Research And Implementation Of RSSI Positioning Algorithms For Underground Based On Wireless Sensor Network

China is rich of coal reserves, and it is the first biggest coal mine production country inthe world. With the development of pervasive computing, sensing technology, big data, andwireless technology, the Internet of things will bring great changes to various industries.Wireless sensor network is a crucial component of the Internet of things, its application usingin "Sensing coal mine" will have a profound impact on the whole coal mine production.As a part of safety monitoring system, underground target localization could remedythe limitations of the satellite positioning, which cannot provide localization services in coalmine. It has brought the important guarantee for safety production. At the moment, China’sunderground positioning systems are mostly still using radio frequency identification system,rather than the practical target tracking system. Therefore, the research work done onlocalization algorithm is briefly analyzed. This paper mainly studied for the wireless sensornetwork localization technology based on RSSI oriented to coal mine.First of all, this paper introduces the basic concept, principle and classification oflocalization algorithm, and several classical algorithms were studied. The main advantagesof parameter estimation localization algorithm are that the signal parameter (such as traveltime, receiving signal strength and receiving angle) can easily be obtained from the hardware,and the work of samples is light. After that using parameter estimation algorithm can obtainthe target location, however signal parameters are easy to be affected in complexenvironment, such as the indoors or underground. Fingerprint matching algorithms utilizethose signals affected by the environment as a part of the fingerprint information. Thepositioning accuracy of using pattern matching algorithms is higher in severe environment.Secondly, aiming at the coal mine channel environment, this paper proposes afingerprint matching positioning algorithm based on virtual data sets and Markov chain. Inthe offline phase, Kalman filter theory has been used in order to obtain more stable datasamples. Because workload of data sampling is heavy by fingerprint matching algorithm, the paper puts forward a scheme constructing virtual data sets. In this paper, fingerprintmatching algorithm is introduced in detail, and probabilistic algorithms consider the RSSIpriori distribution assumptions and statistical characteristics at each sampling point. Toconstruct probability density distribution, each sample at each sampling point is given a"kernel" with itself by kernel function method which can avoid errors caused by determiningmodel. In order to optimize the positioning results, probabilistic algorithms, based onBayesian framework, consider prior probability having influence on posterior probability. Inhence, Markov chain positioning algorithm based on Gaussian model is proposed, thentracking the target by particle filter is simulated.Finally, a positioning network based on ZigBee by using Z-Stack produced by TexasInstruments Company and IOT-2530node build in coal mine. The offline data is stored inMySQL database system through Matlab. Experiments show that the positioning accuracy isimproved by Kalman filtering data sets, and the establishments of a virtual data sets reducethe offline workload. Within length of80m region, the positioning mean square error is4.52m by Markov chain positioning algorithm based on Gaussian model.