Hip joint motion trail prediction method based on Lagrange physical neural network

Abstract

The invention relates to a hip joint motion track prediction method based on a Lagrange physical neural network. The method comprises the following steps: inputting kinematics data of a hip joint at a current moment into a Lagrange physical neural network, and training a kinetic parameter estimation module based on a pre-constructed loss function according to a trajectory prediction value and a real trajectory to obtain a trained Lagrange physical neural network; and predicting a hip joint movement track at the next moment by adopting a Lagrange physical neural network. By adopting the method, the unification of the motion trail prediction interpretability and the algorithm robustness is realized.

Description

Technical Field

[0001] The present application relates to the technical field of hip joint motion trajectory prediction, and in particular to a hip joint motion trajectory prediction method based on Lagrangian physical neural network. Background Art

[0002] Currently, hip joint trajectory prediction methods mainly include physical mechanism models and data-driven models, but both have obvious defects: the physical mechanism model is affected by the complexity of the human skeletal muscle model, and has problems such as cumbersome modeling, low prediction accuracy and poor generalization ability; the data-driven model is overly dependent on input data, resulting in insufficient predictive interpretability and poor stability. Summary of the Invention

[0003] Based on this, it is necessary to provide a hip joint motion trajectory prediction method based on Lagrangian physical neural network to address the above technical problems.

[0004] A hip joint motion trajectory prediction method based on Lagrangian physical neural network, the method comprising:

[0005] Inputting the kinematic data of the hip joint including the joint angle and joint angular velocity acquired by the inertial sensor at the current moment into the Lagrangian physical neural network; the Lagrangian physical neural network includes: a data-driven dynamic parameter estimation module and a physical mechanism-based trajectory prediction module; the dynamic parameter estimation module processes the acquired kinematic data to obtain predicted dynamic parameters of the hip joint including the inertia matrix and the gravity matrix; the predicted dynamic parameters are input into the trajectory prediction module, and the torque and acceleration at the current moment are obtained according to the Lagrangian dynamic equation, and then the trajectory prediction value of the hip joint at the next moment is obtained by subsequent integral calculation;

[0006] According to the trajectory prediction value and the actual trajectory, the dynamic parameter estimation module is trained based on a pre-constructed loss function to obtain a trained Lagrangian physical neural network;

[0007] The Lagrangian physical neural network is used to predict the hip joint motion trajectory at the next moment.

[0008] In one embodiment, the kinematic data includes: joint trajectory q t and joint velocity The predicted dynamic parameters include: inertia matrix and gravity matrix; and further include: processing the kinematic data to obtain the predicted dynamic parameters:

[0009]

[0010] Among them, M trepresents the inertia matrix at time t, G t represents the gravity matrix at time t, q t represents the joint trajectory at time t, ε represents the non-negative relaxation factor, I represents the identity matrix, P t and V t It is the intermediate variable of the hidden layer output in the multilayer perceptron structure in the dynamic parameter estimation module.

[0011] In one embodiment, the method further includes: calculating the kinetic energy at the current moment according to the inertia matrix as:

[0012]

[0013] Among them, T t represents the kinetic energy at time t, represents the joint velocity at time t.

[0014] In one embodiment, the method further includes: calculating the current kinetic energy and the intermediate variable V t , calculate the Lagrangian parameter L t :

[0015] L t =T t -V t ;

[0016] According to the Lagrange parameter L t , joint trajectory q t and the joint velocity Calculate the joint torque as:

[0017]

[0018] Among them, Q t represents the joint torque.

[0019] In one embodiment, the method further includes: calculating the joint angular acceleration according to the inertia matrix, the joint torque, and the gravity matrix:

[0020]

[0021] The joint angular acceleration is calculated by using the Runge-Kutta 4 integral method to obtain the joint velocity and joint trajectory at the next moment.

[0022] In one embodiment, the pre-built loss function is:

[0023]

[0024] Among them, q t+1 and is the actual joint trajectory and speed at the next moment, and The joint trajectory and velocity predicted by the Lagrangian physics neural network are trained by minimizing the loss function to obtain the model with the best prediction performance.

