Simulation-based Analysis of a Localization Algorithm for Wireless Ad-Hoc Sensor Networks

Localization is a problem of estimating the spatial coordinates of wireless nodes in an ad-hoc network. Wireless Sensor Network is an example of such a network, where localization as a problem has been a challenging topic for many years. The position of sensor nodes can be either manually configured before deployment or a GPS receiver can be built into each of these nodes. The former approach is very tedious and error-prone while the latter is a costly proposition in terms of volume, money and power consumption. In this paper we discuss a localization algorithm with few beacons having known position estimates assisting unknown nodes in obtaining their positions based on simulation analysis and performance. We propose a Network model to depict a Wireless Sensor Network with 802.11 based wireless nodes available as part of the OPNET Radio Model. We present simulation scenarios and analyze them based on the number of beacon packets transmitted and the accuracy of position estimate obtained. A simple Node Model and Process Model implemented to carry out the localization algorithm is also discussed in detail. We also provide comparative analysis of the simulation results and the theoretically expected values for the scenarios considered. INTRODUCTION A wireless sensor network consists of a collection of wireless sensor nodes, each of which consists of sensing, processing and communicating components. The sensing component closely interacts with the physical world (e.g. measuring temperature or humidity); the processing component processes the measured data, while the communicating component delivers the processed data to a central agent or base station. Consider a scenario involving a forest fire. Cluster of nodes thrown randomly and equipped with heat sensors detect the fire, and immediately report the event to a base station located at a safe distance. The base station will then have an accurate picture of the event and can trigger an immediate action. Another example could be where thousands of tiny wireless sensor nodes are sprinkled on a battlefield to monitor enemy movements without alerting the enemy to its presence. By self-organizing into a wireless sensor network, the sensor nodes would filter out raw data for relevance before relaying only the important findings to the central command. For a sensor network to be useful, it is vital to know the position of the sensor nodes. In the scenario considered above, the sensor location information is important for the base station to deduce where exactly the event (forest fire) has occurred. The problem of estimating the position or spatial coordinates of wireless sensor nodes is termed as localization. The position of sensor nodes can be either hand-placed, but this is a tedious and error-prone method especially when the number of sensors is increasingly huge. Alternatively, each sensor node could be equipped with a GPS (Global Positioning System) receiver, which is a costly proposition in terms of volume, money and power consumption. In this paper, we consider a novel, robust and RF signal strength based distributed algorithm for localizing wireless sensor nodes. In our network very few nodes have a priori knowledge of their position called beacons while nodes, which estimate their position with the help of beacons, are called unknowns. The main focus of this paper is to give a brief overview of the position estimation algorithm and evaluate its performance using simulation-based analysis. RELATED WORK Numerous localization systems have been developed and deployed during the past few decades. While most of these systems rely upon a fixed infrastructure, fewer exists in the ad-hoc domain. In 1993, the Global Position System (GPS) [1] was introduced which is based on the NAVSTAR satellite constellation. LORAN [2] operates in a similar fashion as GPS but uses ground-based beacons instead of satellites. In 1996, the US Federal Communications Commission (FCC) required that location information of all users in a cellular network be provided for Emergency 911 services. Among the indoor localization systems, RADAR system [3] can track the location of users inside a building and is RF based, while the Cricket [4] location support system is ultrasound based. In [5] an iterative multilateration is considered. The algorithm performs well when a large number of beacons are present, the graph connectivity is high and precise range measurements are available. In [6], the problem of cooperative multilateration is tackled. This method relies on precise range determination technique and then on solving a least square on a large order system. An interesting idea is explored in [7], where the problem of localization is considered in the absence of beacons. The nodes build local coordinate system and further aggregate them into a unique network coordinate system. A method for estimating unknown node positions in a sensor network based exclusively on connectivity-induced constraints is described in [8]. Known peer-to-peer communication in the network is modeled as a set of geometric constraints on the node positions. The global solution of a feasibility problem for these constraints yields estimates for the unknown positions of the nodes in the network. Another original idea is presented in [9] where autocalibration is used to improve the accuracy of a localization algorithm. The authors impose common sense constraints (e.g. the distance from A to B equals the distance from B to A, as well as the triangle inequality) on the position of the nodes, and thus auto-correct the range measurements. LOCALIZATION ALGORITHM In this section, we provide a brief overview of the proposed position estimation algorithm. Consider Fig. 1, which shows an example of a wireless ad-hoc network. A straight line between two wireless nodes shows that the two nodes are within each other’s range. Every node in the network will belong to one of the two categories, beacons and unknowns. Beacons have a priori estimate of their own position, which can be either manually configured before deployment or equipped with a GPS receiver. Unknown nodes estimate their position with the help of assisting beacons. Fig. 1 A Wireless Ad-Hoc Network Beacons send beacon packets, which are packets meant to assist unknown nodes in estimating their position. We assume that a range measurement method is available, i.e. an unknown node receiving a beacon packet will, with some confidence, estimate itself to be located within a ring given by circles of radii 1 − i R and i R . Several range measurement techniques have been proposed. One such method uses ultrasonic impulses emitted by the beacons [10]. The distance is calculated from the propagation delay and the propagation speed, which is usually constant. Another method utilizes received signal strength (RSS) measurements, which is available in most of the current transceivers. Though this method is not as accurate as the acoustic one, it does not require additional hardware and hence a cheap solution. Also, this method is reliable only outdoors and its performance and accuracy falls considerably indoors due to fading, interference and multipath propagation. The position estimation algorithm presented in this section is RSS based and is specifically geared to capture the inaccuracy of radio signal strength measurements. Every beacon node transmits beacon message, which includes its own position estimate, which is a point or a small area corresponding to GPS inaccuracy. An unknown node upon receiving a beacon message from a beacon node computes the constraint, which in this case is a transmission ring, and intersects the constraint with its current estimate to calculate the new position estimate. If the position estimate improves, it will broadcast its estimate to all its neighbors. If the beacon message is from another unknown node, the constraint is slightly difficult to compute. The new constraint in this case is given by the Minkowski sum [11] of the position estimate and the transmission ring. Given two surfaces 1 S and 2 S , their Minkowski sum is given by the union of all translations of 2 S in each and every point