A GNSS and INS adaptive integrated navigation and positioning method and system based on dual optimization

Abstract

The invention discloses a GNSS and INS adaptive combined navigation positioning method and system based on dual optimization, belonging to the field of combined navigation positioning. The method comprises the following steps: combining the vehicle lateral speed output by the BP neural network model with the forward speed and elevation speed constraints of the odometer, and processing through CKF1 to obtain a predicted update value, and using it as the input of the RBF neural network model; at the same time, inputting the speed and position information obtained by the GNSS and INS into CKF2, performing measurement update, obtaining an actual update value, using the actual update value as the output of the RBF neural network model, and training the RBF neural network model; when the GNSS loses lock, inputting the current moment predicted update value into the RBF neural network model to obtain a navigation positioning result. The invention combines the dual optimization method with the motion constraint, which not only improves the observability of the position but also makes the dual optimization meet the real-time performance.

Description

Technical Field

[0001] The invention relates to the field of integrated navigation and positioning, and in particular to a GNSS and INS adaptive integrated navigation and positioning method and system based on dual optimization. Background Art

[0002] The global navigation satellite system (GNSS) and the micro-electromechanical system-based inertial navigation system (MEMS-INS) can be integrated to form a combined navigation system to provide accurate navigation and positioning solutions when GNSS signals are available. The advantages of this combined navigation system include high-precision positioning, accurate velocity prediction, and suppression of inertial navigation divergence. Obviously, the performance of the MEMS-INS / GNSS combined navigation system is much better than that of the GNSS or MEMS-INS standalone system. However, in urban environments, tunnels, and areas of harsh electromagnetic interference where GNSS signal reception is blocked or even non-existent, the positioning accuracy is severely degraded. During the GNSS signal interruption, the integrated navigation system enters the pure INS mode and there is no observation update, which reduces the positioning accuracy.

[0003] The Chinese invention patent with application publication number CN116224407A discloses a GNSS and INS combined navigation positioning method and system, which combines motion constraints and neural network algorithms to improve positioning accuracy and reliability compared to existing combined navigation systems. However, the method adopted in the Kalman filtering process only uses velocity observations for Kalman filtering measurement updates, and there is still an unobservable problem in the position latitude, and the positioning results still need to be further improved. Summary of the invention

[0004] In view of the above technical problems, the present invention proposes a GNSS and INS adaptive integrated navigation and positioning method and system based on dual optimization.

[0005] The technical solution adopted by the present invention is:

[0006] A GNSS and INS adaptive integrated navigation and positioning method based on dual optimization includes the following steps:

[0007] Step 1: construct a BP neural network model, use the vehicle forward speed and heading angular velocity as input, and the vehicle lateral speed as output, train the BP neural network model, and obtain a trained BP neural network model;

[0008] Step 2: Build an RBF neural network model, combine the vehicle lateral speed output by the BP neural network model with the forward speed and elevation speed constraints of the odometer to form a speed observation vector in the three-dimensional direction; input this speed observation vector into the first volumetric Kalman filter, perform measurement update, obtain a predicted update value, and use this predicted update value as the input of the RBF neural network model;

[0009] At the same time, the speed and position information obtained by GNSS and INS are input into the second volumetric Kalman filter to perform measurement updates to obtain actual update values, which are used as the output of the RBF neural network model to train the RBF neural network model.

[0010] Step 3: When GNSS works well, the speed and position information obtained by GNSS and INS are input into the second volumetric Kalman filter, measurement update is performed, and the combined navigation and positioning results of GNSS and INS are output;

[0011] Step 4: When GNSS is locked, the lateral speed of the vehicle at the current moment is predicted by the trained BP neural network model, and then combined with the forward speed of the odometer at the current moment and the elevation speed constraint to form the speed observation vector in the three-dimensional direction at the current moment; this speed observation vector is input into the first volumetric Kalman filter to obtain the predicted update value at the current moment, and the predicted update value at the current moment is input into the RBF neural network model to obtain the navigation positioning result.

[0012] The present invention also provides a GNSS and INS adaptive integrated navigation and positioning system based on dual optimization, comprising a GNSS antenna, a GNSS data processing module, an IMU sensor, a vehicle odometer, a central processing unit and a PC control terminal;

[0013] Among them, the GNSS antenna is connected to the GNSS data processing module, and the IMU sensor is connected to the central processing unit;

[0014] The GNSS data processing module, the central processor and the on-board odometer are respectively connected to the PC control terminal;

[0015] The PC control terminal includes a memory and a processor; the memory stores executable code;

[0016] When the processor executes the executable code, it is used to implement the steps of the GNSS and INS adaptive integrated navigation and positioning method based on dual optimization as described above.

