Vehicle autonomous navigation method based on cascaded LSTM model and computer equipment

Abstract

The invention provides a cascaded LSTM (Long Short Term Memory) model-based vehicle autonomous navigation method and computer equipment. The cascaded LSTM model-based vehicle autonomous navigation method comprises the following steps: designing a first layer, namely an attitude prediction model, of a cascaded LSTM model; designing a second layer of the cascaded LSTM model, namely a speed prediction model; collecting data of an inertial measurement unit and executing inertial navigation calculation; when satellite navigation is effective, inertia / satellite integrated navigation calculation is executed; training an attitude prediction model based on the inertial navigation result and the combined navigation result; training a speed prediction model based on the inertial navigation result, the combined navigation result and the attitude prediction result; predicting vehicle speed and attitude based on a machine learning model when satellite navigation is invalid; and when satellite navigation is invalid, combined navigation is performed based on a machine learning model prediction result. According to the method, the influence of attitude boundaries on a machine learning model is overcome, and integrated navigation is carried out based on a model prediction result, so that full-autonomous high-precision navigation positioning of the vehicle without depending on external reference information is realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of vehicle navigation, and in particular relates to a vehicle autonomous navigation method based on a cascaded LSTM model. Background Art

[0002] Real-time and accurate navigation and positioning are crucial for vehicle control, path planning, and autonomous driving. Satellite / inertial combined navigation technology is currently the main means of vehicle navigation and positioning. However, satellite navigation is affected by environmental occlusion and other factors, making it difficult to remain effective throughout the entire process, and it cannot be applied to indoor scenarios such as underground parking lots. At this time, the integral solution characteristics of inertial navigation cause its errors to accumulate over time, especially for the micro-small, low-cost inertial navigation system used in vehicle navigation, whose sensor accuracy is limited and the error accumulation characteristics are particularly obvious. Therefore, it is urgent to design a vehicle-mounted autonomous navigation method that does not rely on external reference information such as satellite navigation. Summary of the invention

[0003] The present invention aims to solve at least one of the technical problems existing in the prior art.

[0004] To this end, the present invention provides a vehicle autonomous navigation method based on a cascaded LSTM model.

[0005] The technical solution of the present invention is as follows:

[0006] According to one aspect, a vehicle autonomous navigation method based on a cascaded LSTM model is provided, the autonomous navigation method comprising:

[0007] Step 1: Design the first layer of the cascaded LSTM model to obtain a posture prediction model;

[0008] Step 2: Design the second layer of the cascaded LSTM model to obtain the speed prediction model;

[0009] Step 3: Collect inertial measurement unit data and perform inertial navigation solution;

[0010] Step 4: When GNSS is valid, obtain satellite navigation data and perform integrated navigation solution, including:

[0011] Conduct multiple GNSS-valid sports car tests, and perform inertial / satellite combined navigation based on the inertial measurement unit data and GNSS data collected in step 3 to calculate the reference attitude quaternion Atti predQ and baseline navigation system velocity for training the cascaded LSTM model;

[0012] Step 5: Based on the data collected and calculated in steps 3 and 4, train the posture prediction model according to step 1 to obtain Atti predQ ;

[0013] Step 6: Train the speed prediction model based on the inertial navigation results, combined navigation results and attitude prediction results to obtain a complete cascade LSTM model, including:

[0014] The inertial navigation solution result of step 3 and the attitude quaternion Atti predicted in step 5 are predQ , and Atti predQ Navigation system acceleration calculated from auxiliary accelerometer output As the model input, the navigation system velocity calculated by the inertial / satellite integrated navigation in step 4 is used as the model output, and the LSTM model designed in step 2 is trained to form a complete cascade LSTM model;

[0015] Step 7: Based on the inertial navigation information collected and calculated in step 3, the cascaded LSTM model trained in steps 5 and 6 is used to predict the vehicle's attitude and navigation system velocity results;

[0016] Step 8: Based on the prediction results of the cascaded LSTM model, the inertial measurement unit is assisted to perform integrated navigation and obtain vehicle navigation information.