[0025] A device for predicting hip joint motion trajectory based on a Lagrangian physical neural network, comprising:

[0026] A model building module is used to input the kinematic data of the hip joint, including the joint angle and joint angular velocity, acquired by the inertial sensor at the current moment into a Lagrangian physical neural network; the Lagrangian physical neural network includes: a data-driven dynamic parameter estimation module and a physical mechanism-based trajectory prediction module; the dynamic parameter estimation module processes the acquired kinematic data to obtain predicted hip joint dynamic parameters, including an inertia matrix and a gravity matrix; the predicted dynamic parameters are input into the trajectory prediction module, and the torque and acceleration at the current moment are obtained according to the Lagrangian dynamic equation, and then the trajectory prediction value of the hip joint at the next moment is obtained by subsequent integral calculation;

[0027] A training module, configured to train the dynamic parameter estimation module based on a pre-established loss function according to the trajectory prediction value and the actual trajectory, to obtain a trained Lagrangian physical neural network;

[0028] The prediction module is used to predict the hip joint motion trajectory at the next moment by using the Lagrangian physical neural network.

[0029] A computer device includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0030] Inputting the kinematic data of the hip joint including the joint angle and joint angular velocity acquired by the inertial sensor at the current moment into the Lagrangian physical neural network; the Lagrangian physical neural network includes: a data-driven dynamic parameter estimation module and a physical mechanism-based trajectory prediction module; the dynamic parameter estimation module processes the acquired kinematic data to obtain predicted dynamic parameters of the hip joint including the inertia matrix and the gravity matrix; the predicted dynamic parameters are input into the trajectory prediction module, and the torque and acceleration at the current moment are obtained according to the Lagrangian dynamic equation, and then the trajectory prediction value of the hip joint at the next moment is obtained by subsequent integral calculation;

[0031] According to the trajectory prediction value and the actual trajectory, the dynamic parameter estimation module is trained based on a pre-constructed loss function to obtain a trained Lagrangian physical neural network;

[0032] The Lagrangian physical neural network is used to predict the hip joint motion trajectory at the next moment.

[0033] The Lagrangian physical neural network is used to predict the hip joint motion trajectory at the next moment.

[0034] A kinematic data acquisition device, comprising a pair of inertial sensors mounted on a thigh, is characterized in that the acquisition device can implement the following steps:

[0035] The kinematic data of the hip joint at the current moment is input into the above-mentioned Lagrangian physical neural network in real time.

[0036] This Lagrangian physics-based neural network-based hip joint trajectory prediction method overcomes the limitations of traditional single-modeling approaches by combining the advantages of physical mechanisms and data-driven methods. First, the model constructs a loss function by comparing the predicted trajectory with the actual measured value, and inversely optimizes the weight parameters of the data-driven model network. This improves the accuracy of dynamic parameter estimation while enhancing the model's adaptability to multi-scenario data. In this process, the model uses human hip joint kinematic data, such as joint trajectory and joint velocity, as input and utilizes the data-driven model to accurately estimate the core hip joint dynamic parameters, namely the inertia matrix and the gravity matrix. Subsequently, combining the forward and inverse computational framework of the Lagrangian dynamic equations, the theoretical values of joint torque and acceleration are derived, and the joint velocity and trajectory prediction for the next moment is iteratively generated through integral operations. This bidirectional closed-loop mechanism, combining data-driven parameter identification with physical model computational verification, inherits the advantages of neural networks in learning complex features while ensuring the physical plausibility and stability of the prediction results through the constraints of physical laws. Ultimately, it achieves a balance between interpretability and algorithmic robustness in trajectory prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 1 is a flow chart of a method for predicting hip joint motion trajectory based on a Lagrangian physical neural network in one embodiment;

[0038] Figure 2 is a schematic diagram of a Lagrangian physical neural network in one embodiment;

[0039] Figure 3 1 is a block diagram of a hip joint motion trajectory prediction device based on a Lagrangian physical neural network in one embodiment;

[0040] Figure 4 FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0042] In one embodiment, Figure 1 As shown, a hip joint motion trajectory prediction method based on Lagrangian physical neural network is provided, which includes the following steps:

[0043] Step 102: Input the kinematic data of the hip joint including the joint angle and joint angular velocity acquired by the inertial sensor at the current moment into the Lagrangian physical neural network.

[0044] The Lagrangian physical neural network includes: a data-driven dynamic parameter estimation module and a physical mechanism-based trajectory prediction module. The dynamic parameter estimation module processes the collected kinematic data to obtain predicted hip joint dynamic parameters including the inertia matrix and gravity matrix; the predicted dynamic parameters are input into the trajectory prediction module, and the torque and acceleration at the current moment are obtained according to the Lagrangian dynamic equation. The subsequent integral calculation then obtains the predicted trajectory value of the hip joint at the next moment.

