Prediction-based Quantized Kalman Filter For Target Tracking In Wireless Sensor Network

The rapid development of the wireless sensor network containing detecting techniques, pattern recognize, integrated circuit and radio communication is well known now. The wireless sensor network firstly appeared on the battlefield for the target tracking during military operations, which is the one of the primary applications of the wireless sensor network and is now wildly applied in military and civilian fields. However, the wireless sensor network is made up of smart sensor nodes which is a portable embedded device faced with limitations brought by battery power, MCU processing ability, memory space and wireless communication distance. Due to this fact, the application of wireless sensor network for target tracking is met by a challenge from the problem of task completion of target tracking with limited resources.Ad-hoc is a kind of peer-to-peer network structure that indicates the equal status of sensor nodes and needlessness for the prior set up of radio communication infrastructure. In this project, to relieve the burden of network maintenance, all of the sensor nodes of the Ad-Hoc networks are divided into Hierarchical Structures. In addition, for the reduction of computational burden, part of the nodes work together to accomplish the target tracking task. This paper utilizes a scheduling algorithm with relatively low complexity to forbid the waste of resources and shortened life of network potentially caused by the simultaneous measurement of all sensors and the possibility that the limited bandwidth is taken up by the data sent from sensors at the same time. Radio communication is the part consuming highest energy and reducing the cost of communication is one of the important ways to save energy. For the low-cost communication, the measured values are quantized to1bit and the losses of quantized measures are made up by adjusting the Kalman Filter gains. The threshold values of the sent data are set when the estimated values of target location are transmitted to the sink by cluster to forbid the possible increase of communication cost induced by the frequent data transmission. The processing of non-linear characteristics is required for improving accuracy of target tracking and avoids the robust caused by one-order linearization.