Lower extremity exoskeleton control method, device, computer equipment and storage medium for ultra-high voltage power grid

Abstract

Embodiments of the present application provide a lower limb exoskeleton control method and device for an ultrahigh-voltage power grid, computer equipment, a storage medium and a computer program product, relating to the technical field of exoskeletons. The method comprises: acquiring load information data and motion information data of a target object; using a pre-trained load identification model, based on the load information data, acquiring working mode information of a target exoskeleton matched with the target object; using a pre-trained posture estimation model, based on posture information data contained in the motion information data, acquiring motion state information of the target exoskeleton; and according to the working mode information and the motion state information, acquiring an output torque of the target exoskeleton, and adjusting the running state of the target exoskeleton using the output torque. The method improves the efficiency and safety of ultrahigh-voltage power grid operations.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of exoskeleton, and in particular to a lower limb exoskeleton control method and device for an ultrahigh-voltage power grid, a computer device, a storage medium, and a computer program product. BACKGROUND

[0002] In the maintenance of an ultrahigh-voltage power grid, the working conditions and environment involved are often extremely arduous. In order to improve the working environment of a target object, improve work efficiency, and ensure work safety, exoskeleton technology can be introduced into the maintenance work of a high-voltage power grid.

[0003] Currently, exoskeleton technology applied in the maintenance of an ultrahigh-voltage power grid often relies on a pre-set operation mode or manual control, which can cause slow response of the exoskeleton and low efficiency, for example, slow switching of the working mode, difficulty in adapting to a complex and changeable working environment, and slow response to the movements of the target object, which can cause safety hazards and low efficiency and safety. SUMMARY

[0004] Therefore, it is necessary to provide a lower limb exoskeleton control method and device for an ultrahigh-voltage power grid, a computer device, a storage medium, and a computer program product to solve the above technical problems.

[0005] In a first aspect, the present application provides a lower limb exoskeleton control method for an ultrahigh-voltage power grid. The method comprises:

[0006] obtaining load information data and motion information data of a target object;

[0007] using a pre-trained load recognition model to obtain working mode information of a target exoskeleton matched with the target object based on the load information data;

[0008] using a pre-trained posture estimation model to obtain motion state information of the target exoskeleton based on posture information data contained in the motion information data;

[0009] obtaining output torque of the target exoskeleton according to the working mode information and the motion state information, and adjusting the running state of the target exoskeleton using the output torque.

[0010] In one embodiment, the step of using a pre-trained load recognition model to obtain the working mode information of the target exoskeleton matching the target object based on the load information data includes: inputting the load information data into the load recognition model to obtain the working mode information of the target exoskeleton; wherein the load recognition model obtains the load label of the target object based on the load information data, and obtains the working mode information of the target exoskeleton based on the load label.

[0011] In one embodiment, the step of obtaining the load label of the target object by the load recognition model based on the load information data, and obtaining the working mode information of the target exoskeleton based on the load label, includes: a first feature extraction module included in the load recognition model extracting features from the load information data to determine and output the load feature information of the target object; and a first classification module included in the load recognition model determining the load label of the target object based on the load feature information output by the first feature extraction module, and determining and outputting the working mode information of the target exoskeleton based on the load label.

[0012] In one embodiment, the step of using a pre-trained posture estimation model to obtain the motion state information of the target exoskeleton based on the posture information data contained in the motion information data includes: inputting the posture information data into the posture estimation model to obtain the motion state information of the target exoskeleton; wherein, the posture estimation model obtains the posture pattern of the target object based on the posture information data, and obtains the motion state information of the target exoskeleton based on the posture pattern.

[0013] In one embodiment, the step of obtaining the posture pattern of the target object by the posture estimation model based on the posture information data, and obtaining the motion state information of the target exoskeleton based on the posture pattern, includes: a second feature extraction module included in the posture estimation model extracting features from the posture information data to determine and output the posture feature information of the target object; and a second classification module included in the posture estimation model determining the posture pattern of the target object based on the posture feature information output by the second feature extraction module, and determining and outputting the motion state information of the target exoskeleton based on the posture pattern.

[0014] In one embodiment, obtaining the output torque of the target exoskeleton based on the working mode information and the motion state information includes: using a pre-constructed exoskeleton system model of the target exoskeleton system, obtaining the output torque of the target exoskeleton based on the working mode information and the motion state information.

[0015] In one embodiment, the exoskeleton system model includes an objective function and constraints; the step of obtaining the output torque of the target exoskeleton based on the pre-constructed exoskeleton system model and the working mode information and the motion state information includes: under the constraints, solving the objective function based on the working mode information and the motion state information to obtain the output torque of the target exoskeleton in a future time period; the future time period includes at least one future time step; the output torque includes a single output torque for each of at least one of the future time steps.

