Attitude prediction method and device based on artificial intelligence

Abstract

The invention relates to the technical field of computer vision, and discloses an attitude prediction method and device based on artificial intelligence, and the method comprises the steps: obtaining first inertial data in a first state and second inertial data in a second state; determining first attitude information based on the first inertial data and the second inertial data; and adjusting the first attitude information to obtain target attitude information. According to the invention, attitude adjustment can be carried out on the first attitude information according to the inertial data in the large-maneuver motion state, the problem that attitude estimation is abnormal in the large-maneuver motion state is solved, and the robustness of attitude estimation in the large-maneuver motion state is improved.

Description

Technical Field

[0001] This application relates to the field of computer vision technology, and particularly to an attitude prediction method and device based on artificial intelligence. Background Art

[0002] Attitude Estimation has been a hot technology in recent years and is widely used in related fields such as robots, AR / VR, and drones. An IMU (Inertial Measurement Unit) is a sensor that does not rely on external information and usually has a gyroscope and an accelerometer inside. The attitude accuracy of the gyroscope integrated in a short time is relatively high, but serious drift will occur in a long time; the attitude calculation accuracy of the accelerometer is poor, but there will be no serious drift in a long time. In the prior art, there are mainly two ways of attitude calculation: Kalman filtering and complementary filtering. When the object is in a slow motion state, the accuracy of attitude prediction of these two attitude prediction methods can be acceptable, but when the object is in a large maneuver motion state, the estimated attitude will show obvious abnormalities, with low robustness and cannot be used normally. Summary of the Invention

[0003] This application aims to solve at least one of the technical problems existing in the related art. To this end, an embodiment of this application provides an attitude prediction method and device based on artificial intelligence, which can improve the robustness of the estimated attitude in the second state.

[0004] In a first aspect, an embodiment of this application provides an attitude prediction method based on artificial intelligence, including:

[0005] Obtaining first inertial data in a first state and second inertial data in a second state;

[0006] Determining first attitude information based on the first inertial data and the second inertial data;

[0007] Adjusting the first attitude information to obtain target attitude information.

[0008] In a second aspect, an embodiment of this application provides an attitude prediction device based on artificial intelligence, including:

[0009] An obtaining module, configured to obtain first inertial data in a first state and second inertial data in a second state;

[0010] An attitude determination module, configured to determine first attitude information based on the first inertial data and the second inertial data;

[0011] An attitude adjustment module, configured to adjust the first attitude information to obtain target attitude information.

[0012] In a third aspect, an embodiment of the present application further provides an electronic device, including a memory storing multiple computer programs; a processor loads the computer programs from the memory to execute any one of the artificial intelligence-based pose prediction methods provided by the embodiments of the present application.

[0013] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium storing multiple computer programs, and the computer programs are suitable for being loaded by a processor to execute any one of the artificial intelligence-based pose prediction methods provided by the embodiments of the present application.

[0014] In a fifth aspect, an embodiment of the present application further provides a computer program product including a computer program, and when the computer program is executed by a processor, it implements any one of the artificial intelligence-based pose prediction methods provided by the embodiments of the present application.

[0015] The embodiment of the present application can adjust the first pose information according to the inertial data in the large maneuver motion state, solves the problem of abnormal estimated pose in the large maneuver motion state, and improves the robustness of the estimated pose in the large maneuver motion state. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] To more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present application. For those skilled in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0017] Figure 1 is a flowchart of the artificial intelligence-based pose prediction method provided by the embodiment of the present application;

[0018] Figure 2 is a schematic flowchart of the overall solution provided by the embodiment of the present application;

[0019] Figure 3 is a schematic structural diagram of the artificial intelligence-based pose prediction device provided by the embodiment of the present application;

[0020] Figure 4 is a schematic structural diagram of the electronic device provided by the embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application. At the same time, in the description of the embodiments of the present application, terms such as "first" and "second" are only used for distinguishing descriptions and cannot be understood as indicating or implying relative importance. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more features. In the description of the embodiments of the present application, the meaning of "a plurality" is two or more, unless otherwise clearly and specifically defined.

[0022] The embodiments of the present application provide a posture prediction method and device based on artificial intelligence. Specifically, the embodiments of the present application will be described from the perspective of a posture prediction device based on artificial intelligence. The posture prediction device based on artificial intelligence may include intelligent vehicles, intelligent robots, Augmented Reality (AR) devices, Virtual Reality (VR) devices, unmanned aerial vehicles and other intelligent operation devices.

[0023] It should be noted that the description order of the following embodiments does not limit the preferred order of the embodiments. Although the logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order from that shown in the drawings.

[0024] The embodiments of the present application are described with a posture prediction device based on artificial intelligence as the execution subject, and an intelligent vehicle is used as the posture prediction device based on artificial intelligence for illustration. The following will be described in detail with reference to the accompanying drawings. Refer to Figure 1 , Figure 1 is a schematic flowchart of the posture prediction method based on artificial intelligence provided in the embodiments of the present application. The specific process steps of the posture prediction method based on artificial intelligence provided in the embodiments of the present application may be as follows: steps 10 to 30, including:

[0025] Step 10, obtain the first inertial data in the first state and the second inertial data in the second state.

