The HMM-based Modeling for the Energy Level Prediction in Wireless Sensor Networks

Abstract

The energy conservation is the intrinsic issue of the wireless sensor network (WSN). The lack of energy of some necessary sensor nodes will affect the accurate data aggregation as well as the effective lifetime of the network. In this paper, the problem of predicting which energy level the WSNs or nodes stay in is addressed. Since the primitive method that exchanges messages through nodes could not directly predict the energy level, the stochastic modeling based on hidden Markov model (HMM) is proposed to solve the problem. The energy is divided into several energy levels, i.e. different energy ranges of sensor nodes. Since the state of energy level is the random variable with the value of different energy levels at some time, which holds the characteristic of the Markov chain, it can be modeled as a left-right HMM. Since the energy level cannot be obtained directly, the energy consumption of sensor nodes at a certain time is used as observation states in HMM. Since HMM can perform training process during the running of the WSN, the method can adjust itself to make the best adaptation to the desirable one. Then the trained HMM model can be applied to the prediction of the energy level of the network or sensor nodes. The implementation and simulation of the proposed approach are given in this paper. The simulation shows that the model can effectively predict the energy level and reflect the tendency of the energy consumption, and be applied to other protocols such as energy-efficient routing protocols and so on.