[0016] Secondly, this application provides a lower limb exoskeleton control device for ultra-high voltage power grids. The device includes:

[0017] The data acquisition module is used to acquire load information data and motion information data of the target object;

[0018] The first calculation module is used to obtain the working mode information of the target exoskeleton that matches the target object based on the load information data using a pre-trained load recognition model.

[0019] The second calculation module is used to obtain the motion state information of the target exoskeleton based on the posture information data contained in the motion information data by using a pre-trained posture estimation model.

[0020] The third calculation module is used to obtain the output torque of the target exoskeleton based on the working mode information and the motion state information, and to adjust the running state of the target exoskeleton using the output torque.

[0021] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps:

[0022] Acquire load and motion information data of the target object;

[0023] Using a pre-trained load recognition model, based on the load information data, the working mode information of the target exoskeleton that matches the target object is obtained;

[0024] Using a pre-trained pose estimation model, the motion state information of the target exoskeleton is obtained based on the pose information data contained in the motion information data;

[0025] Based on the working mode information and the motion state information, the output torque of the target exoskeleton is obtained, and the operating state of the target exoskeleton is adjusted using the output torque.

[0026] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, performs the following steps:

[0027] Acquire load and motion information data of the target object;

[0028] Using a pre-trained load recognition model, based on the load information data, the working mode information of the target exoskeleton that matches the target object is obtained;

[0029] Using a pre-trained pose estimation model, the motion state information of the target exoskeleton is obtained based on the pose information data contained in the motion information data;

[0030] Based on the working mode information and the motion state information, the output torque of the target exoskeleton is obtained, and the operating state of the target exoskeleton is adjusted using the output torque.

[0031] Fifthly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, performs the following steps:

[0032] Acquire load and motion information data of the target object;

[0033] Using a pre-trained load recognition model, based on the load information data, the working mode information of the target exoskeleton that matches the target object is obtained;

[0034] Using a pre-trained pose estimation model, the motion state information of the target exoskeleton is obtained based on the pose information data contained in the motion information data;

[0035] Based on the working mode information and the motion state information, the output torque of the target exoskeleton is obtained, and the operating state of the target exoskeleton is adjusted using the output torque.

[0036] In the aforementioned lower limb exoskeleton control method, device, computer equipment, storage medium, and computer program product for ultra-high voltage power grids, firstly, load information data and motion information data of the target object can be acquired; nextly, a pre-trained load recognition model can be used to acquire the working mode information of the target exoskeleton matching the target object based on the load information data; furthermore, a pre-trained posture estimation model can be used to acquire the motion state information of the target exoskeleton based on the posture information data contained in the motion information data; finally, the output torque of the target exoskeleton can be acquired according to the working mode information and motion state information, and the operating state of the target exoskeleton can be adjusted using the output torque. In the method provided in this application embodiment, a load recognition model is set in the exoskeleton system of the target exoskeleton. This load recognition model can be used to realize the real-time perception and recognition of the load on the target object, so that the exoskeleton can autonomously adjust its structure and working mode to adapt to different load states. A posture estimation model is set in the exoskeleton system of the target exoskeleton, so that the exoskeleton can naturally change its posture with the movement and operation of the target object and keep in sync with the action of the target object, which can improve comfort and usage efficiency. Furthermore, the torque control module can be used to adjust the output torque of the target exoskeleton in real time according to the working mode information and motion state information of the target exoskeleton, which improves the efficiency and safety of exoskeleton use, and thus improves the efficiency and safety of ultra-high voltage power grid operations. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 A flowchart illustrating a lower limb exoskeleton control method for an ultra-high voltage power grid, provided as an embodiment of this application;

[0039] Figure 2 This application provides a schematic diagram of a process for obtaining working mode information in an embodiment of the present application.

[0040] Figure 3 This is a schematic diagram of a process for obtaining motion state information provided in an embodiment of this application;

[0041] Figure 4 A structural block diagram of a lower limb exoskeleton control device for an ultra-high voltage power grid provided in this application embodiment;

[0042] Figure 5This is an internal structural diagram of a computer device provided in an embodiment of this application. Detailed Implementation

[0043] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0044] In one exemplary embodiment, such as Figure 1 As shown, a method for controlling a lower limb exoskeleton for ultra-high voltage power grids is provided. This embodiment illustrates the method by applying it to a server. It is understood that this method can also be applied to a terminal, or to a system including both a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0045] Step 102: Obtain the load information data and motion information data of the target object.