[0045] Step 104 : According to the trajectory prediction value and the actual trajectory, the dynamic parameter estimation module is trained based on a pre-built loss function to obtain a trained Lagrangian physical neural network.

[0046] Step 106: Use the Lagrangian physical neural network to predict the hip joint motion trajectory at the next moment.

[0047] The aforementioned hip joint trajectory prediction method based on a Lagrangian physics-based neural network overcomes the limitations of traditional single-modeling approaches by integrating the advantages of both physical mechanisms and data-driven methods. First, the model constructs a loss function by comparing the predicted trajectory with the actual measured value, and inversely optimizes the weight parameters of the data-driven model network. This improves the accuracy of dynamic parameter estimation while enhancing the model's adaptability to multi-scenario data. In this process, the model uses human hip joint kinematic data, such as joint trajectory and joint velocity, as input and utilizes the data-driven model to accurately estimate the core hip joint dynamic parameters, namely the inertia matrix and the gravity matrix. Subsequently, combining the forward and inverse computational framework of the Lagrangian dynamic equations, the theoretical values of the joint torque and acceleration are derived, and the joint velocity and trajectory prediction for the next moment is iteratively generated through integral operations. This bidirectional closed-loop mechanism, combining data-driven parameter identification with physical model computational verification, inherits the advantages of neural networks in learning complex features while ensuring the physical plausibility and stability of the prediction results through the constraints of physical laws. Ultimately, it achieves a balance between interpretability and algorithmic robustness in trajectory prediction.

[0048] In one embodiment, Figure 2 As shown in Figure 1, the Lagrangian physical neural network includes (a) a dynamic parameter estimation module for dynamic parameter estimation and (b) a trajectory prediction module for joint velocity and trajectory calculation. For dynamic parameter estimation, the input kinematic data includes joint trajectory and joint velocity; the output predicted dynamic parameters include inertia matrix and gravity matrix. After processing the kinematic data, the predicted dynamic parameters are:

[0049]

[0050] Among them, M t represents the inertia matrix at time t, G t represents the gravity matrix at time t, q t represents the joint trajectory at time t, ε represents the non-negative relaxation factor, I represents the identity matrix, P t and V t It is the intermediate variable of the hidden layer output in the multilayer perceptron structure in the dynamic parameter estimation module.

[0051] It is worth noting that the nonlinear relationship between the input kinematic data and the dynamic parameters is established through the hidden layers of the multilayer perceptron structure. At the same time, the dynamic parameters of the model output are manually set to ensure their physical rationality. The inertia matrix M is guaranteed to be positive definite by adding a non-negative relaxation factor ε multiplied by the identity matrix I, and the dimensionality of the output matrix is guaranteed by setting corresponding structural parameters.

[0052] In one embodiment, the joint velocity and trajectory calculation part is implemented using a trajectory prediction module based on physical mechanisms. For the input inertia matrix, the kinetic energy at the current moment is calculated as:

[0053]

[0054] Among them, T t represents the kinetic energy at time t, Represents the joint velocity at time t. Then, according to the current kinetic energy and the intermediate variable V t , calculate the Lagrangian parameter L t :

[0055] L t =T t -V t

[0056] Then according to the Lagrange parameter L t , joint trajectory and joint velocity, calculate the joint torque as:

[0057]

[0058] Among them, Q t represents the joint torque.

[0059] At this point, the joint angular acceleration is calculated by combining the Lagrange forward and inverse dynamics models:

[0060]

[0061] Finally, the integral calculation of Runge-Kutta 4 obtains the joint velocity and joint trajectory at the next moment.

[0062] In one embodiment, the loss function is:

[0063]

[0064] Among them, q t+1 and is the actual joint trajectory and speed at the next moment, and The joint trajectory and velocity predicted by the Lagrangian physics neural network are trained by minimizing the loss function to obtain the model with the best prediction performance.

