Filter z delci za lokalizacijo v brezžičnih senzorskih omrežjih
Particle filter for localization in wireless sensor networks : doktorska disertacija

slovenian

doktorska dela

2012

Univerza v Mariboru, Fakulteta za elektrotehniko računalništvo in informatiko

This dissertation presents an algorithm for distance determination between a tag node and an anchor node in the Wireless Sensor Network - WSN, with multiple antennas using the Received Signal Strength Indicator - RSSI. Bayesian inference and particle filter are used for distance estimation. With the first order Bayesian inference we determined the distance from RSSI readings, using a preselected radio propagation model (Log-normal or Ground reflection model). The prior within Bayesian inference is modeled using Gauss-Markov Random Field and the likelihood is presented with the Gamma probability density function. Distance estimation is done with Maximum a posterior (MAP) estimation. The second order Bayesian inference was used to estimate the best parameters of the priori and model order (number of antennas), which was done with evidence maximization evaluation. For further distance estimation, we used particle filter with the Sequential Importance Resampling - SIR algorithm. Gaussian probability density function was used to process importance sampling and a comparison between Gamma and Gaussian probability density function was made for importance weights update. Experimental results of the dissertation, which include real RSSI readings andthe estimated distances of Bayesian inference and particle filter, show that it is possible to estimate the distance between the tag node and anchor node with 0,03 m accuracy. The accuracy of the method depends on the space itself and the reflections in it, and from the used models and hardware. Accuracy (distance error) is defined as an absolute value of the difference between the actual and estimated distance.

07.08.2023