Novel positioning algorithms and node configuration optimisation for non-gps wireless systems

This thesis presents a system study on the theory, methodology, and architecture for non-GPS based wireless positioning systems. In particular, the focus is to devise new algorithms and methodologies to improve the positioning accuracy of existing methods, and to develop the theoretical framework for node configuration optimisation. Node configuration is one of the main subjects studied in this thesis. To understand the effect of node configuration on positioning accuracy, we first examine the challenging scenario of tracking swimmers in a swimming pool using a two dimensional (2D) and a three dimensional (3D) distributed sensor network, respectively. By employing the spherical interpolation approach (SIA) as the positioning algorithm and the cumulative distribution function (CDF) as the objective function, a methodology for optimising the configuration of the positioning network for tracking swimmers is proposed. Secondly, we propose a hybrid optimisation algorithm combining the particle swarm optimisation (PSO) and the classical sequential quadratic programming (SQP) method for node placement optimisation using the geometric dilution of precision (GDoP) as the objective function. It is observed that the result of such an optimisation strategy is different from that obtained from the CDF method. This seemingly contradictory finding could potentially be beneficial in practical network deployment: Given a coverage area, the optimal positions of the node obtained from GDoP optimisation would result in the lowest Cramer-Rao lower bound, but such a configuration would require the use of an optimum positioning algorithm. If one doesn't have the option of choosing the algorithm, as in the case when the positioning algorithm is embedded in the device at hand, the CDF based method would result in a positioning network with greater achievable accuracy. Triangulation based positioning methodology employing direction-of-arrival (DoA) estimation suffers from a deficit in the antenna radiation pattern when the impinging signal is close to the antenna plane. To overcome this difficulty, we propose a weighted least squares method (WLSM) with the weighting matrix derived from the angle of incidence of each array for wireless positioning using multi-angulation. The algorithm is suited for both 2D and 3D positioning when a number of sensor arrays are employed, and it is proven to be far superior to the conventional least squares method (LSM). To take advantage of the proliferation of low cost wireless ranging devices, such as wireless LAN and ultra wideband (UWB), a new positioning method employing ranging based sensor arrays is presented and the associated location estimator is developed. Such an array is of low cost and is easy to maintain. The Cramer-Rao lower bound of the proposed location estimator is derived. A closed form expression of the standard deviation is given and proved to reach the Cramer-Rao lower bound. Finally, a co-operative positioning system based on the co-operation between adjacent sensor arrays is proposed. The Cramer-Rao lower bounds for two arrays and three arrays for 2D and 3D cases, respectively, are derived. It is shown that the co-operation between arrays lead to much improved positioning accuracy compared with the employment of a stand-alone array and non-co-operative arrays, thus reaching the Cramer-Rao lower bound.