[0017] Furthermore, the posture prediction model is designed as follows:

[0018] The input of the designed attitude prediction model includes three parts: gyroscope output after deducting harmful components such as the earth's rotation angular velocity Attitude quaternion Atti calculated by pure inertial navigation insQ And the navigation system acceleration f after deducting the gravity acceleration component n ; Among them, the navigation system acceleration f n Attitude calculation through raw accelerometer output combined with pure inertial navigation solution;

[0019] The design attitude prediction model outputs attitude information in the form of quaternion, and the loss function designed for quaternion attitude data is as follows:

[0020]

[0021] Among them, Atti predQ is the quaternion form of the predicted attitude, Atti predQ =[q 0p q 1p q 2p q 3p ] T ;

[0022] θ is the angle difference between the predicted attitude and the reference attitude, and its corresponding quaternion ΔQ=[Δq0Δq1Δq2Δq3] T The calculation formula is:

[0023]

[0024] Among them, Atti referQ =[q 0t q 1t q 2t q 3t ] T is the quaternion corresponding to the reference attitude, where the reference attitude is calculated by inertial / satellite combined navigation.

[0025] Furthermore, the speed prediction model is designed as follows:

[0026] The input of the design speed prediction model includes three parts: gyroscope output after deducting harmful components such as the earth's rotation angular velocity The attitude prediction results (quaternion form) of the first layer of the cascaded LSTM model and the navigation system acceleration after deducting the gravity acceleration component The navigation system acceleration is obtained by combining the original accelerometer output with the attitude Atti predicted by the first layer of the cascaded LSTM model. predQ calculate.

[0027] The model output of the designed speed prediction model is the navigation system speed, and the target benchmark is the navigation system speed calculated by the integrated navigation; the loss function is the root mean square difference (RMSE) between the navigation system speed predicted by the model and the benchmark navigation system speed.

[0028] Furthermore, the step three specifically includes:

[0029] Collect data from the inertial measurement unit, perform inertial navigation solution, and obtain the gyroscope output minus harmful components such as the earth's rotation angular velocity and the attitude quaternion Atti solved by pure inertial navigation insQ ; and based on Atti insQ The navigation system acceleration f is calculated by subtracting the gravity acceleration component from the accelerometer output n , used as input for the cascaded LSTM model.

[0030] Furthermore, the step eight specifically includes:

[0031] Based on the inertial navigation error equation, the attitude and velocity information predicted by the cascaded LSTM model is used as reference information, and LSTM / inertial combined navigation is performed through Kalman filtering.

[0032] According to another aspect, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above method when executing the computer program.

[0033] The above technical solution aims at the problem that the navigation information error of vehicles based on inertial navigation accumulates over time when satellite navigation is invalid. A vehicle autonomous navigation method based on a cascade LSTM model is proposed to achieve fully autonomous high-precision navigation and positioning of vehicles that do not rely on external reference information. When satellite navigation is valid, a cascade LSTM model with two-step prediction of "attitude + speed" is trained. First, the concept of quaternion is introduced into the traditional LSTM model to overcome the boundedness of attitude and achieve high-precision prediction of vehicle attitude. Then, a speed prediction model is designed based on the attitude prediction results to achieve high-precision acquisition of vehicle speed. When satellite navigation is invalid, the trained cascade LSTM model is used to predict vehicle speed and attitude, replacing the reference information to assist inertial navigation for high-precision combined navigation.

[0034] The solution proposed in the present invention does not rely on sensors and reference information other than the inertial measurement unit, can realize fully autonomous navigation of the vehicle, and significantly reduce the error accumulated over time in inertial navigation. It can be applied to satellite navigation failure scenarios that are inevitable during vehicle driving, such as tunnels, indoors, and under building obstructions, and has important practical application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The included drawings are used to provide a further understanding of the embodiments of the present invention, which constitute a part of the specification, are used to illustrate the embodiments of the present invention, and together with the text description, explain the principles of the present invention. Obviously, the drawings in the following description are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0036] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0037] It should be noted that, in the absence of conflict, the embodiments in this application and the features in the embodiments can be combined with each other. The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. The following description of at least one exemplary embodiment is actually only illustrative and is by no means intended to limit the present invention and its application or use. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0038] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0039] Unless otherwise specifically stated, the relative arrangement of the parts and steps described in these embodiments, numerical expressions and numerical values ​​do not limit the scope of the present invention. At the same time, it should be understood that, for ease of description, the sizes of the various parts shown in the accompanying drawings are not drawn according to the actual proportional relationship. The technology, method and equipment known to ordinary technicians in the relevant field may not be discussed in detail, but in appropriate cases, the technology, method and equipment should be regarded as a part of the authorization specification. In all examples shown and discussed here, any specific value should be interpreted as being merely exemplary, rather than as a limitation. Therefore, other examples of exemplary embodiments may have different values. It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once a certain item is defined in an accompanying drawing, it does not need to be further discussed in subsequent drawings.