[0017] The beneficial technical effects of the present invention are as follows:

[0018] (1) The present invention combines the dual optimization neural network algorithm with motion constraints for adaptive combined navigation and positioning, thereby improving positioning accuracy.

[0019] (2) The present invention combines the idea of ​​federated filtering to design two filter loops CKF1 (first volumetric Kalman filter) and CKF2 (second volumetric Kalman filter). Compared with the classic dual optimization algorithm, it is not necessary to introduce another set of reference values ​​of the reference system when training the RBF neural network model.

[0020] (3) The training of the entire window period of the present invention takes place in CKF1, which cleverly combines the dual optimization method with the motion constraint, which not only improves the observability of the position but also makes the dual optimization meet the real-time requirements.

[0021] (4) When GNSS works well, CKF2 works normally to output GNSS / INS combined navigation results and trains the RBF neural network model in real time. When GNSS loses lock, the information output by CKF1 at the corresponding moment is input into the RBF neural network model for prediction, and the predicted speed and position information are obtained, and the positioning prediction result is output. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a method flow chart of the GNSS and INS adaptive integrated navigation and positioning method based on dual optimization of the present invention when the GNSS is not lost;

[0023] Figure 2 A method flow chart of the GNSS and INS adaptive integrated navigation and positioning method based on dual optimization when the GNSS loses lock;

[0024] Figure 3 It is a comparison chart of the solution result trajectory of the GNSS and INS adaptive combined navigation and positioning method based on dual optimization of the present invention, the classic OD / NHC navigation and positioning method, the single BP neural network method and the normal solution result trajectory without loss of lock;

[0025] Figure 4 It is a speed error comparison diagram of the GNSS and INS adaptive combined navigation and positioning method based on dual optimization of the present invention, the classic OD / NHC navigation and positioning method, and the single BP neural network method, wherein (a) is a speed error comparison diagram in the E direction, and (b) is a speed error comparison diagram in the N direction;

[0026] Figure 5 This is a position error comparison diagram of the GNSS and INS adaptive combined navigation and positioning method based on dual optimization of the present invention, the classic OD / NHC navigation and positioning method, and the single BP neural network method, where (a) is a position error comparison diagram in the E direction, and (b) is a position error comparison diagram in the N direction. DETAILED DESCRIPTION

[0027] In conjunction with the accompanying drawings, a GNSS and INS adaptive integrated navigation and positioning method based on dual optimization includes the following steps:

[0028] Step 1, construct a BP neural network model; use the vehicle forward velocity information within a window period output by the GNSS and INS combined navigation system during normal combined solution and the heading angular velocity information measured by the IMU as the input of the BP neural network, use the vehicle lateral velocity information within a window period as the output of the BP neural network, train the constructed BP neural network, and obtain a trained BP neural network model.

[0029] The input layer is the vehicle forward velocity and heading angular velocity calculated by the integrated navigation system within a window period, and the output layer is the vehicle lateral velocity.

[0030] The training of the window period is divided into two parts, the first half and the second half. The first half is used for the training of the BP neural network model, and the second half is used for the training of the RBF neural network model in the following steps.

[0031] The input and output of the BP neural network model are as follows:

[0032] (1)

[0033] (2)

[0034] In the formula, The input of the BP neural network model contains the vehicle forward speed information in the first half of the window period. And heading angular velocity information , is the output of the BP neural network model, which contains the vehicle lateral speed information in the first half of the window period ; ; n is the first half of the window period.

[0035] The number of hidden layers of the BP neural network model is 2, the maximum number of training iterations is 300, the initial learning rate is 0.01, and the training method uses the gradient descent method.

[0036] Step 2: Construct an RBF neural network model, and combine the vehicle lateral speed output by the BP neural network model with the forward speed and elevation speed constraints of the odometer to form a speed observation vector in the three-dimensional direction. Input this speed observation vector into the first volumetric Kalman filter, perform measurement updates, and obtain a predicted update value; use this predicted update value as the input of the RBF neural network model. At the same time, input the speed and position information obtained by GNSS and INS into the second volumetric Kalman filter, perform measurement updates, and obtain an actual update value, which is used as the output of the RBF neural network model to train the RBF neural network model.

[0037] The input and output of the RBF neural network model are as follows:

[0038] (3)

[0039] (4)

[0040] is the input of the RBF neural network model, is the output of the RBF neural network model.

[0041] and They are respectively the speed and position of the predicted update value output by the first volumetric Kalman filter at the corresponding moment; that is, the forward speed and heading angular velocity of the vehicle at the corresponding moment are input into the trained BP neural network model, and the lateral speed at the corresponding moment is predicted, and it is combined with the forward speed and elevation speed constraints of the odometer at the corresponding moment to form a speed observation vector in the three-dimensional direction, which is brought into the first volumetric Kalman filter to obtain the predicted update value at the corresponding moment.

