Effects of anchor placement on mean-CRB for localization N. Salman, H. K. Maheshwari, A.H. Kemp, M. Ghogho School of Electronic and Electrical engineering, University of Leeds. Leeds LS2 9JT, UK email: {elns, elhkm, a.h.kemp, m.ghogho@leeds.ac.uk} Abstract In this paper we discuss the lower bound i.e. the Cramer-rao-bound (CRB) on the accuracy of localization of nodes in a wireless network. We investigate the effects of anchor (or base station) placement on optimal target node positioning. The optimal and worst anchor positions are determined through extended simulation by comparing their mean CRB. Furthermore the ramifications of an additive and multiplicative noise model on the mean CRB are explored. Finally, the least squares (LS) method for localization is used and its performance is compared with the lower bound for optimal anchor positions. I. I NTRODUCTION In recent years, there has been a great interest in research towards positioning of wireless devices. Precise and accurate localization is of great importance in military, health, environment, and commercial ap- plications [1]. For outdoor positioning, the global positioning system (GPS) is most widely used, but it exhibits suboptimal performance in harsh propagation environments (i.e. inside buildings, underground and between heavy vegetative cover) due to the absence of a line of sight (LoS). Also, due to the high power requirements of GPS, it has become impractical to employ GPS technology in low power networks such as wireless sensor networks (WSNs). On the other hand, cellular based positioning systems such as the enhanced 911 (E911) have also been recently implemented. With the E911 service, the position of the calling party can be pinpointed in cases when the caller cannot locate itself (e.g. on a remote highway or during a kidnapping). Two widely used methods for range estimation are the time- of-arrival (ToA) and the received signal strength (RSS). Various techniques have been developed to solve the trilateration distance equations. These include the LS methods [2], the weighted LS method [3] and the maximum likelihood (ML) approach [4]. However the performance of these algorithms is bounded by the CRB. The CRB puts a lower bound on the variance of any unbiased estimator. The CRB for localization is dependent on the geometry of the anchors and the target node. However the results in [6] are based on the additive noise model (aNm) while a modified CRB based on the multiplicative noise model (mNm) is proposed in [7]. In this paper, we investigate the optimal anchor placement for both models. We begin by giving a review of the two noise models in section-II. In section III we discuss the CRB for localization. Our simulation results are presented in section-IV. In section-V the LS method is simulated and compared with the lower bound, which is followed by the conclusion. II. SIGNAL MODELS Let us consider a network consisting of N anchor nodes whose locations θ i =(x i ,y i ) T for i =1, ..., N are known, this can be done by placing these anchors at predefined spots or their position can be determined via GPS. It is desired to determine the location of a target node θ =(x, y) T . Then the estimated distance between each anchor and the target node can be modeled either by the aNm or the mNm. The aNm is a widely accepted signal model, however the 0 10 20 30 40 50 -10 0 10 20 30 40 50 60 Range (m) Estimated Range (m) Range (m) σ 2 = 2 σ 2 = 4 Figure 1. Simulation of estimated range for aNm 10 20 30 40 50 0 10 20 30 40 50 60 Range (m) Estimated Range (m) κ = 0.005 Range (m) η=1.83 η=1.87 η=2 η=2.2 10 20 30 40 50 0 10 20 30 40 50 60 Range (m) Estimated Range (m) κ = 0.008 Range (m) η=1.83 η=1.87 η=2 η=2.2 Figure 2. Simulation of estimated range for mNm mNm is more suitable for practical propagation channels. The two noise models are discussed in the following subsections. A. Additive noise model The signal received at the target node from the i th anchor is given by r i (t)= A i s(t τ i )+ n i (t) (1) where A i is the amplitude or attenuation of the signal, τ i is the propagation delay and n i (t) is the thermal noise. The delay τ i that is dependent on the distance between the anchor and the target node is given by τ i (x, y, ℓ i )= 1 c (x x i ) 2 +(y y i ) 2 + i (2) where c is the speed of the electromagnetic wave ( c 3 × 10 8 ) and i is non line of sight (NLoS) bias. Since in the present work, we only consider the LoS case, thus i =0. From Eq. (2), we note that the distance between i th anchor and the target node is given by d i = i (3) To include distances from N anchors, Eq. (3) is given in vector form: 2011 The 10th IFIP Annual Mediterranean Ad Hoc Networking Workshop 978-1-4577-0899-2/11/$26.00 ©2011 IEEE 115