[0026] It should be noted that the intelligent vehicle in the embodiments of the present application includes at least two data sensors. The data sensors are such as Inertial Measurement Units (IMUs). The inertial measurement unit includes a gyroscope and an accelerometer. Therefore, the intelligent vehicle can obtain the inertial data of the intelligent vehicle during driving through the IMU inertial measurement unit. The inertial data includes gyroscope data and acceleration data.

[0027] Optionally, the intelligent vehicle obtains the inertial data (gyroscope data and acceleration data) of the intelligent vehicle in the first state (current state). Here, the current state can be a uniform (static) state, a slow movement state, or a high-maneuver movement state. The slow movement state means that the driving speed of the intelligent vehicle is less than or equal to the set speed and the acceleration change is less than or equal to the set threshold. The high-maneuver movement state means that the driving speed of the intelligent vehicle is greater than the set speed and the acceleration change is greater than the set threshold.

[0028] Optionally, it can also be understood that the intelligent vehicle obtains the inertial data of the intelligent vehicle at the current time.

[0029] Optionally, the intelligent vehicle obtains the inertial data of the intelligent vehicle in the second state (the next state of the current state). Here, the next state can be a static state, a slow movement state, or a high-maneuver movement state. It should be noted that the inertial data of the next state can be understood as the inertial data between the current state and its next state, that is, the inertial data includes multiple gyroscope data and acceleration data. Optionally, it can also be understood that the intelligent vehicle obtains the inertial data between the current time and the next time of the intelligent vehicle.

[0030] It should be noted that when the next state is a high-maneuver movement state, the artificial intelligence-based attitude prediction method of the embodiment of the present application will be executed. Therefore, before determining whether to execute the artificial intelligence-based attitude prediction method of the embodiment of the application, it is necessary to determine whether the next state is a high-maneuver movement state. The specific steps include step a to step b:

[0031] Step a, calculate the mean value of the measurement data and the variance value of the measurement data based on the second inertial data;

[0032] Step b, if the mean value of the measurement data is greater than or equal to the mean threshold, and the variance value of the measurement data is greater than or equal to the variance threshold, then determine that the second state is a high-maneuver movement state.

[0033] Optionally, the intelligent vehicle calculates the mean value of the measurement data and the variance value of the measurement data according to the acceleration data of the next state. Therefore, it can be understood that the mean value of the measurement data can be understood as the mean acceleration, and the variance value of the measurement data can be understood as the acceleration variance. Specifically, it can be obtained according to the mean value calculation formula and the variance calculation formula, which will not be elaborated here.

[0034] Further, the intelligent vehicle compares the average acceleration value with the average threshold value to obtain a comparison result. The comparison result can be that the average acceleration value is greater than or equal to the average threshold value, or the average acceleration value is less than the average threshold value. If the comparison result is that the average acceleration value is less than the average threshold value, it is determined that the intelligent vehicle is in a uniform (stationary) state. It should be noted that if the intelligent vehicle is in a uniform (stationary) state, the subsequent steps will not be executed. If the comparison result is that the average acceleration value is greater than or equal to the average threshold value, it is determined that the intelligent vehicle is in a slow movement state or a large maneuver movement state, and then the acceleration variance value needs to be compared with the variance threshold value.

[0035] Further, if it is determined that the acceleration variance value is less than the variance threshold value, the intelligent vehicle determines that it is in a slow movement state. If it is determined that the acceleration variance value is greater than or equal to the variance threshold value, the intelligent vehicle determines that it is in a large maneuver movement state. Thus, the embodiment of the present application jointly determines whether the next state of the intelligent vehicle is a large maneuver movement state through the average acceleration value and the acceleration variance value, ensuring the accuracy of the movement state of the intelligent vehicle.

[0036] Step 20: Based on the first inertial data and the second inertial data, determine the first attitude information.

[0037] Optionally, the intelligent vehicle calculates the initial attitude information in the current state according to the acceleration data in the current state. Further, the intelligent vehicle performs attitude recursive prediction on the initial attitude information according to the gyroscope data in the next state and the gyroscope data in the current state to obtain the first attitude information in the next state, as specifically described in Steps 201 to 203.

[0038] Step 30: Adjust the first attitude information to obtain the target attitude information.

[0039] Optionally, the intelligent vehicle determines the attitude compensation amount of the intelligent vehicle in the next state according to the acceleration data and gyroscope data in the next state. Optionally, the intelligent vehicle superimposes the attitude compensation amount of the intelligent vehicle in the next state on the first attitude information of the intelligent vehicle in the next state, adjusts the first attitude information of the intelligent vehicle in the next state, and outputs the target attitude information of the intelligent vehicle in the next state, as specifically described in Steps 301 to 302.

[0040] The embodiment of the present application can determine the attitude compensation amount in the large maneuver movement state according to the inertial data in the large maneuver movement state, and thus perform attitude adjustment on the first attitude information through the attitude compensation amount, solving the problem of abnormal estimated attitude in the large maneuver movement state and improving the robustness of the estimated attitude in the large maneuver movement state.

