Dongli Wang

Posterior CramÉr–Rao Lower Bounds for Target Tracking in Sensor Networks With Quantized Range-Only Measurements

We consider the problem of target tracking in a wireless sensor network (WSN) that consists of randomly distributed range-only sensors. Quantized measurements are usually adopted in such a network to attack the problem of limited power supply and communication bandwidth. Assuming that local sensor noises are mutually independent, we derive the posterior Cramer-Rao lower bound (CRLB) on the mean squared error (MSE) of target tracking in WSNs with quantized range-only measurements. Recursion of posterior CRLB on tracking based on both constant velocity (CV) and constant acceleration (CA) model for target dynamics and a general range-only measuring model for local sensors are obtained. Due to the analytical difficulties, particle filter is applied to approximate the theoretical bounds. To illustrate the posterior CRLB, an example on tracking a target with noisy circular trajectories is given.

Unscented Kalman Filtering based quantized innovation fusion for target tracking in WSN with feedback

The quantized innovation fusion approach to tracking a target with nonlinear Gaussian dynamics in wireless sensor network (WSN) is proposed. A hierarchical innovation fusion structure with feedback from the fusion center (FC) to each deployed sensor is proposed. The measurement innovation in each local sensor node is quantized and then transmitted to the FC. Then the FC estimates the state of the target using the Unscented Kalman Filtering (UKF) strategy. To attack the energy/power source and communication bandwidth constraints, we consider the tradeoff between the communication energy and the global tracking accuracy. A closed-form solution to the optimization problem for bandwidth scheduling is given, where the total energy consumption measure is minimized subject to a constraint on the covariance of the quantization noises. Simulation example is given to illustrate the proposed scheme obtains average percentage of communication energy saving up to 41.5% compared with the uniform quantization, while keeping tracking accuracy very closely to the clairvoyant UKF that relies on analog-amplitude measurements.

Target tracking in wireless sensor networks using adaptive measurement quantization

Quantization/compression is usually adopted in wireless sensor networks (WSNs) since each sensor node typically has very limited power supply and communication bandwidth. We consider the problem of target tracking in a WSN with quantized measurements in this paper. Attention is focused on the design of measurement quantizer with adaptive thresholds. Based on the probability density function (PDF) of the signal amplitude measured at a random location and by maximizing the entropy, an adaptive design method for quantization thresholds is proposed. Due to the nonlinear measuring and quantization models, particle filtering (PF) is adopted in the fusion center (FC) to estimate the target state. Posterior Cramér-Rao lower bounds (CRLBs) for tracking accuracy using quantized measurements are also derived. Finally, a simulation example on tracking single target with noisy circular trajectories is provided to illustrate the effectiveness of the proposed approach.

A scalable support vector machine for distributed classification in ad hoc sensor networks

A scalable support vector machine (SVM) is proposed for distributed classification in ad hoc wireless sensor networks (WSNs) in this paper. The main idea is to train SVM classifier using only the local dataset, and evaluate the global nonlinear classifier via a dynamic consensus algorithm with communication only between neighbors instead of among all agents (sensor node) in the network. Specifically, by introducing a sequential gradient ascent based algorithm and modifying the formulation of the bias, the training process can be executed in a distributed and parallel way without information exchange among agents. After the distributed training of SVM, each node has one set of Lagrange multipliers corresponding to the local dataset. Then we adopt the dynamic consensus algorithm to evaluate the global nonlinear classifier for each agent in the network with only information exchange between neighbors. A novel dynamic consensus formulation is introduced and its convergence is proved. Whats more, since it only exchanges information between neighbors during evaluation, the proposed algorithm is scalable for large-scale ad hoc sensor network and considerable communication energy can be reduced, which will prolong the lifetime of the whole network. Examples from the UCI repository demonstrate the effectiveness of the proposed algorithm.

Collaborative target tracking in wireless sensor networks using quantized innovations and Sigma-Point Kalman Filtering

The decentralized collaborative target tracking problem in wireless sensor network (WSN) is investigated in the fusion of quantized innovations perspective. A hierarchical fusion structure with feedback from the fusion center (FC) to each deployed sensor is proposed for tracking a target with nonlinear Gaussian dynamics. Probabilistic quantization strategy is employed in the local sensor node to quantize the innovation. After the FC received the quantized innovations, it estimates the state of the target using the Sigma-Point Kalman Filtering (SPKF). To attack the energy/power source and communication bandwidth constraints, we consider the tradeoff between the communication energy and the global tracking accuracy. A closed-form solution to the optimization problem for bandwidth scheduling is given, where the total energy consumption measure is minimized subject to a constraint on the covariance of the quantization noises. Simulation results illustrate that the proposed scheme obtains average percentage of communication energy saving up to 40.8% compared with the uniform quantization, while keeping tracking accuracy very closely to the clairvoyant SPKF even when the latter relies on analog-amplitude measurements.