[0046] The target object can be any object equipped with a target exoskeleton, such as a worker equipped with a lower limb exoskeleton. The target exoskeleton can be a human exoskeleton and / or a powered exoskeleton, such as a lower limb exoskeleton. This human exoskeleton and / or powered exoskeleton can be a machine device that uses mechanical devices to enhance the structural strength, motor power, and endurance of human limbs. Its working principle is to detect the human's movement intention and state through sensors, then provide energy from a power unit, and transmit the power to the execution unit through a transmission device, thereby assisting or enhancing the human's movement. The lower limb exoskeleton control method for ultra-high voltage power grids provided in this application embodiment can be applied to the exoskeleton system of this target exoskeleton. This exoskeleton system can include various sensors, a load identification module, a posture estimation module, and a torque control module. These various sensors can include, but are not limited to, force sensors, pressure sensors, accelerometers, and gyroscopes. The force sensor and pressure sensor can be used to measure the external load of the target object and monitor the initial load information data of the target object; the accelerometer and gyroscope can be used to monitor the posture and motion state of the target object and monitor the initial motion information data of the target object. The initial load information data can include force and pressure data, torque data, etc., of the target object. Force and pressure sensors directly measure the external load borne by the target object. For example, when the target object is moving power grid equipment, force sensors in the hands, shoulders, etc., can measure the magnitude of the tensile or compressive force generated by the weight of the equipment, which can intuitively reflect the real-time load of the target object and help the exoskeleton determine the required support force. Some force sensors can measure output torque. For example, sensors at the joints can monitor the torque generated by limb movement and load. When climbing or stretching during ultra-high voltage power grid maintenance, joint torque data helps the exoskeleton adjust the auxiliary force of the corresponding joints to ensure synchronization with the target object's movements. The initial motion information data can include the initial posture information data of the target object. For example, gyroscopes measure the angular velocity of various parts of the target object's body, reflecting the joint rotation speed. When the target object performs twisting, turning, or other movements, angular velocity data can help the exoskeleton sense posture changes, adjust its own posture, maintain the target object's balance, and improve user comfort and safety. The initial motion information data can also include the speed and acceleration of the target object. For example, when the target object walks quickly, suddenly starts or stops, the acceleration data will change accordingly. The target exoskeleton can predict the movement trend of the target object based on this, adjust its own state in advance, provide appropriate assistance, and ensure the smooth movement of the target object.Next, the initial load information data detected by the force sensor and pressure sensor can be preprocessed, for example, by filtering and denoising, to obtain the load information data corresponding to the initial load information data; and the initial motion information data of the target object detected by the accelerometer and gyroscope can be preprocessed, for example, by filtering and denoising, to obtain the motion information data corresponding to the initial motion information data. In one possible implementation, a low-pass filter can be used to smooth the data, and a Kalman filter can be used to fuse multi-sensor data to improve the accuracy of attitude estimation.

[0047] Step 104: Using a pre-trained load recognition model, based on load information data, obtain the working mode information of the target exoskeleton that matches the target object.

[0048] The exoskeleton system includes a load recognition module that can sense changes in the load of the target object in real time. Based on the load state of the target object, the exoskeleton can be switched to the corresponding working mode. The exoskeleton has at least one working mode, and each working mode corresponds to a load state of the target object. The load recognition module is configured with a load recognition model. First, pre-processed load information data monitored by force and pressure sensors can be acquired. Next, using the pre-trained load recognition model, the working mode information of the target exoskeleton matching the target object can be obtained based on the load information data. Inputting this load information data into the pre-trained load recognition model allows the corresponding working mode of the target exoskeleton to be determined and output. The training process of this load recognition model may include: collecting multiple initial sample load data through sensor monitoring, and preprocessing the initial sample load data to obtain multiple sample load data; next, obtaining a load sample training set from the sample load data, which may include sample load data corresponding to each load state under various load conditions, as well as the actual working mode labels of the target exoskeleton corresponding to each load state; further, inputting the multiple sample load data into the load recognition model to be trained, which may be a neural network model, such as a multilayer perceptron, convolutional neural network, or recurrent neural network. The load recognition model may include at least one convolutional layer and at least one pooling layer. The convolutional layer is used to extract features from the input data; the pooling layer is used to sample the input data. Both the convolutional layer and the pooling layer include activation functions.

[0049] Specifically, convolutional layers can be used to extract initial features from multiple sample payload data. First, the multiple sample payload data are vectorized to obtain multiple sample payload data vectors, which can be combined into a vector matrix. Second, this vector matrix is ​​input into the convolutional layer, where a convolution operation is performed between the convolution kernel and the vector matrix, i.e., the inner product of the vector matrix and the convolution kernel, to obtain the convolution result corresponding to the vector matrix. Next, a non-linear transformation is applied to the convolution result based on an activation function, and a bias vector is added to obtain the initial feature vector. Third, the initial feature vector is input into a pooling layer, where feature sampling is performed. Then, a non-linear transformation is applied to the feature sampling result based on an activation function, and a bias vector is added to obtain the sample payload features.