[0041] In an optional embodiment, the steps 201 to 203 are described as follows:

[0042] Step 201: Perform high-pass filtering and low-pass filtering on the first inertial data to obtain the target inertial data of the first state;

[0043] Step 202: Normalize the target inertial data to obtain the second attitude information of the first state;

[0044] Step 203: Perform attitude recursive prediction on the second attitude information based on the first inertial data and the second inertial data to obtain the first attitude information.

[0045] Optionally, the gyroscope data and acceleration data collected by the IMU inertial measurement unit contain various noises. Therefore, in order to improve the accuracy of the first attitude information, it is necessary to remove the noises from the gyroscope data and acceleration data in the current state. Specifically:

[0046] Perform high-pass filtering and low-pass filtering on the first inertial data, that is, remove the gyroscope data with low-frequency noise in the inertial data in the current state through high-pass filtering, and remove the acceleration data with high-frequency noise in the inertial data in the current state through low-pass filtering, to obtain the target inertial data of the intelligent vehicle in the current state.

[0047] Further, the intelligent vehicle normalizes the target inertial data in the current state to obtain the initial attitude information (second attitude information) of the intelligent vehicle in the current state. The process of normalization is as follows: Assume that the gravity is expressed as (0, 0, 9.8), and normalize the acceleration data measured by the accelerometer to obtain the normalized acceleration data as (a x , a y , a z ), where a x is the acceleration value in the x-axis direction, a y is the acceleration value in the y-axis direction, and a z is the acceleration value in the z-axis direction. Further, according to the normalized acceleration data (a x , a y , a z ), obtain the initial attitude information q acc of the intelligent vehicle in the current state. The expression formula of the initial attitude information q acc of the intelligent vehicle in the current state is:

[0048] For the case where a z ≥ 0, the initial attitude information q cc of the intelligent vehicle in the current state is:

[0049]

[0050] For a z In the case of <0, the initial attitude information q of the intelligent vehicle in the current state cc is:

[0051]

[0052] Furthermore, the intelligent vehicle also performs noise removal processing on the gyroscope data and acceleration data of the next state. The specific process is the same as the principle of the noise removal processing of the gyroscope data and acceleration data in the current state, and will not be elaborated here.

[0053] Furthermore, the intelligent vehicle performs attitude recursive prediction on the initial attitude information according to the gyroscope data between the current state and the next state, and obtains the first attitude information of the intelligent vehicle in the next state.

[0054] In one embodiment, it can also be understood that the intelligent vehicle performs attitude recursive prediction on the attitude information at time tk-1 according to the attitude information at time tk-1 and the gyroscope data between time tk-1 and time tk, and obtains the first attitude information of the intelligent vehicle at time tk. The specific representation formula is as follows:

[0055]

[0056] where q acc,tk is the first attitude information at time tk-1, q acc,tk-1 is the initial attitude information at time tk-1, is the attitude change rate from time tk-1 to time tk, which can be obtained from the gyroscope data, and Δt is the time difference between time tk-1 and time tk.

[0057] The embodiment of the present application accurately infers the first attitude information of the object in the second state according to the first inertial data and the second inertial data, providing a data basis for the follow-up.

[0058] In an alternative embodiment, steps 301 to 302 are described

[0059] Step 301, based on the second inertial data, determine the attitude compensation amount of the second state;

[0060] Step 302, based on the attitude compensation amount, adjust the first attitude information to obtain the target attitude information.

[0061] Optionally, the intelligent vehicle calculates the gravity deviation value and the direction deviation value of the intelligent vehicle in the next state based on the acceleration data and the gyroscope data in the next state, where the gravity deviation value is the modulus difference between the modulus of the acceleration data of the intelligent vehicle in the next state and the modulus of the gravity data of the intelligent vehicle in the next state. Further, the intelligent vehicle determines the attitude compensation amount of the intelligent vehicle in the next state according to the gravity deviation value and the direction deviation value in the next state, specifically as steps 301 to 303.

[0062] Optionally, the intelligent vehicle superimposes the attitude compensation amount of the intelligent vehicle in the next state onto the first attitude information of the intelligent vehicle in the next state, adjusts the first attitude information of the intelligent vehicle in the next state, and outputs the target attitude information of the intelligent vehicle in the next state.

[0063] The embodiment of the present application can determine the attitude compensation amount of an object in a large maneuvering motion state according to the inertial data of the object in the large maneuvering motion state, so as to perform attitude adjustment on the first attitude information of the object through the attitude compensation amount, solve the problem of abnormal attitude estimation in the large maneuvering motion state, and improve the robustness of attitude estimation of the object in the large maneuvering motion state.

[0064] In an alternative embodiment, the descriptions of steps 3011 to 3013 are as follows:

[0065] Step 3011: Determine the gravity deviation value of the second state based on the gravity data of the second state and the acceleration data in the second inertial data;

[0066] Step 3012: Determine the direction deviation value of the second state based on the gyroscope data and the acceleration data in the second inertial data;

[0067] Step 3013: Determine the attitude compensation amount of the second state based on the gravity deviation value and the direction deviation value.