Support vector machine for distributed classification: A dynamic consensus approach

A totally distributed and scalable support vector machine (DSVM) for classification in ad hoc wireless sensor networks (WSNs) is proposed. A sequential gradient ascent based algorithm is first introduced and adapted for distributed and parallel SVM training using only the local dataset for each classification agent. Then the global nonlinear classifier is evaluated via a dynamic consensus algorithm with communication between neighbors instead of among all agents in the network. The proposed algorithm is totally distributed and parallel, thus the requirement of data privacy can be satisfied since it requires no information exchange among agents during training. Whats more, since it only exchanges information between neighbors during evaluation, the proposed algorithm is scalable for large-scale sensor network and considerable communication energy can be reduced which will prolong the lifetime of the whole network.

Consensus 3-D bearings-only tracking in switching senor networks

Abstract Target tracking in bearings-only sensor networks (BOSNs) has obtained distinct interest in the last decade. In this situation, the scalability of the tracking algorithm and robustness against network topologies due to moving platform or node/communication fault are two important issues. This motivates the present work on distributed bearings-only tracking in switching BOSNs adopting consensus-based unscented Kalman filters (CoUKFs). First, information unscented Kalman filters (IUKFs) for bearings-only measurements are derived by statistical linearization approach. Then the IUKF is distributed by computing the average consensus on information contribution with only message exchange between one-hop neighbors. To accelerate the convergence in switching networks, adaptive updating of the weights in terms of gradient is proposed for the consensus strategy. Finally, an example of tracking by a network of mixed static and moving bearings-only sensors with switching topologies is given to demonstrate the effectiveness of the proposed method.

Binary tree of posterior probability support vector machines for hyperspectral image classification

The problem of hyperspectral remote sensing images classification is revisited by posterior probability support vector machines (PPSVMs). To address the multiclass classification problem, PPSVMs are extended using binary tree structure and boosting with the Fisher ratio as class separability measure. The class pair with larger Fisher ratio separability measure is separated at upper nodes of the binary tree to optimize the structure of the tree and improve the classification accuracy. Two approaches are proposed to select the class pair and construct the binary tree. One is the so-called some-against-rest binary tree of PPSVMs (SBT), in which some classes are separated from the remaining classes at each node considering the Fisher ratio separability measure. For the other approach, named one-against-rest binary tree of PPSVMs (OBT), only one class is separated from the remaining classes at each node. Both approaches need only to train n – 1 (n is the number of classes) binary PPSVM classifiers, while the average convergence performance of SBT and OBT are O(log2n) and O[(n! − 1)/n], respectively. Experimental results show that both approaches obtain classification accuracy if not higher, at least comparable to other multiclass approaches, while using significantly fewer support vectors and reduced testing time.

Binary tree of posterior probability support vector machines

Posterior probability support vector machines (PPSVMs) prove robust against noises and outliers and need fewer storage support vectors (SVs). Gonen et al. (2008) extended PPSVMs to a multiclass case by both single-machine and multimachine approaches. However, these extensions suffer from low classification efficiency, high computational burden, and more importantly, unclassifiable regions. To achieve higher classification efficiency and accuracy with fewer SVs, a binary tree of PPSVMs for the multiclass classification problem is proposed in this letter. Moreover, a Fisher ratio separability measure is adopted to determine the tree structure. Several experiments on handwritten recognition datasets are included to illustrate the proposed approach. Specifically, the Fisher ratio separability accelerated binary tree of PPSVMs obtains overall test accuracy, if not higher than, at least comparable to those of other multiclass algorithms, while using significantly fewer SVs and much less test time.

Fuzzy Logic Based Interactive Multiple Model Fault Diagnosis for PEM Fuel Cell Systems

The problem of fault detection and diagnosis (FDD) in dynamic systems has received considerable attention in last decades due to the growing complexity of modern engineering systems and ever increasing demand for fault tolerance, cost efficiency, and reliability (Willsky, 1976; Basseville, 1988). Existing FDD approaches can be roughly divided into two major categories including model-based and knowledge-based approaches (Venkatasubramanian et al., 2003a; Venkatasubramanian et al., 2003b). Model-based approaches make use of the quantitative analytical model of a physical system. Knowledgebased approaches do not need full analytical modeling and allow one to use qualitative models based on the available information and knowledge of a physical system. Whenever the mathematical models describing the system are available, analytical model-based methods are preferred because they are more amenable to performance analysis. Generally, there are two steps in the procedure of model-based FDD. First, on the basis of the available observations and a mathematical model of the system, the state variable x and test statistics are required to be obtained. Then, based on the generated test statistics, it is required to decide on the potential occurrence of a fault. For linear and Gaussian systems, the Kalman filter (KF) is known to be optimal and employed for state estimation. The innovations from the KF are used as the test statistics, based on which hypothesis tests can be carried out for fault detection (Belcastro & Weinstein, 2002). In reality, however, the models representing the evolution of the system and the noise in observations typically exhibit complex nonlinearity and non-Gaussian distributions, thus precluding analytical solution. One popular strategy for estimating the state of such a system as a set of observations becomes available online is to use sequential Monte-Carlo (SMC) methods, also known as particle filters (PFs) (Doucet et al., 2001). These methods allow for a complete representation of the posterior probability distribution function (PDF) of the states by particles (Guo & Wang, 2004; Li & Kadirkamanathan, 2001). The aforementioned FDD strategies are single-model-based. However, a single-model-based FDD approach is not adequate to handle complex failure scenarios. One way to treat this