[0050] The load recognition model may include at least one fully connected layer, which can classify the sample load features to obtain the predicted working mode label of the target exoskeleton corresponding to the sample load data. Additionally, the fully connected layer may include an activation function, which includes a weight matrix and a bias constant.

[0051] Specifically, the sample load features can be input into a fully connected layer. Based on the weight matrix and bias vector of the activation function, the sample load features are nonlinearly transformed and then normalized to obtain the predicted working mode label of the target exoskeleton corresponding to the sample load data.

[0052] Furthermore, the load recognition model to be trained can be trained based on the difference between the predicted working mode label of the target exoskeleton corresponding to the sample load data and the actual working mode label of the target exoskeleton corresponding to the sample load data. The difference between the predicted working mode label and the actual working mode label can be obtained by using the loss function of the load recognition model to be trained. The training process of the load recognition model to be trained can be found in equation (1):

[0053]

[0054] in, These are model parameters. It is a loss function. It is the model's predicted output. It's a real label. It is a regularization parameter. It is a regularization term.

[0055] Therefore, the trained load recognition model can be deployed in the load recognition module of the exoskeleton system of the target exoskeleton.

[0056] Step 106: Using a pre-trained pose estimation model, based on the pose information data contained in the motion information data, obtain the motion state information of the target exoskeleton.

[0057] The exoskeleton system includes a posture estimation module that can sense the posture changes of the target object in real time and adjust the motion state of the exoskeleton accordingly. First, posture information data, including preprocessed motion data monitored by accelerometers and gyroscopes, can be acquired. Next, a pre-trained posture estimation model can be used to obtain the posture feature information of the target object based on this posture information data. In one possible implementation, quaternions and / or direction cosine matrices can be used to estimate the posture changes of the target object. Further, the pre-trained posture estimation model can be used to determine the posture category of the target object, such as standing, walking, and squatting, based on the posture feature information. Then, the motion state of the exoskeleton can be adjusted according to the posture category of the target object.

[0058] Step 108: Based on the working mode information and motion state information, obtain the output torque of the target exoskeleton, and use the output torque to adjust the running state of the target exoskeleton.

[0059] In this step, Model Predictive Control (MPC) can be used to predict the output torque of the target exoskeleton in the future. First, an exoskeleton system model of the target exoskeleton system can be established. This exoskeleton system model can be a dynamic model, which can describe the behavior of each component of the target exoskeleton (e.g., the multiple joints contained in the target exoskeleton) based on the kinematic and dynamic principles of the target exoskeleton. In one possible implementation, the dynamic model of the exoskeleton system can be constructed based on the Lagrange equation or the Newton-Euler rule. The target exoskeleton can include multiple joints, and the state of each joint can be represented in spatial state form, as shown in equations (2) and (3).

[0060]

[0061] in, It could be the target exoskeleton at time step The state vector can include, for example, information such as the position, velocity, and acceleration of multiple joints; It can be at the time step The control input may include the joint torque of each joint in the plurality of shutdowns; It could be that the exoskeleton system is at a certain time step. The output vector can include measurable variables in the exoskeleton system, such as the angles of the joints of the target exoskeleton; , , , It is the system matrix.

[0062] Next, in MPC, the motion state of the target exoskeleton in the future can be predicted based on the current state information of the target exoskeleton at each time step, i.e., the state vector. The current state information may include, but is not limited to, the working mode information and motion state information of the target exoskeleton, etc. The working mode information of the target exoskeleton is associated with the load state information of the target object; the future time period, i.e., the future time period, may include multiple future time steps. In one possible implementation, a rolling time window with a prediction window of 𝑁 can be used to predict the motion state of the target exoskeleton in N future time steps, as shown in equations (4) and (5):

[0063]

[0064] in, This could be the target exoskeleton in future time steps. The state vector can include, for example, information such as the position, velocity, and acceleration of multiple joints; It could be a step in the future. The control input may include the joint torque of each joint in the plurality of shutdowns; This exoskeleton system could be a step in the future. The output vector can include measurable variables in the exoskeleton system, such as the angles of the joints of the target exoskeleton; , , , It is the system matrix;

[0065] Furthermore, based on the motion state of the target exoskeleton at N future time steps, i.e., the output vector, the control input, i.e., the joint matrix, of the target exoskeleton at N future time steps can be predicted. In one possible implementation, an objective function for the target exoskeleton can be constructed, and constraints can be set. Under these constraints, the objective function can be solved. By minimizing the objective function, an optimal sequence of control inputs can be obtained. This sequence of control inputs can include the control inputs at N future time steps. The objective function can be found in equation (6), and the constraints can include control input constraints, as shown in equation (7).

[0066]

[0067] in, It could be the desired output of the exoskeleton system (e.g., the joint positions and joint velocities of the target exoskeleton). It could be a future time step The control input, namely the joint torque; and It is a weight matrix.