[0068] Optionally, the intelligent vehicle separates gravity from the acceleration data of the intelligent vehicle in the next state, obtains the component of the acceleration data in the vertical direction, and calculates the magnitude of the component of the acceleration data in the vertical direction to obtain the magnitude |G1| of the component of the acceleration data in the vertical direction. Further, the intelligent vehicle calculates the magnitude of the gravity data of the intelligent vehicle in the next state to obtain the gravity magnitude |G2| of the intelligent vehicle in the next state. Further, the intelligent vehicle calculates the magnitude difference between the magnitude |G1| of the component of the acceleration data in the vertical direction and the gravity magnitude |G2| of the intelligent vehicle in the next state to obtain the magnitude difference, and this magnitude difference is the gravity deviation value of the intelligent vehicle in the next state, that is, the gravity deviation value = |G1| - |G2|, where the gravity deviation value characterizes the difference between the component value of the acceleration data output by the sensor in the vertical direction and the true gravity value.

[0069] It should be noted that in the slow motion state, the smaller the gravity deviation value, the higher the confidence level of the attitude estimated by the acceleration measured by the accelerometer, and the compensation coefficient should be adjusted larger; otherwise, the compensation coefficient should be adjusted smaller. Therefore, by dynamically adjusting the compensation coefficient, the correctness of compensating with the accelerometer measurement value is ensured.

[0070] Further, to determine the direction deviation value of the intelligent vehicle in the next state, gyroscope data and acceleration data are required to calculate the attitude information of the intelligent vehicle in the next state, and then the intelligent vehicle determines the direction deviation value of the intelligent vehicle in the next state according to the attitude information of the intelligent vehicle in the next state. The direction deviation value characterizes the degree of direction deviation of the intelligent vehicle in the next state. Therefore, it can be understood that the intelligent vehicle needs to obtain the acceleration data and gyroscope data of the intelligent vehicle in the next state, and then determine the direction deviation value of the intelligent vehicle in the next state according to the acceleration data and gyroscope data of the intelligent vehicle in the next state, that is, according to the acceleration data and gyroscope data of the intelligent vehicle in the next state.

[0071] In one embodiment, the attitude solution algorithm of the intelligent vehicle in the next state is as follows:

[0072] S1: Obtain gyroscope data and acceleration data.

[0073] S2: Process the acceleration data to remove the gravitational acceleration to obtain the component of the acceleration in the body coordinate system.

[0074] S3: Use the acceleration vector a (a x , a y , a z ) measured by the accelerometer to calculate the z-axis direction vector a' in the current body coordinate system, that is, unitize the measured acceleration vector.

[0075]

[0076] Among them, ||[a x , a y , a z || represents the modulus length of [a x , a y , a z .

[0077] S4: Calculate the angular velocity vector ω = [w x , w y , w z using the gyroscope data, that is, the rotational angular velocity at the current moment.

[0078] S5: Calculate the rotation angle θ = [θ x , θ y , θ z of the current body coordinate system relative to the inertial coordinate system according to the Euler angle sequence (such as YXZ).

[0079] S6: Convert the rotation angle to a rotation matrix R, that is, the rotation matrix R = R z (θ z ) * R y (θ y ) * R x (θ x ), where R z (θ z ) represents the rotation matrix corresponding to rotating by the angle θ z around the z-axis, R y (θ y ) represents the rotation matrix corresponding to rotating by the angle θ y around the y-axis, and R x (θ x ) represents the rotation matrix corresponding to rotating by the angle θ x around the x-axis. R = R z (θ z ) * R y (θ y ) * R x (θ x ) can be understood as: first rotate around the x-axis, then rotate around the y-axis, and finally rotate around the z-axis to obtain the final rotation matrix R.

[0080] S7: Convert the acceleration vector a' to a vector g = R * a' in the inertial coordinate system, that is, find the component of the gravity direction at the current moment in the inertial coordinate system.

[0081] S8: Calculate the angle δ between the vector g = and the z-axis direction vector, that is, the deviation angle δ.

[0082]

[0083] It should be noted that in the state of large maneuvering motion, there is a scenario where the magnitude of the accelerometer measurement value is not much different from the magnitude of gravity, but the direction difference is relatively large. In this scenario, only through the magnitude deviation value, the attitude confidence estimated from the accelerometer measurement value cannot be truly reflected. If the compensation coefficient is still calculated only based on the magnitude deviation value, it will lead to incorrect attitude estimation. Therefore, the direction deviation value is introduced. In the state of large maneuvering motion, the smaller the direction deviation value, the higher the attitude confidence estimated using the acceleration measurement value, and the compensation coefficient is increased; conversely, the compensation coefficient is decreased.

[0084] Furthermore, the intelligent vehicle determines the attitude compensation amount of the intelligent vehicle in the next state according to the gravity deviation value and the direction deviation value of the intelligent vehicle in the next state, specifically as steps 30131 to 30136.