[0065] It should be understood that although Figure 1 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. In addition, Figure 1At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0066] In one embodiment, Figure 3 As shown, a hip joint motion trajectory prediction device based on Lagrangian physical neural network is provided, including: a model building module 302, a training module 304 and a prediction module 306, wherein:

[0067] The model building module 302 is used to input the kinematic data of the hip joint, including the joint angle and joint angular velocity, acquired by the inertial sensor at the current moment into a Lagrangian physical neural network; the Lagrangian physical neural network includes: a data-driven dynamic parameter estimation module and a physical mechanism-based trajectory prediction module; the dynamic parameter estimation module processes the acquired kinematic data to obtain predicted hip joint dynamic parameters, including an inertia matrix and a gravity matrix; the predicted dynamic parameters are input into the trajectory prediction module, and the torque and acceleration at the current moment are obtained according to the Lagrangian dynamic equation. Then, a subsequent integral calculation is performed to obtain a predicted trajectory value of the hip joint at the next moment;

[0068] A training module 304 is configured to train the dynamic parameter estimation module based on the trajectory prediction value and the actual trajectory and a pre-established loss function to obtain a trained Lagrangian physical neural network;

[0069] The prediction module 306 is used to predict the hip joint motion trajectory at the next moment using the Lagrangian physical neural network.

[0070] In one embodiment, the training module 304 trains the best-performing model based on the set loss function:

[0071]

[0072] In one embodiment, the kinematic data includes: joint trajectories and joint velocities; the predicted dynamic parameters include: inertia matrix and gravity matrix; the model building module 304 is further used to process the kinematic data to obtain the predicted dynamic parameters as follows:

[0073]

[0074] Among them, M t represents the inertia matrix at time t, G t represents the gravity matrix at time t, qt represents the joint trajectory at time t, ε represents the non-negative relaxation factor, I represents the identity matrix, P t and V t It is the intermediate variable of the hidden layer output in the multilayer perceptron structure in the dynamic parameter estimation module.

[0075] In one embodiment, the model building module 304 is further configured to calculate the kinetic energy at the current moment according to the inertia matrix:

[0076]

[0077] Among them, T t represents the kinetic energy at time t, represents the joint velocity at time t.

[0078] In one embodiment, the model building module 304 is further configured to calculate the current kinetic energy and the intermediate variable V t , calculate the Lagrangian parameter L t :

[0079] L t =T t -V t

[0080] According to the Lagrange parameter L t , joint trajectory and the joint velocity, calculate the joint torque as:

[0081]

[0082] Among them, Q t represents the joint torque.

[0083] In one embodiment, the model building module 304 is further configured to calculate the joint angular acceleration according to the inertia matrix, the joint torque and the gravity matrix.

[0084]

[0085] In one embodiment, the prediction module 306 uses Runge-Kutta 4 integration calculation based on the joint angular acceleration to obtain the joint velocity and the joint trajectory at the next moment.

[0086] Regarding the specific limitations of the hip joint motion trajectory prediction device based on the Lagrangian physical neural network, please refer to the limitations of the hip joint motion trajectory prediction method based on the Lagrangian physical neural network mentioned above, which will not be repeated here. The various modules in the above-mentioned hip joint motion trajectory prediction device based on the Lagrangian physical neural network can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0087] In one embodiment, a kinematic data acquisition device is provided. The kinematic data acquisition device may be a terminal, and its internal structure diagram may be as follows: Figure 4 As shown. The kinematic data acquisition device includes a processor, a memory, an inertial sensor and an input device connected via a system bus. The processor of the kinematic data acquisition device is used to provide computing and control capabilities. The memory of the kinematic data acquisition device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The inertial sensor of the kinematic data acquisition device is used to collect kinematic data of the hip joint including joint angle and joint angular velocity. When the computer program is executed by the processor, a hip joint motion trajectory prediction method based on Lagrangian physical neural network is implemented. The input device of the computer device can be a touch layer covered on the display screen, or a button, trackball or touchpad provided on the computer device housing, or an external keyboard, touchpad or mouse.

[0088] Those skilled in the art will understand that Figure 4 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0089] In one embodiment, a kinematic data acquisition device is provided, comprising a pair of inertial sensors mounted on the thigh, and inputting the acquired kinematic data including joint angles and joint angular velocities into a Lagrangian physical neural network.