[0068]

[0069] in, and It controls the upper and lower limits of input.

[0070] In each prediction cycle, MPC needs to determine the optimal control input sequence by solving an optimization problem. This optimization process is usually solved using numerical optimization algorithms, such as gradient descent or least squares. The result is the optimal control input sequence for a future time period. This sequence can include control inputs for multiple future time steps. Typically, in the multiple future time steps of the current prediction cycle, the control input of the target exoskeleton at the first of these multiple future time steps, i.e., joint torque, can be used. When the exoskeleton system enters the next time step of the current time step, the next prediction cycle can be started, which will yield the control input of the first of the new multiple future time steps, and so on.

[0071] In addition, the exoskeleton system may include a safety protection module that immediately activates a protection mode if an abnormal operation of the target exoskeleton is detected. In one possible implementation, a detection threshold can be set for the target exoskeleton based on historical sensor data to detect abnormal operation of the target exoskeleton. For example, torque threshold, acceleration threshold, and angular velocity threshold, etc., can be found in equation (8):

[0072]

[0073] in, It is the average value of the sensor data. It is the standard deviation of the sensor data. It is a constant used to adjust the sensitivity. In another possible implementation, machine learning models can be used for anomaly detection, or statistical methods can be employed to identify anomalies.

[0074] The design of the protection mode may include:

[0075] 1. Torque Limitation: Immediately limit the output torque of the exoskeleton to prevent excessive mechanical load on the target object. This can be achieved by dynamically adjusting the output limit of the torque controller, see formula (9):

[0076]

[0077] in, It is the safety torque after being limited. This is the currently calculated torque. It is the set maximum allowable torque.

[0078] 2. Movement Stop: Upon detection of a severe abnormality, all movement of the exoskeleton is immediately stopped to prevent further injury. This can be achieved by setting the control input to zero, see equation (10):

[0079]

[0080] 3. Alarm system: Triggers audible and visual alarms or sends alarm signals to the target object and monitoring system so that further safety measures can be taken in a timely manner.

[0081] 4. Data logging: Record data on abnormal events for post-event analysis and system improvement.

[0082] In the method of this embodiment, firstly, load information data and motion information data of the target object can be obtained; nextly, a pre-trained load recognition model can be used to obtain the working mode information of the target exoskeleton that matches the target object based on the load information data; further, a pre-trained posture estimation model can be used to obtain the motion state information of the target exoskeleton based on the posture information data contained in the motion information data; finally, the output torque of the target exoskeleton can be obtained according to the working mode information and motion state information, and the operating state of the target exoskeleton can be adjusted using the output torque. In the method provided in this application embodiment, a load recognition model is set in the exoskeleton system of the target exoskeleton. This load recognition model can be used to realize the real-time perception and recognition of the load on the target object, so that the exoskeleton can autonomously adjust its structure and working mode to adapt to different load states. A posture estimation model is set in the exoskeleton system of the target exoskeleton, so that the exoskeleton can naturally change its posture with the movement and operation of the target object and keep in sync with the action of the target object to improve comfort and usage efficiency. Furthermore, the torque control module can be used to adjust the output torque of the target exoskeleton in real time according to the working mode information and motion state information of the target exoskeleton, thereby improving the efficiency and safety of exoskeleton use, and thus improving the efficiency and safety of ultra-high voltage power grid operations.

[0083] In one exemplary embodiment, step 104 may include:

[0084] The load information data is input into the load recognition model to obtain the working mode information of the target exoskeleton; the load recognition model obtains the load label of the target object based on the load information data, and obtains the working mode information of the target exoskeleton based on the load label.

[0085] In one exemplary embodiment, such as Figure 2 As shown, the load recognition model obtains the load label of the target object based on the load information data, and obtains the working mode information of the target exoskeleton based on the load label, which may include steps 202 to 204. Wherein:

[0086] Step 202: The first feature extraction module included in the load identification model extracts features from the load information data, determines and outputs the load feature information of the target object.

[0087] The first feature extraction module may include at least one convolutional layer and at least one pooling layer. The convolutional layer is used to extract features from the input data; the pooling layer is used to sample the input data. Both the convolutional and pooling layers include activation functions. The payload data may include multiple textual information.

[0088] Specifically, convolutional layers can be used to extract initial features from multiple textual pieces of information. The first step involves vectorizing the multiple textual pieces of information to obtain multiple textual information vectors, which can be combined into a single textual information vector matrix. The second step involves inputting this textual information vector matrix into the convolutional layer, where a convolution operation is performed between the kernel and the textual information vector matrix. This involves performing an inner product operation between the kernel and the textual information vector matrix to obtain the convolution result. Next, a non-linear transformation is applied to the convolution result based on an activation function, and a bias vector is added to obtain the initial feature vector. The third step involves inputting the initial feature vector into a pooling layer to sample the initial feature vector. Then, a non-linear transformation is applied to the sampled feature result based on an activation function, and a bias vector is added to obtain the loaded feature information.