[0042] and They are respectively the speed and position updated by the normal measurement obtained by the second volumetric Kalman filter at the corresponding moment.

[0043] ; n+10 is the entire window period.

[0044] The RBF neural network model was trained using the newrb function in the MATLAB neural network toolbox. The training parameters of the RBF neural network model were: maximum number of neurons 1000, error tolerance 0.01, radial basis expansion speed 0.5, and increment factor 1.

[0045] Step 3: When GNSS works well, the speed and position information obtained by GNSS and INS are input into the second volumetric Kalman filter, measurement update is performed, and the combined navigation and positioning results of GNSS and INS are output.

[0046] Step 4: When GNSS is unlocked, the lateral speed of the vehicle at the current moment is predicted by the trained BP neural network model, and then combined with the forward speed of the odometer at the current moment and the elevation speed constraint to form the speed observation vector in the three-dimensional direction at the current moment. This speed observation vector is input into the first volumetric Kalman filter to obtain the current moment prediction update value, and the prediction update value is input into the real-time trained RBF neural network model to obtain the navigation positioning result.

[0047] Specifically, in step 4:

[0048] When GNSS loses lock, the vehicle forward speed calculated by INS and the vehicle heading angular velocity measured by IMU are input into the BP neural network model, and the pseudo-measured value of the vehicle's lateral speed at the moment of GNSS loss of lock is directly obtained through the predicted BP neural network model. , the three-dimensional velocity observation vector at this time is:

[0049] (5)

[0050] is the vehicle lateral speed output by the BP neural network model, is the forward speed output by the on-board odometer, It is the elevation speed constraint.

[0051] The effect of the GNSS and INS adaptive integrated navigation and positioning method based on dual optimization of the present invention is experimentally verified as follows:

[0052] The experimental data uses a set of real vehicle data sets collected in an open industrial zone in a certain city. In order to fully verify the effectiveness of the method of the present invention, the GNSS data is randomly interrupted during the vehicle driving process. The number of interruptions is 17 times, and the duration of each interruption is 15s to 20s. The inertial sensor is ICM20602 (TDK, Japan).

[0053] During the experiment, the duration of artificial interruption was mainly distributed in 25-30s. The GNSS interruption process included common vehicle motion states such as straight driving, turning, large arc turning, U-turn, acceleration, deceleration, etc. The maximum speed of the vehicle was about 10m / s and the average speed was about 6.4m / s. The main performance indicators of inertial devices are shown in Table 1 below.

[0054] Table 1

[0055]

[0056] In order to evaluate the performance of the adaptive integrated navigation and positioning method of the present application, the following three schemes are used to solve the experimental data: the classic NHC / OD algorithm, the INS / OD algorithm of single BP neural network assisted NHC, and the dual optimization algorithm proposed in the present application.

[0057] Solution 1: Adopt the NHC algorithm, add lateral and astronomical speed constraints, introduce the OD odometer, add forward speed observation and combine it with INS.

[0058] Solution 2: Based on Solution 1, a BP neural network is constructed to assist in obtaining the lateral velocity. The obtained lateral velocity is used to replace the constant lateral velocity in Solution 2, and then combined with INS for solution.

[0059] Solution 3: The dual optimization algorithm based on motion constraints proposed in this application brings the result value solved by Solution 2 into the trained RBF neural network to obtain the secondary optimization result value.

[0060] The post-processed GNSS-RTK solution results are used as the reference true values ​​of position and velocity. The position error and velocity error of the solution results of each algorithm in the experiment are statistically analyzed, and the anti-error effects of the three schemes are evaluated.

[0061] Figure 3 The trajectory diagrams are obtained by using three algorithms and reference values. The speed and position errors of the experiment are shown in Figure 4 As shown in the figure, it can be seen that compared with the NHC / OD method, the positioning accuracy has been significantly improved after the introduction of the BP neural network. As shown in Table 2, compared with Scheme 1, the positioning accuracy of Scheme 2 is improved from meter level to sub-meter level in terms of speed and position. The further introduction of the dual optimization algorithm (i.e., Scheme 3) not only improves the positioning accuracy, but also enhances the reliability of the system. Compared with Scheme 2, Scheme 3 improves the standard deviation (STD) of position and velocity errors by 60% and 80% respectively; and in terms of the root mean square error (RMSE) of position and velocity errors, Scheme 3 improves by 21.8% and 34.7% respectively.

[0062] Table 2

[0063]

[0064] The present invention also provides a GNSS and INS adaptive combined navigation and positioning system based on dual optimization, comprising a GNSS antenna, a GNSS data processing module, an IMU sensor, a vehicle odometer, a central processing unit and a PC control terminal. The GNSS antenna is connected to the GNSS data processing module, and the IMU sensor is connected to the central processing unit. The GNSS data processing module and the central processing unit are respectively connected to the PC control terminal. The PC control terminal comprises a memory and a processor; an executable code is stored in the memory. When the processor executes the executable code, it is used to implement the steps of the GNSS and INS adaptive combined navigation and positioning method based on dual optimization as described above.