[0090] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0091] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0092] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and such modifications and improvements are intended to fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A hip joint motion trajectory prediction method based on Lagrangian physical neural network, characterized in that: The method comprises: Inputting the kinematic data of the hip joint including the joint angle and joint angular velocity acquired by the inertial sensor at the current moment into the Lagrangian physical neural network; the Lagrangian physical neural network includes: a data-driven dynamic parameter estimation module and a physical mechanism-based trajectory prediction module; the dynamic parameter estimation module processes the acquired kinematic data to obtain predicted dynamic parameters of the hip joint including the inertia matrix and the gravity matrix; the predicted dynamic parameters are input into the trajectory prediction module, and the torque and acceleration at the current moment are obtained according to the Lagrangian dynamic equation, and then the trajectory prediction value of the hip joint at the next moment is obtained by subsequent integral calculation; According to the trajectory prediction value and the actual trajectory, the dynamic parameter estimation module is trained based on a pre-constructed loss function to obtain a trained Lagrangian physical neural network; The Lagrangian physical neural network is used to predict the hip joint motion trajectory at the next moment.

2. The method according to claim 1, characterized in that The kinematic data includes: joint trajectory and joint velocity; the predicted dynamic parameters include: inertia matrix and gravity matrix; The kinematic data is processed to obtain predicted kinetic parameters, including: M t =P t P t T +εI Among them, M t represents the inertia matrix at time t, G t represents the gravity matrix at time t, q t represents the joint trajectory at time t, ε represents the non-negative relaxation factor, I represents the identity matrix, P t and V t It is the intermediate variable of the hidden layer output in the multi-layer perceptron structure of the dynamic parameter estimation module.

3. The method according to claim 2, characterized in that After inputting the predicted dynamic parameters into the trajectory prediction module, the method further includes: According to the inertia matrix, the kinetic energy at the current moment is calculated as: Among them, T t represents the kinetic energy at time t, represents the joint velocity at time t.

4. The method according to claim 3, characterized in that After inputting the predicted dynamic parameters into the trajectory prediction module, the method further includes: According to the kinetic energy at the current moment and the intermediate variable V t , calculate the Lagrangian parameter L t ; According to the Lagrange parameter L t , joint trajectory and the joint velocity, calculate the joint torque as: Among them, Q t represents the joint torque.

5. The method according to claim 4, characterized in that Inputting the predicted dynamic parameters into the trajectory prediction module to obtain the trajectory prediction value of the hip joint at the next moment includes: Calculate the joint angular acceleration based on the inertia matrix, joint torque and gravity matrix; The joint angular acceleration is calculated by using the Runge-Kutta 4 integral method to obtain the joint velocity and joint trajectory at the next moment.

6. The training module according to claim 5, wherein the pre-built loss function is: in, q t+1 and is the actual joint trajectory and speed at the next moment, and The joint trajectory and velocity predicted by the Lagrangian physics neural network are trained by minimizing the loss function to obtain the model with the best prediction performance.

7. A hip joint motion trajectory prediction device based on Lagrangian physical neural network, characterized in that: The device comprises: A model building module is used to input the kinematic data of the hip joint, including the joint angle and joint angular velocity, acquired by the inertial sensor at the current moment into a Lagrangian physical neural network; the Lagrangian physical neural network includes: a data-driven dynamic parameter estimation module and a physical mechanism-based trajectory prediction module; the dynamic parameter estimation module processes the acquired kinematic data to obtain predicted hip joint dynamic parameters, including an inertia matrix and a gravity matrix; the predicted dynamic parameters are input into the trajectory prediction module, and the torque and acceleration at the current moment are obtained according to the Lagrangian dynamic equation, and then the trajectory prediction value of the hip joint at the next moment is obtained by subsequent integral calculation; A training module, configured to train the dynamic parameter estimation module based on a pre-established loss function according to the trajectory prediction value and the actual trajectory, to obtain a trained Lagrangian physical neural network; The prediction module is used to predict the hip joint motion trajectory at the next moment by using the Lagrangian physical neural network.

8. The calculation module according to claim 7, characterized in that: The kinematic data includes: joint trajectory and joint velocity; the predicted dynamic parameters include: inertia matrix and gravity matrix; the calculation module is further used to process the kinematic data to obtain the predicted dynamic parameters: M t =P t P t T +εI Among them, M t represents the inertia matrix at time t, G t represents the gravity matrix at time t, q t represents the joint trajectory at time t, ε represents the non-negative relaxation factor, I represents the identity matrix, P t and V t It is the intermediate variable of the hidden layer output in the multilayer perceptron structure in the dynamic parameter estimation module.

9. A kinematic data acquisition device comprising a pair of inertial sensors mounted on the thigh, characterized in that: The acquisition device can realize the input data required by the method according to any one of claims 1 to 6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images