[0089] Step 204: The first classification module included in the load recognition model determines the load label of the target object based on the load feature information output by the first feature extraction module, and determines and outputs the working mode information of the target exoskeleton based on the load label.

[0090] The first classification module may include at least one fully connected layer, which can classify the load feature information to obtain the working mode information of the target exoskeleton. Additionally, the fully connected layer may include an activation function, which includes a weight matrix and a bias constant.

[0091] Specifically, load feature information can be input into a fully connected layer, and the load feature information can be nonlinearly transformed based on the weight matrix and bias vector of the activation function. Then, through normalization, the working mode information of the target exoskeleton can be obtained.

[0092] In the method of this embodiment, a load recognition model is set in the exoskeleton system of the target exoskeleton. The load recognition model can be used to realize the real-time perception and recognition of the load on the target object, so that the exoskeleton can autonomously adjust its structure and working mode to adapt to different load states, thereby improving the efficiency of exoskeleton use.

[0093] In one exemplary embodiment, step 106 may include:

[0094] The posture information data is input into the posture estimation model to obtain the motion state information of the target exoskeleton; the posture estimation model obtains the posture pattern of the target object based on the posture information data, and obtains the motion state information of the target exoskeleton based on the posture pattern.

[0095] In one exemplary embodiment, such as Figure 3 As shown, the pose estimation model obtains the pose pattern of the target object based on the pose information data, and acquires the motion state information of the target exoskeleton based on the pose pattern, which may include steps 302 to 304. Wherein:

[0096] Step 302: The second feature extraction module included in the attitude estimation model extracts features from the attitude information data, determines and outputs the attitude feature information of the target object.

[0097] The second feature extraction module may include at least one convolutional layer and at least one pooling layer. The convolutional layer is used to extract features from the input data; the pooling layer is used to sample the input data. Both the convolutional and pooling layers include activation functions. The pose information data may include multiple textual information.

[0098] Specifically, convolutional layers can be used to extract initial features from multiple textual pieces of information. First, multiple textual pieces of information are transformed into vectors, resulting in multiple textual information vectors, which can be combined into a textual information vector matrix. Second, this textual information vector matrix is ​​input into the convolutional layer, where a convolution operation is performed between the convolution kernel and the textual information vector matrix, i.e., the inner product of the textual information vector matrix and the convolution kernel, yielding the convolution result. Next, a nonlinear transformation is applied to the convolution result based on an activation function, and a bias vector is added to obtain the initial feature vector. Third, the initial feature vector is input into a pooling layer for feature sampling. Then, a nonlinear transformation is applied to the feature sampling result based on an activation function, and a bias vector is added to obtain the pose feature information.

[0099] Step 304: The second classification module included in the pose estimation model determines the pose pattern of the target object based on the pose feature information output by the second feature extraction module, and determines and outputs the motion state information of the target exoskeleton based on the pose pattern.

[0100] The second classification module may include at least one fully connected layer, which can classify pose feature information to obtain the motion state information of the target exoskeleton. Additionally, the fully connected layer may include an activation function, which includes a weight matrix and a bias constant.

[0101] Specifically, the pose feature information can be input into the fully connected layer, and the pose feature information can be nonlinearly transformed based on the weight matrix and bias vector of the activation function. Then, through normalization, the motion state information of the target exoskeleton can be obtained.

[0102] In the method of this embodiment, a posture estimation model is set in the exoskeleton system of the target exoskeleton, which enables the exoskeleton to change its posture naturally with the movement and operation of the target object, and keep in sync with the action of the target object, thereby improving comfort and usage efficiency.

[0103] In one exemplary embodiment, step 108 may include:

[0104] Using a pre-built exoskeleton system model of the target exoskeleton, the output torque of the target exoskeleton is obtained based on the working mode information and motion state information.

[0105] Among these methods, Model Predictive Control (MPC) can be used to predict the output torque of the target exoskeleton in the future. First, an exoskeleton system model of the target exoskeleton can be established. This model can be a dynamic model, which describes the behavior of each component of the target exoskeleton (e.g., the multiple joints it contains) based on the kinematics and dynamics principles of the target exoskeleton. In one possible implementation, the dynamic model of the exoskeleton system can be constructed based on the Lagrange equation or the Newton-Euler rule. The target exoskeleton can include multiple joints, and the state of each joint can be represented using spatial state forms, as shown in equations (2) and (3). Furthermore, the output torque of the target exoskeleton can be obtained using this exoskeleton system model based on the working mode information and motion state information.

