Research On The Key Techniques Of Tightly Coupled GNSS/INS Integration

The tightly coupled Global Navigation Satellite System(GNSS)and Inertial Navigation System(INS)integration can maintain effective output of position and velocity when the GNSS signal is blocked,and can prevent the accumulation of INS errors.It makes up for the shortcomings of both parties.However,with the rapid development of large-scale intelligent systems that are sensitive to positioning performance,such as autonomous vehicles and unmanned transportation equipment,there is a higher demand for combined systems.It can be summarized in three directions:high precision,high stability and high real-time.In this paper,the requirements of the above three directions in tightly coupled GNSS/INS integration are studied and discussed.The research work and innovations are as follows:1、An adaptive singular value decomposition algorithm is proposed to solve the problem that the Kalman filter algorithm based on singular value decomposition is strongly dependent on the initial prior information and easy to diverge in complex environments.Since innovation represents the difference between the current estimated value and the observed value,it also implies the state of the current filter.Using the innovation to obtain the correction,and doing singular value decomposition.Correcting the noise covariance real-timely,can be adaptive to get a better noise estimation.Secondly,the information filtering algorithm based on singular value decomposition is deduced to reduce the rounding error.Comparative experiments show that the proposed filtering method and the information filtering based on singular value decomposition have higher positioning accuracy under the 32-bit variable’s storage width.2、A factor graph algorithm based on adaptive covariance is proposed.The weight coefficient in the pseudorange measurement covariance matrix can be obtained by using the pseudo-code phase locking parameters,carrier phase locking parameters and frame calibration parameters in GNSS receiver,which can maintain the high positioning accuracy and system stability of the integrated navigation system when GNSS signal refuses or the signal is unlocked.The simulation and experimental results show that,compared with other improved factor graph algorithms,the standard deviation of position is reduced by 40.6% in 15 seconds of interference time,which is more suitable for the scene with complex environment and changeable interference in practical engineering.3、Aiming at the problem that the number of nodes increases linearly with time in factor graph algorithm,a low storage factor graph algorithm is proposed.The generalized likelihood ratio detection statistics is used to dectect the zero-velocity of vehicle.it is composed of the three-dimensional acceleration and angular velocity output by the inertial measurement unit and the pseudorange-rate measurement output by the GNSS receiver.When the carrier is detected to be stationary,the current state node is not added to the existing graph,and the zero-velocity correction variable node is added to the factor graph to constrain the system state value,and then solve it.The vehicle experiment shows that the number of main nodes decreases from 1800 to 1185,reducing by 34.1% within 600 seconds.Under the premise of keeping the positioning accuracy unchanged,this algorithm effectively reduces the number of nodes,and the storage and real-time performance is guaranteed.It is suitable for the computing platform of lightweight storage and computing.In summary,this paper focuses on the algorithm research of high precision,high stability and high real-time,studies the robust Kalman filter and graph optimization method based on tightly coupled GNSS/INS integration.And the paper obtains a series of useful conclusions,which strongly supports engineering practice.