[0085] In the embodiment of the present application, the attitude compensation amount of the object in the second state is jointly determined by the gravity deviation value and the direction deviation value, which ensures the accuracy of the inertial data of the object in the state of large maneuvering motion, enables the attitude compensation amount of the object in the state of large maneuvering motion to be determined according to the inertial data of the object in the state of large maneuvering motion, and thus adjusts the first attitude information of the object through the attitude compensation amount, solving the problem of abnormal attitude estimation in the state of large maneuvering motion and improving the robustness of attitude estimation of the object in the state of large maneuvering motion.

[0086] In an alternative embodiment, the descriptions of steps 30131 to 30136 are as follows:

[0087] Step 30131, if the gravity deviation value is less than or equal to the first error threshold, determine whether the direction deviation value is greater than the second error threshold;

[0088] Step 30132, if the direction deviation value is greater than the second error threshold, determine the attitude compensation amount of the second state based on the direction deviation value; or,

[0089] Step 30133, if the direction deviation value is less than or equal to the second error threshold, determine the attitude compensation amount of the second state based on the gravity deviation value;

[0090] Step 30134, if the gravity deviation value is greater than the first error threshold, determine the non-gravitational acceleration compensation value based on the second inertial data;

[0091] Step 30135, if the direction deviation value is greater than the second error threshold, determine the attitude compensation amount of the second state based on the direction deviation value and the non-gravitational acceleration compensation value;

[0092] Step 30136, if the direction deviation value is less than or equal to the second error threshold, determine the attitude compensation amount of the second state based on the non-gravitational acceleration compensation value.

[0093] Optionally, the intelligent vehicle compares the gravity deviation value with the first error threshold to obtain a comparison result. Therefore, the comparison result can be that the gravity deviation value is less than or equal to the first error threshold, or the gravity deviation value is greater than the first error threshold, where the first error threshold is set according to the actual situation.

[0094] Therefore, for the case where the comparison result can be that the gravity deviation value is less than or equal to the first error threshold:

[0095] If it is determined that the gravity deviation value is less than or equal to the first error threshold, at this time, the influence of non-gravitational acceleration does not need to be considered. Therefore, the intelligent vehicle compares the direction deviation value with the second error threshold to determine whether the direction deviation value is greater than the second error threshold, and obtains a comparison result, where the second error threshold is set according to the actual situation.

[0096] If it is determined that the comparison result is that the direction deviation value is greater than the second error threshold, determining the attitude compensation amount through the gravity deviation value at this time will cause an error in attitude estimation. Therefore, it is necessary to determine the attitude compensation amount through the direction deviation value. Therefore, the intelligent vehicle obtains a direction deviation value - attitude compensation amount table, where the direction deviation value - attitude compensation amount table is a pre-established association relationship table between the direction deviation value and its corresponding attitude compensation amount, as shown in Table 1 in an embodiment.

[0097] Table 1 Direction deviation value - attitude compensation amount table

[0098] Direction deviation value Attitude compensation amount Direction deviation value Attitude compensation amount L1 M1 L2 M2 L3 M3 L4 M4

[0099] In an embodiment, L1 < L2 < L3 < L4, M1 > M2 > M3 > M4.

[0100] Furthermore, the intelligent vehicle performs data matching in the direction deviation value - attitude compensation amount table according to the direction deviation value to obtain the attitude compensation amount of the intelligent vehicle in the next state.

[0101] Furthermore, if it is determined that the direction deviation value is less than or equal to the second error threshold, at this time, the influence of the direction deviation value on determining the attitude compensation amount is small, and the attitude compensation amount of the intelligent vehicle in the next state can be determined only through the gravity deviation value. Therefore, the intelligent vehicle obtains a gravity deviation value - attitude compensation amount table, where the gravity deviation value - attitude compensation amount table is a pre-established association relationship table between the gravity deviation value and its corresponding attitude compensation amount, as shown in Table 2 in an embodiment.

[0102] Table 2 Gravity deviation value - attitude compensation amount table

[0103] Gravity deviation value Attitude compensation amount Gravity deviation value Attitude compensation amount H1 M1 H2 M2 H3 M3 H4 M4

[0104] In one embodiment, H1 < H2 < H3 < H4, and M1 > M2 > M3 > M4.

[0105] Furthermore, the intelligent vehicle performs data matching in the gravity deviation value - attitude compensation scale according to the gravity deviation value to obtain the attitude compensation amount of the intelligent vehicle in the next state.

[0106] Therefore, for the case where the comparison result may be that the gravity deviation value is greater than the first error threshold:

[0107] Optionally, if the gravity deviation value is greater than the first error threshold, the influence of non - gravitational acceleration needs to be considered at this time. Therefore, the intelligent vehicle estimates the non - gravitational acceleration compensation value according to the inertial data of the intelligent vehicle in the next state, and determines the non - gravitational acceleration compensation value based on the second inertial data, including steps d to f:

[0108] Step d: Based on the dynamic behavior mathematical model and the initial attitude information of the first state, determine the initial state estimate value and covariance information;

[0109] Step e: Update the Kalman gain of the initial state estimate value and the covariance information based on the second inertial data to obtain the final state estimate value;

[0110] Step f: Determine the final state estimate value as the non - gravitational acceleration compensation value.

