Approved for public release; distribution is unlimited Self Localization of Acoustic Sensor Networks Randolph L. Moses, Robert M. Patterson, and Wendy Garber Department of Electrical Engineering, The Ohio State University 2015 Neil Avenue, Columbus, OH 43210 USA Abstract We present algorithms for self-localization of a network of sensors. We consider the case when no “anchor” nodes with known locations are present. We use source signals in the scene, also at unknown locations, to estimate time-of-arrival and direction-of-arrival between sources and sensors. These measurements are used to compute maximum likelihood relative calibration solutions, in which sensor nodes are localized and oriented with respect to one another. Prior location information, in the form of uncertain aimpoints for a subset of the sensors, are then be used to obtain maximum a posteriori estimates of absolute locations and orientations. We derive analytical statistical performance bounds for the two estimators, and present examples that illustrate the performance of the algorithms. 1. Introduction Sensor networks are becoming increasingly important for distributed sensing in a large number of military and nonmilitary applications [1]. A sensor network consists of a large number of low-cost, self-powered sensors that are capable of sensing signals, processing those signals, and communicating with other sensors or higher-level processing centers for data fusion and collaborative decision making. In order to effectively fuse sensor information, it is often important to know the location and orientation of each sensor in the network. However, accurate sensor location and orientation is difficult to provide in many types of sensor deployment. Thus, there is interest in developing methods to find the locations and orientations (that is, to self-calibrate the sensor network) after the sensors have been deployed. Self-localization in sensor networks is an active area of current research (see, e.g., [2, 3, 4, 5] and the references therein). Iterative multilateration-based techniques are considered in [5]. Bulusu et al. [2, 6] consider a low-cost localization methods that use a number of beacon signals at known locations. Research on blind beamforming considers a related problem of forming a maximum power beam to a source without computing the source locations [7]. Cevher and McClellan consider sensor network self-calibration using