Factorial Design Analysis for Localization Algorithms

Abstract

Localization is a fundamental problem in Wireless Sensor Networks, as it provides useful information regarding the detection of an event. There are different localization algorithms applied in single-hop or multi-hop networks; in both cases their performance depends on several factors involved in the evaluation scenario such as node density, the number of reference nodes and the log-normal shadowing propagation model, determined by the path-loss exponent (η) and the noise level (σdB) which impact on the accuracy and precision performance metrics of localization techniques. In this paper, we present a statistical analysis based on the 2k factorial methodology to determine the key factors affecting the performance metrics of localization techniques in a single-hop network to concentrate on such parameters, thus reducing the amount of simulation time required. For this proposal, MATLAB simulations are carried out in different scenarios, i.e., extreme values are used for each of the factors of interest and the impact of the interaction among them in the performance metrics is observed. The simulation results show that the path-loss exponent (η) and noise level (σdB) factors have the greatest impact on the accuracy and precision metrics evaluated in this study. Based on this statistical analysis, we recommend estimating the propagation model as close to reality as possible to consider it in the design of new localization techniques and thus improve their accuracy and precision metrics.

Localization Based on Probabilistic Multilateration Approach for Mobile Wireless Sensor Networks

“…Assuming that the NOI Z is moving, the range δ A from the RN A 0 to the NOI Z will be affected by a beta random variable ϕ A due to the mobility of the NOI Z. Therefore, the range δ A from the RN A 0 to the NOI Z is computed through (22) where the beta random variable ϕ A is weighted by the factor λ A proportional to the real separation distance between the NOI Z and the RN A 0 , i.e., 0 < ϕ A < 1 λ A . Now, we assume in equation 22that the Gaussian and beta random variables are independent.…”

“…. , t, then, by using the equation (22) we can estimate the position of the NOI Z regardless of the value of the instant of time τ . through the Pythagorean theorem as follows…”

“…Taking Figure 4 as a reference, the distance δ A from the RN A 0 to NOI Z can be estimated using equation (22). However, in a network with N number of RNs, we get the estimated distance δ i between the mobile node and the RN i given by…”

“…The RNs are nodes whose position is known, and we can achieve this by either equipping them with a Global Positioning System (GPS) device or distributing them in strategic zones with a known location. GPS is a technology able to estimate the position of a mobile device on a particular geographic area; but due to the system higher costs, higher power consumption and inefficiency in indoor environments, it is not the best fit [22]. In addition to the GPS issue, cell phone and WiFi systems do not perform well in certain scenarios such as highlands, underground and disaster zones where the satellite signals or signals from the mobile infrastructure cannot be received [23].…”

Average Consensus over Mobile Wireless Sensor Networks: Weight Matrix Guaranteeing Convergence without Reconfiguration of Edge Weights

“…When compared to static WSNs, MWSNs find a wider variety of applications, since the mobility of entities results in many advantages, such as reliability, cost, improved coverage, connectivity, energy efficiency, dynamic topology, increased channel capacity, etc. [ 26 , 27 , 28 ]. Moreover, MWSNs can easily monitor also moving targets (e.g., animals, people, chemical clouds, etc.)…”