[0065] The parts not mentioned in the above methods can be realized by adopting or drawing on existing technologies.

[0066] The above-described embodiments are merely descriptions of preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should also fall within the scope of protection determined by the claims of the present invention.

Claims

1. A GNSS and INS adaptive integrated navigation and positioning method based on dual optimization, characterized in that The following steps are involved: Step 1: construct a BP neural network model, use the vehicle forward speed and heading angular velocity as input, and the vehicle lateral speed as output, train the BP neural network model, and obtain a trained BP neural network model; Step 2: Build an RBF neural network model, combine the vehicle lateral speed output by the BP neural network model with the forward speed and elevation speed constraints of the odometer to form a speed observation vector in the three-dimensional direction; input this speed observation vector into the first volumetric Kalman filter, perform measurement update, obtain a predicted update value, and use this predicted update value as the input of the RBF neural network model; At the same time, the speed and position information obtained by GNSS and INS are input into the second volumetric Kalman filter to perform measurement updates to obtain actual update values, which are used as the output of the RBF neural network model to train the RBF neural network model. Step 3: When GNSS works well, the speed and position information obtained by GNSS and INS are input into the second volumetric Kalman filter, measurement update is performed, and the combined navigation and positioning results of GNSS and INS are output; Step 4: When GNSS is locked, the lateral speed of the vehicle at the current moment is predicted by the trained BP neural network model, and then combined with the forward speed of the odometer at the current moment and the elevation speed constraint to form the speed observation vector in the three-dimensional direction at the current moment; this speed observation vector is input into the first volumetric Kalman filter to obtain the predicted update value at the current moment, and the predicted update value at the current moment is input into the RBF neural network model to obtain the navigation positioning result.

2. The GNSS and INS adaptive integrated navigation and positioning method based on dual optimization according to claim 1 is characterized in that: In step one: The training of the window period is divided into two parts, the first half and the second half. The first half is used for the training of the BP neural network model, and the second half is used for the training of the RBF neural network model. The input and output of the BP neural network model are as follows: （1） （2） In the formula, The input of the BP neural network model contains the vehicle forward speed information in the first half of the window period. And heading angular velocity information , is the output of the BP neural network model, which contains the vehicle lateral speed information in the first half of the window period ; ; n is the first half of the window period; The number of hidden layers of the BP neural network model is 2, the maximum number of training iterations is 300, the initial learning rate is 0.01, and the training method uses the gradient descent method.

3. The GNSS and INS adaptive integrated navigation and positioning method based on dual optimization according to claim 1 is characterized in that: In step 2: The input and output of the RBF neural network model are as follows: （3） （4） is the input of the RBF neural network model, is the output of the RBF neural network model; and are respectively the speed and position of the predicted update value output by the first volumetric Kalman filter at the corresponding moment; and are respectively the speed and position of the normal measurement update obtained by the second volumetric Kalman filter at the corresponding moment; ; n+10 is the entire window period.

4. The GNSS and INS adaptive integrated navigation and positioning method based on dual optimization according to claim 3 is characterized in that: The RBF neural network model was trained using the newrb function in the MATLAB neural network toolbox. The training parameters of the RBF neural network model were: maximum number of neurons 1000, error tolerance 0.01, radial basis expansion speed 0.5, and increment factor 1.

5. The GNSS and INS adaptive integrated navigation and positioning method based on dual optimization according to claim 1 is characterized in that: In step 4: When GNSS loses lock, the vehicle forward speed calculated by INS and the vehicle heading angular velocity measured by IMU are input into the BP neural network model, and the pseudo-measured value of the vehicle lateral speed at the moment of GNSS loss of lock is directly obtained through the trained BP neural network model. , the three-dimensional velocity observation vector at this time is: （5） is the vehicle lateral speed output by the BP neural network model, is the forward speed output by the on-board odometer, It is the elevation speed constraint.

6. A GNSS and INS adaptive integrated navigation and positioning system based on dual optimization, including a GNSS antenna, a GNSS data processing module, an IMU sensor, a vehicle odometer, a central processing unit and a PC control terminal; in, The GNSS antenna is connected to the GNSS data processing module, and the IMU sensor is connected to the central processing unit; The GNSS data processing module, the central processor and the on-board odometer are respectively connected to the PC control terminal; The PC control terminal includes a memory and a processor; the memory stores executable code; It is characterized in that when the processor executes the executable code, it is used to implement the steps of the GNSS and INS adaptive combined navigation and positioning method based on dual optimization as described in any one of claims 1 to 5 above.

Images