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

With the rapid development of multi-source information fusion theory,the performance of long endurance navigation system has been continuously improved,and it can provide real-time navigation for the carrier.The environmental adaptability has also been strengthened.At present,the demand of indoor autonomous navigation based on pedestrian is increasing,and the inertial navigation scheme in indoor autonomous navigation system has been widely used.But the indoor positioning accuracy of pure inertial navigation will become worse over time,so it is difficult to ensure the positioning accuracy of long-time work.All the errors related to time accumulation will eventually affect the real-time performance of the system.Due to the introduction of a variety of navigation sources,the calculation terminal based on wearable pedestrian inertial navigation system has a large workload and many calculation tasks,and it is difficult to achieve real-time.So it is particularly sensitive to the error based on time error.In order to solve the problem of long endurance cumulative error of inertial navigation system,In this thesis,the inertial navigation system is used as the leading navigation source,and using ultra wideband technology and visual sensor as the guidance navigation source.The location information and heading information are modified by factor graph multi-source information fusion algorithm.The positioning accuracy and real-time performance are effectively improved.The main contents of this thesis are as follows:Firstly,the basic principles of inertial navigation system,ultra wideband technology and vision sensor are introduced.The multi-source information fusion system model is constructed for multi-source information fusion technology.The navigation source information is preprocessed and the method of system estimation and correction is designed.Secondly,the basic concept of the factor graph model and its closely related sum product algorithm are studied.And the indirect method is used to deal with the navigation system model,and the system equation and measurement equation are obtained.At the same time,three main models of the factor graph model are given: non recursive model,global model and recursive model,and the corresponding multi-source model is derived by solving the probability graph.In this thesis,an improved method is proposed to reduce the complexity of multi-sensor information fusion system.Finally,the multi-source fusion inertial measurement system is used as the verification platform to calculate the data from asynchronous measurement sampling.The results show that the cumulative error of inertial navigation system in long endurance is smaller than that of single inertial navigation system with the assistance of ultra wideband technology and vision sensor.Compared with the factor graph algorithm based on recursive model,the accuracy of the two axial positioning based on the improved factor graph multi-source information fusion algorithm is improved by16.55% and 17.19% respectively.