[0040] refer to Figure 1 This embodiment provides a vehicle autonomous navigation method based on a cascaded LSTM model, which specifically includes the following steps:

[0041] Step 1: Design the first layer of the cascaded LSTM model, namely the posture prediction model

[0042] Among them, the first layer of the cascade LSTM model, namely the posture prediction model, is designed as follows:

[0043] The model input consists of three parts: gyroscope output after deducting harmful components such as the earth's rotation angular velocity Attitude quaternion Atti calculated by pure inertial navigation insQ , the navigation system acceleration f after deducting the gravity acceleration component n ; Among them, the navigation system acceleration f n Attitude calculation through raw accelerometer output combined with pure inertial navigation solution;

[0044] The model outputs attitude information in the form of quaternion, and the loss function designed for quaternion attitude data is as follows:

[0045]

[0046] Among them, Atti predQ is the quaternion form of the predicted attitude, AttipredQ =[q 0p q 1p q 2p q 3p ] T ;

[0047] θ is the angle difference between the predicted attitude and the reference attitude, and its corresponding quaternion ΔQ=[Δq0Δq1Δq2Δq3] T The calculation formula is:

[0048]

[0049] Among them, Atti referQ =[q 0t q 1t q 2t q 3t ] T is the quaternion corresponding to the reference attitude, where the reference attitude is calculated by inertial / satellite combined navigation;

[0050] Step 2: Design the second layer of the cascaded LSTM model, namely the speed prediction model

[0051] Based on the first layer of the cascaded LSTM model, i.e. the posture prediction model, a speed prediction model is designed:

[0052] The model input consists of three parts: gyroscope output after deducting harmful components such as the earth's rotation angular velocity The attitude prediction results (quaternion form) of the first layer of the cascaded LSTM model and the navigation system acceleration after deducting the gravity acceleration component The navigation system acceleration is obtained by combining the original accelerometer output with the attitude Atti predicted by the first layer of the cascaded LSTM model. predQ calculate;

[0053] The model output is the navigation system velocity; the target reference is the navigation system velocity calculated by the integrated navigation, and the loss function is the root mean square difference (RMSE) between the navigation system velocity predicted by the model and the reference navigation system velocity;

[0054] Step 3: Collect IMU data and perform inertial navigation solution

[0055] Collect data from the inertial measurement unit, perform inertial navigation solution, and obtain the gyroscope output minus harmful components such as the earth's rotation angular velocity Attitude quaternion Atti calculated by pure inertial navigation insQ ; and based on Atti insQ The navigation system acceleration f is calculated by subtracting the gravity acceleration component from the accelerometer output n , used as input for the cascaded LSTM model;

[0056] Step 4: When GNSS is valid, obtain satellite navigation data and perform integrated navigation solution

[0057] Conduct multiple GNSS-valid sports car tests, perform inertial / satellite combined navigation based on the collected inertial measurement unit data and GNSS data, and calculate the reference attitude quaternion Atti referQ =[q 0t q 1t q 2t q 3t ] T and baseline navigation system velocity for training the cascaded LSTM model;

[0058] Step 5: Train the attitude prediction model based on the inertial navigation results and combined navigation results

[0059] Based on the data collected and calculated in steps 3 and 4, the posture prediction model in the cascade LSTM model is trained according to step 1 to obtain Atti predQ ;

[0060] Step 6: Train the speed prediction model based on the inertial navigation results, combined navigation results and attitude prediction results to obtain a complete cascade LSTM model;