[0106] In an exemplary embodiment, the exoskeleton system model includes an objective function and constraints; the steps of obtaining the output torque of the target exoskeleton based on the pre-built exoskeleton system model and the working mode information and motion state information may include:

[0107] Under constraints, the objective function is solved based on the working mode information and motion state information to obtain the output torque of the target exoskeleton in the future time period; the future time period includes at least one future time step; the output torque includes a single output torque in each of the at least one future time step.

[0108] In MPC, the motion state of the target exoskeleton in the future can be predicted based on the current state information of the target exoskeleton at each time step, i.e., the state vector. The current state information may include, but is not limited to, the working mode information and motion state information of the target exoskeleton. The working mode information of the target exoskeleton is associated with the load state information of the target object. The future time period, i.e., the future time interval, may include multiple future time steps. In one possible implementation, a rolling time window with a prediction window of 𝑁 can be used to predict the motion state of the target exoskeleton in N future time steps, as shown in equations (4) and (5). Further, the control input of the target exoskeleton in N future time steps, i.e., the joint matrix, can be predicted based on the motion state of the target exoskeleton in N future time steps, i.e., the output vector. In one possible implementation, an objective function of the target exoskeleton can be constructed, and constraints can be set. Under these constraints, the objective function can be solved. By minimizing the objective function, an optimal control input sequence can be obtained. This control input sequence may include the control input of N future time steps, i.e., the output torque. In each prediction cycle, MPC needs to determine the optimal control input sequence by solving an optimization problem. This optimization process is usually solved using numerical optimization algorithms, such as gradient descent or least squares. The result is the optimal control input sequence for a future time period. This sequence can include control inputs for multiple future time steps. Typically, in the multiple future time steps of the current prediction cycle, the control input of the target exoskeleton at the first of these multiple future time steps, i.e., joint torque, can be used. When the exoskeleton system enters the next time step of the current time step, the next prediction cycle can be started, which will yield the control input of the first of the new multiple future time steps, and so on.

[0109] In the method of this embodiment, the torque control module can adjust the output torque of the target exoskeleton in real time according to the working mode information and motion state information of the target exoskeleton, thereby improving the efficiency and safety of the exoskeleton use, and thus improving the efficiency and safety of ultra-high voltage power grid operations.

[0110] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0111] Based on the same inventive concept, this application also provides a lower limb exoskeleton control device for ultra-high voltage power grids, which implements the aforementioned lower limb exoskeleton control method for ultra-high voltage power grids. The solution provided by this device is similar to the implementation described in the above method. Therefore, the specific limitations of one or more embodiments of the lower limb exoskeleton control device for ultra-high voltage power grids provided below can be found in the limitations of the lower limb exoskeleton control method for ultra-high voltage power grids described above, and will not be repeated here.

[0112] In one embodiment, such as Figure 4 As shown, a lower limb exoskeleton control device for ultra-high voltage power grids is provided, comprising: a data acquisition module 402, a first calculation module 404, a second calculation module 406, and a third calculation module 408, wherein:

[0113] Data acquisition module 402 is used to acquire load information data and motion information data of the target object;

[0114] The first calculation module 404 is used to obtain the working mode information of the target exoskeleton that matches the target object based on the load information data using a pre-trained load recognition model.

[0115] The second calculation module 406 is used to obtain the motion state information of the target exoskeleton based on the posture information data contained in the motion information data by using a pre-trained posture estimation model.

[0116] The third calculation module 408 is used to obtain the output torque of the target exoskeleton based on the working mode information and the motion state information, and to adjust the running state of the target exoskeleton using the output torque.

[0117] In one embodiment, the first calculation module 404 is further configured to: input the load information data into the load recognition model to obtain the working mode information of the target exoskeleton; wherein the load recognition model obtains the load tag of the target object based on the load information data, and obtains the working mode information of the target exoskeleton based on the load tag.

[0118] In one embodiment, the first calculation module 404 is further configured to: extract features from the load information data by a first feature extraction module included in the load recognition model, determine and output the load feature information of the target object; and, based on the load feature information output by the first feature extraction module, determine the load label of the target object by a first classification module included in the load recognition model, and determine and output the working mode information of the target exoskeleton based on the load label.

[0119] In one embodiment, the second calculation module 406 is further configured to: input the posture information data into the posture estimation model to obtain the motion state information of the target exoskeleton; wherein the posture estimation model obtains the posture pattern of the target object based on the posture information data, and obtains the motion state information of the target exoskeleton based on the posture pattern.

[0120] In one embodiment, the second calculation module 406 is further configured to: extract features from the posture information data by the second feature extraction module included in the posture estimation model, determine and output the posture feature information of the target object; and, based on the posture feature information output by the second feature extraction module, determine the posture pattern of the target object by the second classification module included in the posture estimation model, and determine and output the motion state information of the target exoskeleton based on the posture pattern.

