Strong Tracking Filter with__ Bandwidth Constraint

X.Feng is with College of Automation, Hangzhou Dianzi University, Hangzhou 310018, China (e-mail:ioptmyloving@).
C.Wen is with College of Automation, Hangzhou Dianzi University, Hangzhou 310018, China (e-mail:wencl@).
I. INTRODUCTION
Wireless Sensor Networks (WSNs) is a new frontier technology for signal processing and is becoming one of the hot research fields. With the rapid development of some new technologies, such as communication, computer, and information processing, and sensor technology and so on, the degree by using multi-sensor information fusion technology has been greatly expended and enhanced. It has a wide and important application foreground in automatic target recognition, environmental monitoring, and robots, and image processing etc. Synchronously, it has attracted the attention from researchers and engineers [1-5].
Manuscript received October 30, 2009. This work was supported in part by National Nature Science Foundation (Grant.60804064), Technology Department of Zhejiang Province (No.2009C34016) and Zhejiang Graduate Innovation Project (YK.2008061).
Tingliang Xu, Quanbo Ge Member, IEEE , Xiaoliang Feng and Chenglin Wen Member, IEEE
Abstract—Limited bandwidth is an unavoidable constraint for the information transmission from local sensors to the estimation center over sensor networks. For a networked target tracking system with single sensor, which consists of a state-vector and an observation-vector, the quantized state estimation is researched under the centralized frame with bandwidth constrained in this paper. Firstly, by introducing Strong Tracking Filtering (STF) to deal with uncertain measurement model induced by the current variance approximation of quantization error for adaptive bits quantization error, a novel quantized strong tracking filter is presented. Secondly, the estimation performance for quantized filter with different quantization intervals is discussed. Finally, two numerical examples are demonstrated. Index Terms—Sensor networks, bandwidth constrained, quantization, Kalman filter, strong tracking filtering.
Aiming at the above problem and in order to improve the estimation accuracy, a strong tracking filter with quantized measurement is presented by introducing strong tracking filtering (STF) theory. It can improve effectively the estimation precision of the current Kalman Filter with Quantized Measurement in [10]. More importantly, the quantized estimator which adopts STF has effective strong tracking ability to the target maneuver or state abrupt change. Synchronously, we also compare the filter performance for different quantization intervals at the same quantization bandwidth bit.
communication bandwidth between nodes and estimation center is not sufficient to transmit integrated measurement data, it is necessary to quantize the measuremenቤተ መጻሕፍቲ ባይዱ data. Otherwise, data will be blocked or lost in the process of transmission, which easily makes the data quality reduced and the performance of the whole target system influenced.
The problem of bandwidth constrained is unavoidable in the research of wireless sensor networks. It is also the key problem which needs to be solved urgently in the fields of networked information fusion. Due to constraints on bandwidth and energy, local sensors usually cannot directly send the original observation to estimation center under the centralized frame, it must be preprocessed firstly. When the
2010 8th IEEE International Conference on Control and Automation Xiamen, China, June 9-11, 2010
ThA5.4
Strong Tracking Filter with Bandwidth Constraint for Sensor Networks
In order to avoid energy and bandwidth constrained, some quantitative estimation algorithms have been given for sensor networks system. Z.Q Luo firstly presented a universal decentralized estimation scheme [6], and then the result has been extended to the self-organizing network environment [7]. However, among these studies, all parameters to be estimated are assumed to be scalars. Thereby, the recursive distributed state estimation based on the sign of innovations is researched for the vector state system in [8], in which the transmission from each sensor to estimation center must be only 1-bit message. Different from [8], each sensor is allowed to send multiple-bit messages to the estimation center in [9], and then two quantized Kalman filters based on quantized observation and quantized innovation are presented. However, this work only considers a class of state-vector and observation-scalar model. In order to overcome the shortcoming of which is only to deal with the scalar measurement system, the design of quantized Kalman filter was solved for the system with state-vector and measurement vector in [10]. It adopts the adaptive quantization strategy for every component of measurement vector according to the bandwidth size and uses upper bound technology to approximate the variance of quantization error. Accordingly, the Kalman filter for uncertain quantized message model is established. Unfortunately, the approximate makes the quantized message model mismatch with the original accurate system. It will greatly influence the estimation performance and should be further improved.