Route Planning And Monitor Data Predicting Based Approach For WSN Optimizing

In 21 st century, Wireless Sensor Network(WSN) develops a lot, and has already been used or showed great value in various areas, such as environment monitoring, biomedical, etc. WSN based monitoring is one of the most prevalent application currently. With the development of technologies and demand of higher performance, several defects have revealed in traditional WSN monitoring systems, such as communication collision and short lifetime, which obviously hinder the further progresses of WSN.In this paper, research work is conducted in the following three crucial procedures of wireless sensor monitoring:(1) data acquisition and transmission;(2) monitor data prediction;(3) monitor data visualization.(1) To improve the reliability of data transmission, lifetime and performance of real-time communication in WSN, this paper takes a study of data transmission route planning. Due to the problem of transmission route planning in WSN is NP-hard, an Integer Linear Program(ILP) based approach is designed in this paper. The experiment results show that the proposed transmission route planning method is efficient to improve the reliability of WSN transmission routes.(2) Machine learning is introduced to predict monitor data for restoring the corrupted data and forecasting the upcoming data in WSN. In this paper, analyses are firstly conducted from different aspects on two real monitor datasets obtained from citrus plantation and industrial plant. Next, an improved Hidden Markov Model(HMM) is proposed, which combines K-means clustering algorithm and Particle Swarm Optimization(PSO). To predict monitor data accurately, the Improved HMM will cluster all the monitor data into several classes based on different varying patterns, and then generate optimized HMM parameter settings for each specific class. Therefore, the Improved HMM has the capability to accurately predict WSN monitor data which contain different varying patterns. For comparison, a number of methods are considered: Naive Bayes, Grey System, BP Neural Network and traditional HMM. Experiment results demonstrate that the proposed HMM is able to provide higher accuracy for one-step and multi-step monitor data predicting than the other methods.(3) A background monitor software is implemented with Microsoft.Net platform and Rich Internet Application technology for monitor data visualization. Latest monitor data will be continuously presented to users through this background software. Moreover, various thresholds are embedded in this background monitor software for real-time alarming. Overall, this software provides a basic data visualization platform in “Application Layer” for WSN monitoring systems.