[0111] Optionally, the intelligent vehicle determines the initial state estimate value and covariance information according to the dynamic behavior mathematical model of the intelligent vehicle and the initial attitude information of the intelligent vehicle in the current state. Among them, the dynamic behavior mathematical model describes the state change of the vehicle during movement. The dynamic behavior mathematical model may include state variables such as the position, speed, and acceleration of the vehicle, and combines physical laws. The observable quantity in the dynamic behavior mathematical model is the non - gravitational acceleration compensation value.

[0112] Furthermore, the intelligent vehicle updates the Kalman gain of the initial state estimate value and the covariance information through the acceleration data of the intelligent vehicle in the next state to obtain the final state estimate value, and determines the final state estimate value as the non - gravitational acceleration compensation value.

[0113] Furthermore, the intelligent vehicle compares the direction deviation value with the second error threshold to determine whether the direction deviation value is greater than the second error threshold, and obtains the comparison result.

[0114] Optionally, if the comparison result is determined to be that the direction deviation value is greater than the second error threshold, then the influence of the direction deviation value and the non-gravity acceleration compensation needs to be considered at this time, and the smart car obtains the direction deviation value-attitude compensation scale, and performs data matching in the direction deviation value-attitude compensation scale according to the direction deviation value to obtain the attitude compensation amount of the smart car in the next state. Further, the smart car determines the non-gravity acceleration compensation and the attitude compensation amount of the smart car in the next state as the final attitude compensation amount of the smart car in the next state, that is, the final attitude compensation amount of the smart car in the next state = non-gravity acceleration compensation + attitude compensation amount of the smart car in the next state.

[0115] Optionally, if the comparison result is determined to be that the direction deviation value is less than or equal to the second error threshold, then only the influence of non-gravitational acceleration needs to be considered at this time. Therefore, the smart car determines the non-gravitational acceleration compensation as the posture compensation amount of the smart car in the next state.

[0116] The embodiment of the present application jointly determines the posture compensation amount of the object in the second state through the gravity deviation value and the direction deviation value, thereby ensuring the accuracy of the inertial data of the object in the large maneuvering motion state, so that the posture compensation amount of the object in the large maneuvering motion state is determined according to the inertial data of the object in the large maneuvering motion state, and the posture adjustment of the first posture information of the object is performed through the posture compensation amount, which solves the problem of abnormal estimation of posture in the large maneuvering motion state and improves the robustness of the estimated posture of the object in the large maneuvering motion state.

[0117] Therefore, refer to Figure 2 , Figure 2 This is a schematic diagram of the overall solution flow provided by the embodiment of the present application, which can be specifically understood as follows:

[0118] 1. Data preprocessing: Filter the gyroscope data and acceleration data collected in the current state, as well as the gyroscope data and acceleration data collected in the next state. Specifically, remove the low-frequency noise of the gyroscope data through high-pass filtering, and remove the high-frequency noise of the acceleration data through low-pass filtering to obtain the target inertial data.

[0119] 2. Posture initialization: The initial posture is calculated through acceleration data to obtain the initial posture information.

[0120] 3. Attitude recursive prediction: Perform attitude recursive prediction on the initial attitude information through gyroscope data to obtain the first attitude information.

[0121] 4. Attitude compensation calculation: The attitude compensation amount is calculated based on the error and dynamic compensation coefficient between the attitude information predicted by the gyroscope data and the attitude information estimated by the accelerometer.

[0122] 5. Attitude Output: Superimpose the calculated attitude compensation amount on the first attitude information and output the target attitude information.

[0123] Next, the attitude prediction device based on artificial intelligence provided in the embodiments of the present application will be described. The attitude prediction device based on artificial intelligence described below can be correspondingly referred to the attitude prediction method based on artificial intelligence described above. Refer to Figure 3 as shown in Figure 3 FIG. 7 is a schematic structural diagram of an attitude prediction device based on artificial intelligence provided in an embodiment of the present application. The attitude prediction device based on artificial intelligence may include:

[0124] An acquisition module 301, configured to acquire first inertial data of a first state and second inertial data of a second state;

[0125] An attitude determination module 302, configured to determine first attitude information based on the first inertial data and the second inertial data;

[0126] An attitude adjustment module 303, configured to adjust the first attitude information to obtain target attitude information.

[0127] The embodiments of the present application can determine the attitude compensation amount in the large maneuver motion state according to the inertial data in the large maneuver motion state, so as to perform attitude adjustment on the first attitude information through the attitude compensation amount, solve the problem of abnormal estimated attitude in the large maneuver motion state, and improve the robustness of the estimated attitude in the large maneuver motion state.

[0128] In an optional example, the attitude adjustment module 303 is further configured to:

[0129] Determine the attitude compensation amount of the second state based on the second inertial data;

[0130] Adjust the first attitude information based on the attitude compensation amount to obtain target attitude information.

[0131] In an optional example, the attitude adjustment module 303 is further configured to:

[0132] Determine the gravity deviation value of the second state based on the gravity data of the second state and the acceleration data in the second inertial data;

[0133] Determine the direction deviation value of the second state based on the gyroscope data and the acceleration data in the second inertial data;

[0134] Determine the attitude compensation amount of the second state based on the gravity deviation value and the direction deviation value.

