Characterisation of the Accuracy of Indoor Positioning via Received Signal Strength Method

Abstract

Recent RF technological advances in the past decades have made possible the use of wireless sensors networks (WSNs) for a variety of monitoring applications. However, many of these applications depend on the accurate positioning of the sensor nodes. Several methods have been developed for position estimation which include the received signal strength (RSS) method where distance and position of a sensor node can be estimated using appropriate RSS propagation model. However, the RSS is widely known to be easily affected by environment conditions and equipment uncertainties. This thesis focuses on three main objectives which are (i) to investigate whether it is possible to statistically characterise the accuracy of indoor position estimations using the RSS; (ii) to investigate the relationship between limited accuracy in RSS measurements to positioning accuracy; and (iii) to identify the best achievable positioning accuracy based on the RSS in which the uncertainties due to environment and measurement hardware are minimised. To achieve these objectives, a derivation of the statistical models for distance and position estimation of a sensor node from a classical RSS propagation model has been presented. Subsequently, a performance measure which characterises the radius of error of the position estimates as a function of indoor channel parameters has been derived as having a Nakagami-m distribution. The performance of the proposed statistical models has been examined and validated via rigorous Kolmogorov-Smirnov (KS) hypothesis tests using simulated RSS and the RSS measured extensively in three indoor environments. The limits of applicability of the proposed models have also been identified. From the investigation, the proposed radius of error r model has been shown to perform better than the Circular Error Probability (CEP) when the channel parameters are within the identified limit. Modifications to the proposed models have been suggested to improve position estimation in harsh propagation channel conditions. The best achievable positioning accuracy using the RSS measured in a controlled environment (anechoic chamber) where the uncertainties due to environment and hardware are minimised is identified to be 11 cm for 95% of the estimates. For positioning in typical environments, 95% of the estimates are found to reside within 0.99 m from the actual location after modifications have been applied.