Reinforcement learning-based real time search algorithm for routing optimisation in wireless sensor networks using fuzzy link cost estimation

Abstract

Internet of things is a technological advancement of wireless sensor networks (WSNs) which are characterised by highly complex, large scale, heterogeneous, dynamically changing and asymmetric networks. Such constraints make routing in WSNs a difficult task. This paper introduces fuzzy link cost estimation-based real time search routing algorithm (fuzzy RTS) in which link cost estimation is obtained from physical and MAC layer parameters like residual energy, packet drop rate and RSSI. Its performance has been evaluated with traditional reinforcement learning-based algorithms like real time search, adaptive tree, ant routing and constrained flooding algorithms on the basis of metrics like throughput, loss rate, success rate, energy consumption, energy efficiency and node battery life. The simulation results reveal that fuzzy RTS algorithm is most appropriate reinforcement learning-based routing algorithm among given algorithms for ensuring energy efficient and QoS aware routing in dynamically changing, asymmetric and unreliable environment of WSNs.