[0135] In an optional example, the attitude adjustment module 303 is further configured to:

[0136] If the gravity deviation value is less than or equal to the first error threshold, determine whether the direction deviation value is greater than the second error threshold;

[0137] If the direction deviation value is greater than the second error threshold, determine the attitude compensation amount of the second state based on the direction deviation value; or,

[0138] If the direction deviation value is less than or equal to the second error threshold, determine the attitude compensation amount of the second state based on the gravity deviation value.

[0139] In an optional example, the attitude adjustment module 303 is further configured to:

[0140] If the gravity deviation value is greater than the first error threshold, determine the non-gravity acceleration compensation value based on the second inertial data;

[0141] If the direction deviation value is greater than the second error threshold, determine the attitude compensation amount of the second state based on the direction deviation value and the non-gravity acceleration compensation value; or

[0142] If the direction deviation value is less than or equal to the second error threshold, determine the attitude compensation amount of the second state based on the non-gravity acceleration compensation value.

[0143] In an optional example, the attitude adjustment module 303 is further configured to:

[0144] Determine the initial state estimate value and covariance information based on the dynamic behavior mathematical model and the initial attitude information of the first state;

[0145] Perform Kalman gain update on the initial state estimate value and the covariance information based on the second inertial data to obtain the final state estimate value;

[0146] Determine the final state estimate value as the non-gravity acceleration compensation value.

[0147] In an optional example, the attitude determination module 302 is further configured to:

[0148] Perform high-pass filtering and low-pass filtering on the first inertial data to obtain the target inertial data of the first state;

[0149] Perform normalization processing on the target inertial data to obtain the second attitude information of the first state;

[0150] Perform attitude recursive prediction on the second attitude information based on the first inertial data and the second inertial data to obtain the first attitude information.

[0151] The specific embodiments of the posture prediction device based on artificial intelligence provided in this application are basically the same as those of the posture prediction method based on artificial intelligence in each embodiment, and will not be elaborated here.

[0152] Optionally, as Figure 4 shown, Figure 4 is a schematic structural diagram of an electronic device provided in an embodiment of the present application. The electronic device may include: a processor 410, a communication interface 420, a memory 430, and a communication bus 440. Among them, the processor 410, the communication interface 420, and the memory 430 communicate with each other through the communication bus 440. The processor 410 may call a computer program in the memory 430 to execute the steps of the posture prediction method based on artificial intelligence, for example, including:

[0153] Obtain first inertial data in a first state and second inertial data in a second state;

[0154] Based on the first inertial data and the second inertial data, determine first posture information;

[0155] Adjust the first posture information to obtain target posture information.

[0156] In addition, when the logical computer program in the above-mentioned memory 430 can be implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several computer programs to enable a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present application. The foregoing storage medium includes: various media such as a USB flash drive, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk, or an optical disc that can store program codes.

[0157] On the other hand, an embodiment of the present application also provides a non-transitory computer-readable storage medium. The non-transitory computer-readable storage medium includes a computer program. The computer program can be stored on the non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the steps of the posture prediction method based on artificial intelligence provided in the above-mentioned various embodiments, for example, including:

[0158] Obtain the first inertial data in the first state and the second inertial data in the second state;

[0159] Determine the first attitude information based on the first inertial data and the second inertial data;

[0160] Adjust the first attitude information to obtain the target attitude information.

[0161] On the other hand, an embodiment of the present application also provides a computer product. The computer product includes a computer program that can be stored on the computer product. When the computer program is executed by a processor, the computer can execute the steps of the attitude prediction method based on artificial intelligence provided in the above embodiments, for example, including:

[0162] Obtain the first inertial data in the first state and the second inertial data in the second state;

[0163] Determine the first attitude information based on the first inertial data and the second inertial data;

[0164] Adjust the first attitude information to obtain the target attitude information.

[0165] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative labor.

[0166] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution, in essence, or the part that contributes to the prior art can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several computer programs to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0167] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than limiting them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An attitude prediction method based on artificial intelligence, characterized in that, it includes: Obtain the first inertial data of the first state and the second inertial data of the second state; Based on the first inertial data and the second inertial data, determine the first attitude information; Adjust the first attitude information to obtain the target attitude information.

2. The attitude prediction method based on artificial intelligence according to claim 1, characterized in that, The adjusting the first attitude information to obtain the target attitude information includes: Based on the second inertial data, determine the attitude compensation amount of the second state; Based on the attitude compensation amount, adjust the first attitude information to obtain the target attitude information.

3. The attitude prediction method based on artificial intelligence according to claim 2, characterized in that, The determining the attitude compensation amount of the second state based on the second inertial data includes: Based on the gravity data of the second state and the acceleration data in the second inertial data, determine the gravity deviation value of the second state; Based on the gyroscope data and acceleration data in the second inertial data, determine the direction deviation value of the second state; Based on the gravity deviation value and the direction deviation value, determine the attitude compensation amount of the second state.

