Performance Analysis of Received Signal Strength Indicator-Based Localization for Underground Mines

Abstract

Wireless sensor networks (WSNs) have been implemented in a variety of scenarios, including tracking, smart grid, and equipment monitoring. Often times, devices in WSNs are required to know their locations with respect to other devices in the network. This act, known as localization, is the art of predicting the location of a device using anchor devices with known locations in the same network. Localization is essential for many critical industrial applications including underground mines, as the location of devices and employees is vital in helping to guarantee the safety of mining personnel. To the best of our knowledge, existing works have not yet performed extensive research into the use of numerical lters for localization accuracy enhancement. Other works have looked into the use of a speci c numerical lter, however none have covered all of the lters used in this paper. In this thesis, a WSN system that performs localization in an underground mine is proposed. This system uses the Zigbee communication protocol, received signal strength indicator (RSSI) measurements and the trilateration algorithm. The focus of the system is to maximize the accuracy of the predicted location by minimizing the radial error, which is the distance between the predicted location and actual location of a node. Point-to-point tests are performed to determine the best choice of hardware and RSSI model. These were determined to be Texas Instrument's (TI) CC2538 devices and the Piecewise RSSI model respectively. Next, trilateration experiments were performed using nodes containing the CC2538 devices positioned in

a star topology. The localization results showed that the system was able to predict the location of a node with some error, which can be improved. Lastly, this thesis discusses the use of numerical ltering algorithms to enhance the accuracy of the predicted location by decreasing the radial error. These lters are as follows: the Statistical Average (SA) lter, the Gaussian lter, the Gaussian and Statistical Average (GSA) lter, the Kalman lter, the Kalman and Statistical Average (KSA) lter, as well as the Alpha-Trimmed Mean (ATM) lter.