Information Fusion Algorithms For Multisensor System With Spatiotemporal Bias

In the multisensor information fusion system,the uncertainty of data sources and the possibly existing system bias are the two key issues that may affect the performance of information fusion.To solve the uncertainty of data sources,universal and efficient multisensor data association method is need to be studied.And in the research of the registration of unknow system bias,most algorithms study the problems of spatial bias registration and time alignment,temporal bias registration as well as spatiotemporal bias simultaneous registration haven’t been studied.While in the practice the timestamp of measurement has the time delay,temporal bias thus exists between sensors.In order to guarantee the performance of fusion,the spatial and temporal bias are then needed to be registrated.In this paper,the following three aspects about data association and bias estimation problems in the multisensor information fusion system are mainly studied.First,the data association method based on multidimensional assignment(MDA)algorithm.Aiming at multiple passive sensor multitarget tracking problem,the data association method based on multidimensional assignment algorithm is studied.In this algorithm,the prior information are used to establish the validation gate,only the measurements that fall into the validation gate are used to generate the association hypotheses and are assigned cost values.Based on the relationship that each measurement is assigned at most one target or declared false and each target is assigned at most one measurement from each sensor,the constraint conditions are obtained,thus the generalized MDA problem are formulated.The Lagrangian Relaxation method are used to successively relax the contranits to formulate the two dimensional assignment problem.The obtained two dimensional assignment problem can be solved in polynomial time complexity and get the assignment results according to that problem.Successively enforce the constraints based on the assignment results to eventually obtained the measurement association results from multiple sensors.The simulation results show that the proposed algorithm can reduce the time complexity without loss of data association accuracy,and the proposed data association can apply to other types of multisensor multitarget tracking system.Secondly,the multisensor system spatiotemporal bias registration method with known data rate.In the case of two sensors,the relation between sensor measurement and target state as well as spatiotemporal bias is analyzed,and the sensor spatiotemporal bias and target state eatimation joint estimation model is formulated.The situation where sensors have equal and unequal data rate is studied,and two different algorithms of the sensor spatiotemporal bias and target state eatimation joint estimation based on block processing and sequential processing,respectively,are proposed.The simulation results shows the effectiveness of the proposed algorithms,and the spatiotemporal bias registration and the target state fusion estimation can be achieved simultaneously.Among which sequential processing scenario has better estimation performace.Besides,noticing the algorithm has a slow convergence speed when estimate the spatiotemporal bias,the spatiotemporal bias estimations from multiple targets at the same time instant are weighted fusion in the principle of minimum root mean square error.The simulation results shows multiple targets weighted fusion can obtain more accurate spatiotemporal bias estimation and faster convergence speed.Finally,the multisensor system spatiotemporal bias registration method with known data rate.In the case of two sensors,in the situation where only the timestamps with constant time delay are known,the sensor spatiotemporal bias and target state eatimation joint estimation with unknown data rate model is formulated,the sensor spatiotemporal bias and target state eatimation joint estimation based on sequential processing algorithm is proposed.The simulation results shows the proposed algorithm can obtain the spatiotemporal bias registration and the target state fusion estimation can be achieved simultaneously.And in the principle of minimum root mean square error,weighted fusion is carried on the spatiotemporal bias estimations from multiple targets at the same time instant to improve the convergence speed.The simulation results shows multiple targets weighted fusion can remarkably improve the estimation accuracy and convergence speed of spatiotemporal bias.