A wireless sensor network (WSN) consists of a large number of spatially distributed signal processing devices (nodes), each with finite battery lifetime and thus limited computing and communication capabilities. When properly programmed and networked, nodes in a WSN can cooperate to perform advanced signal processing tasks with unprecedented robustness and versatility, thus making WSN an attractive low-cost technology for a wide range of remote sensing and environmental monitoring applications [1], [32].

Abstract—In this paper, we study the problem of optimal trajectory generation for a team of mobile sensors tracking a moving target using distance-only measurements. This problem is shown to be NP-Hard, in general, when constraints are imposed on the speed of the sensors. We propose two algorithms, modified Gauss-Seidel-relaxation and LP-relaxation, for determining the set of feasible locations that each sensor should move to in order to collect the most informative measurements; i.e., distance measurements that minimize the uncertainty about the position of the target. Furthermore, we prove that the motion strategy that minimizes the trace of the position error covariance matrix is equivalent to the one that maximizes the minimum eigenvalue of its inverse. The two proposed algorithms are applicable regardless of the process model that is employed for describing the motion of the target, while the computational complexity of both methods is linear in the number of sensors. Extensive simulation results are presented demonstrating that the performance attained with the proposed methods is comparable to that obtained with gridbased exhaustive search, whose computational cost is exponential in the number of sensors, and significantly better than that of a random, towards the target, motion strategy.

Local data aggregation is an effective means to save sensor node energy and prolong the lifespan of wireless sensor networks. However, when a sensor network is used to track moving objects, the task of local data aggregation in the network presents a new set of challenges such as the necessity to estimate, usually in real-time, the constantly-changing state of the target based on information acquired by the nodes at different time instants. To address these issues, we propose a distributed object tracking system which employs a cluster-based Kalman filter in a network of wireless cameras. When a target is detected, cameras that can observe the same target interact with one another to form a cluster and elect a cluster head. Local measurements of the target acquired by members of the cluster are sent to the cluster head, which then estimates the target position via Kalman filtering and periodically transmits this information to

Abstract — This paper addresses the problem of cooperative localization (CL) under severe communication constraints. Specifically, we present minimum mean square error (MMSE) and maximum a posteriori (MAP) estimators that can process measurements quantized with as little as one bit per measurement. During CL, each robot quantizes and broadcasts its measurements and receives the quantized observations of its teammates. The quantization process is based on the appropriate selection of thresholds, computed using the current state estimates, that minimize the estimation error metric considered. Extensive simulations demonstrate that the proposed Iteratively-Quantized Extended Kalman filter (IQEKF) and the Iteratively Quantized MAP (IQMAP) estimator achieve performance indistinguishable of that of their real-valued counterparts (EKF and MAP, respectively) when using as few as 4 bits for quantizing each robot measurement. I.

Abstract — In this paper we study the problem of optimal trajectory generation for a team of mobile robots that tracks a moving target using range-only measurements. We propose an adaptive-relaxation algorithm for determining the set of feasible locations that each robot must move to in order to collect the most informative measurements; i.e., distance measurements that minimize the uncertainty about the position of the target. We prove that the motion strategy that minimizes the trace of the position error covariance matrix is equivalent to the one that minimizes its maximum eigenvalue. The proposed method is applicable regardless of the process model employed for describing the motion of the target while its computational complexity is linear in the number of robots. Extensive simulation results are presented, demonstrating that the performance attained with the proposed method is comparable to that obtained with exhaustive search whose computational cost is exponential in the number of robots. I.

Abstract—We consider state estimation of a dynamic process using a wireless sensor network (WSN). Due to bandwidth constraints, analog amplitude observations are quantized to m-bit messages, based on which the state is estimated. We introduce a novel iterative quantization approach such that at time n, the i-th bit is iteratively formed using the sign of the difference between the n-th observation and its estimate based on past observations (up to time n − 1) along with previous bits (up to i − 1) of the current observation. Analysis and simulations confirm that KF-like tracking based on only 2 to 4 bits of iteratively quantized data communicated among sensors exhibits meansquare error (MSE) performance identical to a KF based on analog-amplitude data. Technical Areas: C.2 Wireless sensor networks, E.6 Estimation and detection.

We consider state estimation of a Markov stochastic process using an ad hoc wireless sensor network (WSN) based on noisy linear observations. Due to power and bandwidth constraints present in resourcelimited WSNs, the observations are quantized before transmission. We derive a distributed recursive mean-square error (MSE) optimal quantizer-estimator based on the quantized observations. The resultant Kalman-like algorithm based on quantized observations exhibits MSE performance and computational complexity comparable to the Kalman filter based on un-quantized observations even for 2-3 bits of quantization per observation. Index Terms — wireless sensor networks, distributed state estimation, Kalman filtering, target tracking, limited-rate communication. 1.

Abstract — In this paper, we study the accuracy of Cooperative

Abstract—Estimation of nonstationary dynamical processes is of paramount importance in various applications including target tracking and navigation. The goal of this paper is to perform such tasks in a distributed fashion, using data collected at power-limited sensors which either communicate with a fusion center (FC) over noisy links, or, communicate with each other over nonideal channels in an ad hoc setting. In FC-based wireless sensor networks (WSNs) with a prescribed power budget, linear dimensionality reducing operators which account for the sensor-to-FC channel are derived per sensor to minimize the mean-square error (MSE) of Kalman filtered state estimates formed at the FC. Using these operators and state predictions fed back from the FC online, sensors reduce the dimensionality of their local innovation sequences and communicate them to the FC where tracking estimates are corrected. Analytical and numerical results advocate that the novel channel-aware distributed tracker outperforms competing alternatives. In ad hoc WSNs deployed to perform distributed tracking, one sensor broadcasts reduced-dimensionality data per time slot, according to a prespecified transmission order. The dimensionality reducing operators employed by the broadcasting sensor are selected to meet its transmit-power budget, while minimizing the state estimation MSE of the sensor with the lowest receiving SNR. Based on the received reduced-dimensionality data from the broadcasting sensor, every sensor in range performs the MSE optimal tracking. Corroborating distributed target tracking simulations based on distance-only observations illustrate that the novel scheme provides sensors with accurate estimates at affordable communication cost. Index Terms—Distributed tracking, wireless sensor networks (WSNs), Kalman filtering, target tracking. I.

Abstract—Source extraction was traditionally done by sensor arrays. Recently, sensor networks have been considered as promising candidates for extraction of multiple sources. In a sensor network, each sensor observes an instantaneous linear mixture of the sources and their observations are corrupted by additive white Gaussian noise. Two sensor network models are adopted. The first one is cluster based, in which a sensor acts as cluster head and performs local extraction of the sources based on its own observation and the received quantized data from the cluster members. Then, the extracted signal is quantized and the quantized data are sent to the sink while the sink performs global extraction of the sources. The other one is cluster free, in which data collected by the sensors are quantized and sent to the sink directly. Then, the sink performs global extraction of the sources. The proposed schemes are evaluated against the benchmarking case where the sensor observations are undistorted. Index Terms—Blind source extraction, distributed estimation, wireless sensor network. I.