Research On Multi-source Information Fusion Algorithm Based On Factor Graph

Because of the continuous development of multi-source information fusion theory,the future multi-source Navigation system mainly based on redundant measurement will become the mainstream with the continuous construction and improvement of Position Navigation and Timing(PNT)services.In the future,multi-source navigation system can provide continuous navigation information for the carrier and has a strong ability to adapt to the environment.Among them,due to the different working conditions of sensors,the measurement information available to the carrier in different environments or in different time periods of the same environment keeps changing.Multi-source information fusion technology is the key technology of dynamic combination dosability measurement.It is of great significance to study how to change the structure of filtering algorithm to improve the dynamic adaptability,real-time performance,reliability and accuracy of navigation system.Aiming at the problem that the existing information fusion methods are not flexible enough and have low precision when dealing with the dynamic integration of sensors,this paper uses the factor graph probability model to model the integrated navigation system,and proposes a plug and play filtering algorithm suitable for sensors around multi-source information fusion.The main contents of this paper are as follows:Firstly,the basic principles of INS,GPS,OD and EC are introduced.The mechanical programming equation of strapdown inertial navigation is derived,and the combination,estimation and correction methods of multi-source navigation system are briefly introduced.Secondly,the spatiotemporal registration technology of multi-source information is introduced.The arm errors of different sensors are analyzed and the compensation method is given.The errors in time registration due to incomplete synchronization of measurement information are analyzed.INS motion information is used to interpolate other sensors to reduce the influence of time errors.Aiming at the phenomenon of GPS measurement delay,a measurement lag compensation method based on multiple sensors is proposed to improve the system accuracy without abandoning the lag measurement.Then,the basic concept of factor graph and the summation algorithm of conditional probability density function are introduced.The factor graph model of navigation system under non-recursive model,global model and recursive model is given.An improved factor graph model is proposed for plug and play problems of sensors.The model is solved by recursive structure method.The method not only supports dynamic changes of sensors but also processes asynchronous measurement information more easily and reduces system complexity.Finally,car experiment was carried out,using the measurement lag compensation method was verified on the actual data and the factor graph information fusion method,the simulation results show that under the condition of GPS measurement lag,slightly followed by measurement accuracy of OD output and the change of the INS quantity to compensate the hysteresis of GPS information,the effect is better than that of INS alone compensation with abandon measurement method;In the case of dynamic change of measurable information,the information fusion method based on plug and play factor graph can effectively filter the variable measured information and ensure the system accuracy under the premise of flexible combination of sensors.