[0061] Based on the inertial navigation solution in step 3 Step 5 Predicted attitude quaternion Atti predQ , Atti predQ Navigation system acceleration calculated from auxiliary accelerometer output As the model input, the navigation system velocity calculated by the inertial / satellite integrated navigation in step 4 is used as the model output, and the LSTM model designed in step 2 is trained to form a complete cascade LSTM model;

[0062] Step 7: Based on the trained cascade LSTM model, use the inertial navigation data to predict the vehicle attitude and speed information when GNSS is invalid;

[0063] Based on the inertial navigation information collected and calculated in step 3, the cascaded LSTM model trained in steps 5 and 6 is used to predict the vehicle's attitude and navigation system velocity results;

[0064] Step 8: Based on the prediction results of the cascaded LSTM model, the inertial measurement unit is assisted in integrated navigation to obtain high-frequency and high-precision navigation information of the vehicle, including:

[0065] Based on the inertial navigation error equation, the attitude and velocity information predicted by the cascaded LSTM model is used as reference information, and LSTM / inertial combined navigation is performed through Kalman filtering. The Kalman filter model used is as follows

[0066] The state equation is X k =Φ k,k-1 X k-1 +w k-1 .

[0067] In the formula, the state quantity Including the north east misalignment angle 、North celestial east velocity error δv n =[δV n δV u δV e ] T , Latitude Error Altitude error δh, longitude error δλ, carrier system gyro drift ε b =[ε x ε y ε z ] T , load system accelerometer zero bias ▽ b =[▽ x ▽ y ▽ z ] T , Loading system arm error L out , time delay T d ;

[0068] Measurement equation: Z k =H k X k +V k ;

[0069] In the formula, the measurement Z k It is the difference between the attitude and velocity information predicted by the cascade LSTM model and the inertial navigation solution result.

[0070] That is to say, through the combined navigation of cascaded LSTM and inertial navigation, higher-frequency full-dimensional, high-precision vehicle motion information can be obtained without relying on external reference information such as satellite navigation.

[0071] According to another embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above method when executing the computer program.

[0072] It can be seen that the present invention aims at the demand that the inertial navigation error accumulates over time when the reference information such as satellite navigation is invalid, and it is difficult to achieve high-precision navigation of the vehicle. A cascade LSTM model of "attitude + speed" two-step prediction is proposed, and the model prediction results are used to assist the inertial navigation to perform combined navigation, and finally a vehicle autonomous navigation method based on the cascade LSTM model is formed. The scheme proposed by the present invention does not rely on external sensors and reference information, and only uses an inertial measurement unit, which can realize the full autonomous navigation of the vehicle and significantly reduce the error accumulated over time by inertial navigation. It can be applied to satellite navigation failure scenarios such as tunnels, indoors, and building obstructions, and has important practical application value.

[0073] Features described and / or illustrated above for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with or used in place of features in other embodiments.

[0074] It should be emphasized that the term "include / comprises" when used herein refers to the presence of features, integers, steps or components, but does not exclude the presence or addition of one or more other features, integers, steps, components or combinations thereof.

[0075] The above method of the present invention can be implemented by hardware, or by hardware combined with software. The present invention relates to such a computer-readable program, which, when executed by a logic component, enables the logic component to implement the above-mentioned device or component, or enables the logic component to implement the above-mentioned various methods or steps. The present invention also relates to a storage medium for storing the above program, such as a hard disk, a magnetic disk, an optical disk, a DVD, a flash memory, etc.

[0076] The many features and advantages of these embodiments are apparent from this detailed description, and thus the appended claims are intended to cover all such features and advantages of these embodiments that fall within their true spirit and scope. Furthermore, since numerous modifications and changes will readily occur to those skilled in the art, it is not intended that the embodiments of the invention be limited to the exact construction and operation illustrated and described, but rather all suitable modifications and equivalents falling within the scope thereof are intended to be covered.

[0077] Parts of the present invention that are not described in detail are well known to those skilled in the art.

