Research On Maneuvering Target Tracking Algorithm In Sensor Networks

Multi-source information fusion is a technology to combine data and information from multiple sources in order to obtain accurate description and estimate of targets.By introducing resource management in multi-source information fusion system,the multi-level closed-loop structure of the system is built to improve system performance.According to the maneuvering characteristics of the stealth target,etc.,a variable structure multiple model algorithm which updates model sets online for single sensor is designed in this paper.On this basis,a collaborative tracking algorithm based on Rényi information gain for multi-radar network is proposed.Two kinds of distributed target tracking algorithms are simulated in sensor network.The main work can be summarized as follows:Firstly,the research background and significance of this thesis is briefly described.The current progress of maneuvering target collaborative tracking algorithms in sensor networks and distributed target tracking algorithms are reviewed.The theories and methods of maneuvering target tracking are also introduced.Secondly,the maneuvering target collaborative tracking algorithms based on hybrid grid and Rényi information gain is designed.To solve the problem caused by the inaccuracy estimation of target mode and a sudden maneuver of acceleration in the variable structure multiple model algorithm,a hybrid grid multiple model algorithm based on current statistics model is adopted and a fading factor is introduced to adjust the upper and lower limits of target acceleration.An improved tracking algorithm of maneuvering targets based on CSHGMM is presented.Simulation results demonstrate the effectiveness of the algorithm.To solve the nonlinear observation problem when tracking one maneuvering target in multi-radar network,a maneuvering target collaborative tracking algorithm based on Rényi information gain is designed.Firstly,a variable structure multiple model algorithm which combines with current statistics model and interacting multiple model unscented kalman filter is proposed to estimate the state of maneuvering target.One sensor is then selected to perform target tracking according to maximal Rényi information gain.Grid partition is finally performed by estimation of the optimal acceleration to update possible model sets of the target.The performance of the proposed algorithm is analyzed in general and strong maneuvering scenarios,and the simulation results demonstrate that the proposed algorithm can select the optimal sensor reasonably and improves the accuracy for maneuvering target tracking.The proposed algorithms are also transplanted and tested in the multi-radar network simulation software platform based on HLA.Thirdly,two kinds of distributed target tracking algorithms are simulated,which involve Kalman consensus filter and distributed particle filter.Simulation results show that the algorithms can realize the consistency of target tracking.Finally,the main work and further research of this thesis are summarized.