility. We propose a novel approach, named Predicting-Scheduling-Tracking (PST), to perform charging tasks in this case. In the proposed scheme, different from the existing work, we guide the mobile charger to chase the sensor and recharge it. In our work, the base station runs an improved LSTM to predict the future locations of nodes periodically. Then, the mobile charger can select an appropriate node as the charging target by a charging scheduling algorithm. During the energy transferring, a Kalman-filter-based tracking algorithm is used to ensure the charging-required distance between the mobile charger and the target node. The simulation results show that the proposed charging scheme can fulfil the charging tasks in WRSN of nodes with non-deterministic mobility.
Highlights
Wireless sensor networks (WSN) comprise one or more base stations and many sensor nodes placed in a large area to monitor a physical environment cooperatively [1]
For this case, [13] proposed an interesting solution by exacting some fixed locations in the application area, called hotspots which are frequently visited by sensors and scheduling the mobile charger (MC) to wait and charge sensors at these hotspots based on a Reinforcement Learning (RL)-based algorithm
PERFORMANCE COMPARISON we evaluate the performance of the proposed charging scheme by comparing it with RL-based Charging Algorithm (RLC) [13]
Summary
Wireless sensor networks (WSN) comprise one or more base stations and many sensor nodes placed in a large area to monitor a physical environment cooperatively [1]. The nodal movement is out of control and non-deterministic, even the mobility model itself is the target of the network For this case, [13] proposed an interesting solution by exacting some fixed locations in the application area, called hotspots which are frequently visited by sensors and scheduling the MC to wait and charge sensors at these hotspots based on a Reinforcement Learning (RL)-based algorithm. This approach can’t adapt the changes of nodal mobility pattern, since the hotspots never change once they are selected.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