[0121] In one embodiment, the third calculation module 408 is further configured to: obtain the output torque of the target exoskeleton based on the working mode information and the motion state information using a pre-constructed exoskeleton system model of the target exoskeleton system.

[0122] In one embodiment, the third calculation module 408, wherein the exoskeleton system model includes an objective function and constraints, is further configured to: under the constraints, solve the objective function based on the working mode information and the motion state information to obtain the output torque of the target exoskeleton in a future time period; the future time period includes at least one future time step; the output torque includes a single output torque for each of at least one of the future time steps.

[0123] The modules in the aforementioned lower limb exoskeleton control device for ultra-high voltage power grids can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0124] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 5 As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database stores data related to lower limb exoskeleton control for ultra-high voltage power grids. The network interface communicates with external terminals via a network connection. When executed by the processor, the computer program implements a lower limb exoskeleton control method for ultra-high voltage power grids.

[0125] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0126] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0127] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0128] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0129] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0130] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0131] The technical features of the above embodiments can be combined in any way. For the sake of brevity, 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.

[0132] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for controlling a lower limb exoskeleton for ultra-high voltage power grids, characterized in that, The method includes: Acquire load and motion information data of the target object; Using a pre-trained load recognition model, based on the load information data, the working mode information of the target exoskeleton matching the target object is obtained; wherein, the load information data is input into the load recognition model to obtain the working mode information of the target exoskeleton; wherein, the first feature extraction module included in the load recognition model extracts features from the load information data to determine and output the load feature information of the target object; the first classification module included in the load recognition model determines the load label of the target object based on the load feature information output by the first feature extraction module, and determines and outputs the working mode information of the target exoskeleton based on the load label; Using a pre-trained pose estimation model, the motion state information of the target exoskeleton is obtained based on the pose information data contained in the motion information data; wherein, the pose information data is input into the pose estimation model to obtain the motion state information of the target exoskeleton; wherein, the pose estimation model obtains the pose pattern of the target object based on the pose information data, and obtains the motion state information of the target exoskeleton based on the pose pattern. Based on the working mode information and the motion state information, the output torque of the target exoskeleton is obtained, and the operating state of the target exoskeleton is adjusted using the output torque.

2. The method according to claim 1, characterized in that, The step of obtaining the pose pattern of the target object from the pose estimation model based on the pose information data, and obtaining the motion state information of the target exoskeleton based on the pose pattern, includes: The second feature extraction module included in the attitude estimation model performs feature extraction on the attitude information data, determines and outputs the attitude feature information of the target object; The second classification module included in the posture estimation model determines the posture pattern of the target object based on the posture feature information output by the second feature extraction module, and determines and outputs the motion state information of the target exoskeleton based on the posture pattern.

3. The method according to claim 1, characterized in that, The step of obtaining the output torque of the target exoskeleton based on the working mode information and the motion state information includes: Using a pre-constructed exoskeleton system model of the target exoskeleton, the output torque of the target exoskeleton is obtained based on the working mode information and the motion state information.

4. The method according to claim 3, characterized in that, The exoskeleton system model includes an objective function and constraints; The exoskeleton system model utilizing the pre-constructed exoskeleton system of the target exoskeleton, based on the working mode information and the motion state information, obtains the output torque of the target exoskeleton, including: Under the constraints, the objective function is solved based on the working mode information and the motion state information to obtain the output torque of the target exoskeleton in a future time period; the future time period includes at least one future time step; the output torque includes a single output torque for each of at least one of the future time steps.

5. A lower limb exoskeleton control device for ultra-high voltage power grids, characterized in that, The device includes: The data acquisition module is used to acquire load information data and motion information data of the target object; The first calculation module is used to obtain the working mode information of the target exoskeleton matching the target object based on the load information data using a pre-trained load recognition model. Specifically, the load information data is input into the load recognition model to obtain the working mode information of the target exoskeleton. The first feature extraction module included in the load recognition model extracts features from the load information data to determine and output the load feature information of the target object. The first classification module included in the load recognition model determines the load label of the target object based on the load feature information output by the first feature extraction module, and determines and outputs the working mode information of the target exoskeleton based on the load label. The second calculation module is used to obtain the motion state information of the target exoskeleton based on the posture information data contained in the motion information data using a pre-trained posture estimation model; wherein, the posture information data is input into the posture estimation model to obtain the motion state information of the target exoskeleton; wherein, the posture estimation model obtains the posture pattern of the target object based on the posture information data, and obtains the motion state information of the target exoskeleton based on the posture pattern; The third calculation module is used to obtain the output torque of the target exoskeleton based on the working mode information and the motion state information, and to adjust the running state of the target exoskeleton using the output torque.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1-4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method described in any one of claims 1-4.

Images