Research On Multi-sensor Data Fusion Estimation Algorithm In Noise Correlation Environment

With the development of science and technology and the rapid rise of artificial intelligence,data fusion technology plays an increasingly important role in the field of intelligent robots and attracts many scholars to study it.Data fusion technology mainly uses multiple data resources to effectively evaluate the state,and the state estimation is based on filtering algorithm.Although data fusion technology can effectively estimate the state of the system,there are still many problems in the process of data processing.In the fusion estimation algorithm,the correlation between noises,the nonlinear problem of the system,and the excessive number of sensors and the large amount of calculation need to be considered.In view of the above problems,linear filtering algorithm and nonlinear filtering algorithm are designed respectively to complete the state estimation and trajectory tracking of the target.The main research contents are as follows:Firstly,the fusion estimation of one-step correlation between process noise and measurement noise is studied.Aiming at the problems of large computation and small flexibility of centralized filtering algorithm,a sequential filtering algorithm with global optimality is designed,and its global optimality is proved in detail.Furthermore,under the condition of satisfying the experimental requirements,a global suboptimal distributed matrix weighted filtering algorithm is proposed,which uses the covariance between local states to determine the weight to obtain the final state estimation.The algorithm has high flexibility,and the loss of measurement data of a sensor does not affect the estimation accuracy.Secondly,the problem of state estimation for nonlinear systems with non-Gaussian interference and noise correlation is studied.By using the Bayesian filtering framework,the interference of non-Gaussian terms is introduced into the likelihood probability,that is,the interference of not only Gaussian noise but also non-Gaussian term is introduced in the sensor measurement process,so the untracked information filtering under the condition of non-Gaussian term interference is proposed.Aiming at the one-step correlation of measurement noise of process noise,the filtering of untraced information under noise correlation is proposed.Furthermore,considering the two cases of non-Gaussian term and noise correlation,the filtering of untracked information under the condition of non-Gaussian interference and noise correlation is proposed.For these three algorithms,the stability of the algorithm is proved to ensure the validity of the algorithm.Finally,a distributed cluster fusion estimation problem is studied.Considering that there are a large number of sensors and the noise interferes with each other,the interference of the noise measured by sensors is divided into a group,and the local state estimation is obtained by centralized information filtering in the cluster.Among clusters,the weights are determined according to the covariance of local state estimates,and the final state estimates are obtained by matrix weighting.The distributed cluster fusion estimation has a two-layer fusion structure,which ensures the accuracy of state estimation and the flexibility of the algorithm.