UWB technology for indoor localization: from theory to practice

Abstract

In this dissertation, the localization of targets in indoor environment by means of the Ultra Wide Band (UWB) technology has been investigated from analytical, simulative and experimental points of view.

In Chapter 1, the motivations of the importance of indoor localization are presented and an overview of the literature on this topic is given. A variety of approaches suitable to address this problem is described and the advantages brought by the UWB technology in this context are emphasized.

Aiming at finding strategies to improve the localization accuracy, the presented work initially focuses on optimized placement of Anchor Nodes (ANs) used for target localization. In particular, in Chapter 2, an analytical approach to optimized ANs placement for UWB-based localization of a target moving in a large indoor scenario is proposed. Imposing (realistic) constraints on the ANs positions and assuming that the target moves along a straight line in the middle of a corridor, a closed-form expression for the optimal distance between consecutive ANs is derived. The validity of the analytical framework is confirmed by simulations, which also show that the proposed placement strategy is effective even when the TN follows generic paths.

However, in practical application scenarios it is not always possible to decide the ANs positions. For this reason, to overcome the limitations of geometric techniques, which can suffer from ill-conditioning, an optimization-based approach to localization is proposed. More precisely, in Chapter 3, three approaches to UWB-based localization of nodes are considered: two stem from the geometry oriented localization literature and one originates from the soft computing literature, namely the Particle Swarm Optimization (PSO) algorithm. The obtained results show that the PSO approach guarantees, with respect to the other algorithms, a better accuracy in the position estimate. Furthermore, an improved version of the PSO algorithm is proposed. Simulation results show that the convergence speed of the improved algorithm is significantly higher than that of the PSO algorithm, leading to a relevant reduction of the total computational cost.

In order to perform a more realistic analysis of the localization accuracy, experimental results are then considered. In Chapter 4, a statistical model for the distance estimates in Line-of-Sight (LoS) conditions is derived on the basis of an experimental campaign of range measurements obtained using Time Domain PulsON 410 Ranging and Communications Modules (RCMs). The range estimate error between pairs of RCMs is statistically analyzed and the results show that its average and standard deviation are well approximated as linearly increasing functions of the true distance between the considered pair of nodes. The derived statistical model is applied to correct the predictions of two illustrative localization algorithms in two realistic localization scenarios. Results show that the distance corrections introduced by the proposed model allow improving the performance of the considered algorithms, thus confirming the validity of the model.

In Chapter 5, a real-world scenario is considered, where the presence of a metal obstacle introduces Non-Line-of-Sight (NLoS) effects making localization harder. First, two variants of a geometrically inspired algorithm for the localization of a target by means of three ANs are proposed. The results obtained in our experiments with PulsON 410 RCMs are shown for different target positions. Then, the adopted localization algorithm is rearranged for the case of four ANs. Results show that the use of an additional AN allows improving the performance of the localization in terms of the average distance between the true target positions and their estimates.

In conclusion, the possibility of using UWB technology for accurate indoor localization is investigated in this dissertation. The obtained results show that it is indeed possible to achieve 2-3 cm accuracy if all nodes are in LoS and 20-40 cm accuracy in NLoS scenarios.