Research On Navigation Algorithm Of Multi-sensor Information Fusion Based On Factor Graph

With the popularization of drones in the military and civilian fields,as well as the diversification and precision of their application,the accuracy of drone navigation systems has become increasingly demanding.In order to improve the navigation accuracy of drones,on the one hand,hardware can be used to improve the measurement accuracy of sensors,which requires further improvement of the production process.Accordinly,the cost of the entire system will increase.On the other hand,it can be derived from the algorithm.Applying modern data fusion algorithms to improve the navigation accuracy of drones does not necessarily increase additional cost.This paper focuses on the multisensor data fusion algorithm,combines with improved factor graph model theory and further improves the navigation accuracy of the UAV integrated navigation system.This paper firstly discusses the theory of the factor graph model,including the analysis of the factor graph model framework.The derivation of the message passing algorithm lays a theoretical foundation for the subsequent application of the factor graph algorithm in the multi-sensor integrated navigation system.Secondly,the SINS/GPS integrated navigation system is analyzed,and a multisensor information fusion error model is established,and then the linear state error equation of the navigation system is deduced.In order to improve the accuracy of navigation filtering,a self-adjusting interval RTS filtering algorithm is proposed.Aiming at the problem of unsynchronized multi-sensor measurement time in integrated navigation system,a multi-sensor information fusion framework of integrated navigation system was constructed based on factor graph theory.The navigation state was updated in real time,and the plug-and-play of sensors was realized.Finally,the feasibility of factor graph filtering is verified by simulation.Aiming at the problem that the conventional factor graph algorithm can't recognize the observable anomaly,this paper proposes a reliability-based robust estimation algorithm,dynamically adjusting the observation covariance matrix,reducing the weight of the anomaly observation,and ensuring the normal operation of the system.Simulations show that the robustness of the factor graph smoothing algorithm based on the robust estimation algorithm is good.