The Fusion Estimation Method For Multi-Sensor With Intermittent Measurements

With the further development of the multi-sensor target tracking system,and the widely use of the networked control in general,it leads that the detection probability of sensor is often less than one.Accordingly,the accessibility of observation in different situations is different,which means that the observation set is a random set.Under the condition of the intermittent measurement,the traditional Kalman filter often cannot obtain reliable filtering results.According to the intermittent measurement problem,the uncertainty of the ordinary availability and the uncertainty of generalized availability were defined respectively,the sources of two kinds invalid measurement were analyzed in this thesis.Then the Bernoulli model and the Markov model were established to describe the intermittent measurement problem.To solve the estimation problem with intermittent measurements,an availability function was introduced.Finally,the simulation results analyzed the scene that three obstructions appeared to influence the process of vehicle traffic verified the applicability of Markov model.In order to solve the uncertainty of the ordinary availability problem,this thesis introduced a DIMM algorithm combined with KCF algorithm with a distributed sensor structure,which adapted the multi-sensors system to the intermittent measurement.Two algorithms with different structures were proposed: The first one,each fused estimation was obtained and then the consensus algorithm was used to get final estimations.The second one,the consensus algorithm was used on each model filter,and then fused all models to get final estimations.It proved the equivalence of two algorithms in the scenario of intermittent measurement,and then two kinds algorithm still had a certain similarity in the estimation error covariance,calculation efficiency of two structures was analyzed at the same time.The simulation results also proved the above conclusion.In order to solve the uncertainty of the generalized availability problem,this thesis regarded the unobservable part as the invalid part,then heterogeneous sensors of unequal dimensions problem could be solved by constructing the generalized equivalent sensor,thus we introduced a modified unified method to solve both of heterogeneous sensor and the uncertainty of the availability problems.Firstly,the availability function was constructed to get the generalized observation vector and error variance.Then we use the equivalent sensor method regarding the unobservable components as the invalid components.After that,measurement fusion method is used to get the generalized equivalent pseudo-observation.Finally,the Kalman filter is used to obtain the system state estimation.Finally,simulation results of four different simulation scene showed that the proposed unified method could effectively solve the uncertainty of generalized availability the problem.