4. The attitude prediction method based on artificial intelligence according to claim 3, characterized in that, The determining the attitude compensation amount of the second state based on the gravity deviation value and the direction deviation value includes: If the gravity deviation value is less than or equal to the first error threshold, determine whether the direction deviation value is greater than the second error threshold; If the direction deviation value is greater than the second error threshold, determine the attitude compensation amount of the second state based on the direction deviation value; or, If the direction deviation value is less than or equal to the second error threshold, determine the attitude compensation amount of the second state based on the gravity deviation value.

5. The attitude prediction method based on artificial intelligence according to claim 3, characterized in that, The determining the attitude compensation amount of the second state based on the gravity deviation value and the direction deviation value includes: If the gravity deviation value is greater than the first error threshold, determine the non-gravitational acceleration compensation value based on the second inertial data; If the direction deviation value is greater than the second error threshold, determine the attitude compensation amount of the second state based on the direction deviation value and the non-gravitational acceleration compensation value; or If the direction deviation value is less than or equal to the second error threshold, determine the attitude compensation amount of the second state based on the non-gravitational acceleration compensation value.

6. The attitude prediction method based on artificial intelligence according to claim 5, characterized in that, The determining the non-gravitational acceleration compensation value based on the second inertial data includes: Based on the dynamic behavior mathematical model and the initial attitude information of the first state, determine the initial state estimation value and covariance information; Based on the second inertial data, perform Kalman gain update on the initial state estimation value and the covariance information to obtain the final state estimation value; Determine the final state estimation value as the non-gravitational acceleration compensation value.

7. The artificial intelligence-based attitude prediction method according to any one of claims 1 to 6, characterized in that determining the first attitude information based on the first inertial data and the second inertial data includes: Performing high-pass filtering and low-pass filtering on the first inertial data to obtain the target inertial data of the first state; Performing normalization processing on the target inertial data to obtain the second attitude information of the first state; Performing attitude recursive prediction on the second attitude information based on the first inertial data and the second inertial data to obtain the first attitude information.

8. An artificial intelligence-based attitude prediction device, characterized in that comprising: An acquisition module for acquiring the first inertial data of the first state and the second inertial data of the second state; An attitude determination module for determining the first attitude information based on the first inertial data and the second inertial data; An attitude adjustment module for adjusting the first attitude information to obtain the target attitude information; Preferably, the attitude adjustment module adjusts the first attitude information to obtain the target attitude information, including: Determining the attitude compensation amount of the second state based on the second inertial data; Adjusting the first attitude information based on the attitude compensation amount to obtain the target attitude information; Preferably, the attitude adjustment module determines the attitude compensation amount of the second state based on the second inertial data, including: Determining the gravity deviation value of the second state based on the gravity data of the second state and the acceleration data in the second inertial data; Determining the direction deviation value of the second state based on the gyroscope data and the acceleration data in the second inertial data; Determining the attitude compensation amount of the second state based on the gravity deviation value and the direction deviation value; Preferably, the attitude adjustment module determines the attitude compensation amount of the second state based on the gravity deviation value and the direction deviation value, including: If the gravity deviation value is less than or equal to the first error threshold, determine whether the direction deviation value is greater than the second error threshold; If the direction deviation value is greater than the second error threshold, determine the attitude compensation amount of the second state based on the direction deviation value; or, If the direction deviation value is less than or equal to the second error threshold, determine the attitude compensation amount of the second state based on the gravity deviation value; Preferably, the attitude adjustment module determines the attitude compensation amount of the second state based on the gravity deviation value and the direction deviation value, including: If the gravity deviation value is greater than the first error threshold, determine the non-gravitational acceleration compensation value based on the second inertial data; If the direction deviation value is greater than the second error threshold, determine the attitude compensation amount of the second state based on the direction deviation value and the non-gravitational acceleration compensation value; or If the direction deviation value is less than or equal to the second error threshold, determine the attitude compensation amount of the second state based on the non-gravitational acceleration compensation value; Preferably, the attitude adjustment module determines a non-gravitational acceleration compensation value based on the second inertial data, including: Determining an initial state estimate value and covariance information based on a dynamic behavior mathematical model and the initial attitude information of the first state; Performing Kalman gain update on the initial state estimate value and the covariance information based on the second inertial data to obtain a final state estimate value; Determining the final state estimate value as the non-gravitational acceleration compensation value; Preferably, the attitude determination module determines first attitude information based on the first inertial data and the second inertial data, including: Performing high-pass filtering and low-pass filtering on the first inertial data to obtain target inertial data of the first state; Performing normalization processing on the target inertial data to obtain second attitude information of the first state; Performing attitude recursive prediction on the second attitude information based on the first inertial data and the second inertial data to obtain the first attitude information.

9. An electronic device Characterized in that It includes a processor and a memory, and the memory stores multiple computer programs; the processor loads the computer programs from the memory to execute the artificial intelligence-based attitude prediction method according to any one of claims 1 to 7.

10. A computer-readable storage medium Characterized in that The computer-readable storage medium stores multiple computer programs, and the computer programs are suitable for being loaded by a processor to execute the artificial intelligence-based attitude prediction method according to any one of claims 1 to 7.

Images