Claims

1. A vehicle autonomous navigation method based on a cascaded LSTM model, characterized in that: The vehicle autonomous navigation method comprises: Step 1: Design the first layer of the cascaded LSTM model to obtain a posture prediction model; Step 2: Design the second layer of the cascaded LSTM model to obtain the speed prediction model; Step 3: Collect inertial measurement unit data and perform inertial navigation solution; Step 4: When GNSS is valid, obtain satellite navigation data and perform integrated navigation solution, including: Conduct multiple GNSS-valid sports car tests, and perform inertial / satellite combined navigation based on the inertial measurement unit data and GNSS data collected in step 3 to calculate the reference attitude quaternion Atti predQ and baseline navigation system velocity for training the cascaded LSTM model; Step 5: Based on the data collected and calculated in steps 3 and 4, train the posture prediction model according to step 1 to obtain Atti predQ ; Step 6: Train the speed prediction model based on the inertial navigation results, combined navigation results and attitude prediction results to obtain a complete cascade LSTM model, including: The inertial navigation solution result of step 3 and the attitude quaternion Atti predicted in step 5 are predQ , and Atti predQ Navigation system acceleration calculated from auxiliary accelerometer output As the model input, the navigation system velocity calculated by the inertial / satellite integrated navigation in step 4 is used as the model output, and the LSTM model designed in step 2 is trained to form a complete cascade LSTM model; Step 7: Based on the inertial navigation information collected and calculated in step 3, the cascaded LSTM model trained in steps 5 and 6 is used to predict the vehicle's attitude and navigation system velocity results; Step 8: Based on the prediction results of the cascaded LSTM model, the inertial measurement unit is assisted to perform integrated navigation and obtain vehicle navigation information.

2. A vehicle autonomous navigation method based on a cascaded LSTM model according to claim 1, characterized in that: The design of the posture prediction model is as follows: The input of the designed attitude prediction model includes three parts: gyroscope output after deducting harmful components such as the earth's rotation angular velocity Attitude quaternion Atti calculated by pure inertial navigation insQ And the navigation system acceleration f after deducting the gravity acceleration component n ; Among them, the navigation system acceleration f n Attitude calculation through raw accelerometer output combined with pure inertial navigation solution; The design attitude prediction model outputs attitude information in the form of quaternion, and the loss function designed for quaternion attitude data is as follows: Among them, Atti predQ is the quaternion form of the predicted attitude, Atti predQ =[q 0p q 1p q 2p q 3p ] T ; θ is the angle difference between the predicted attitude and the reference attitude, and its corresponding quaternion ΔQ=[Δq0Δq1Δq2Δq3] T The calculation formula is: Among them, Atti referQ =[q 0t q 1t q 2t q 3t ] T is the quaternion corresponding to the reference attitude, where the reference attitude is calculated by inertial / satellite combined navigation.

3. A vehicle autonomous navigation method based on a cascaded LSTM model according to claim 1 or 2, characterized in that: The speed prediction model is designed as follows: The input of the design speed prediction model includes three parts: gyroscope output after deducting harmful components such as the earth's rotation angular velocity The attitude prediction results (quaternion form) of the first layer of the cascaded LSTM model and the navigation system acceleration after deducting the gravity acceleration component The navigation system acceleration is obtained by combining the original accelerometer output with the attitude Atti predicted by the first layer of the cascaded LSTM model. predQ calculate. The model output of the designed speed prediction model is the navigation system speed, and the target benchmark is the navigation system speed calculated by the integrated navigation; the loss function is the root mean square difference (RMSE) between the navigation system speed predicted by the model and the benchmark navigation system speed.

4. A vehicle autonomous navigation method based on a cascaded LSTM model according to claim 2 or 3, characterized in that: The step three specifically includes: Collect data from the inertial measurement unit, perform inertial navigation solution, and obtain the gyroscope output minus harmful components such as the earth's rotation angular velocity and the attitude quaternion Atti solved by pure inertial navigation insQ ; and based on Atti insQ The navigation system acceleration f is calculated by subtracting the gravity acceleration component from the accelerometer output n , used as input for the cascaded LSTM model.

5. A vehicle autonomous navigation method based on a cascaded LSTM model according to any one of claims 1 to 4, characterized in that: The step eight specifically includes: Based on the inertial navigation error equation, the attitude and velocity information predicted by the cascaded LSTM model is used as reference information, and LSTM / inertial combined navigation is performed through Kalman filtering.

6. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 5 is implemented.

Images