Control method and system of lower limb exoskeleton power assisting device for iron tower climbing

Abstract

The invention discloses a control method and system of a lower limb exoskeleton power assisting device for iron tower climbing. The method comprises the steps that electromyographic signals are processed, and muscle activation degree indexes are calculated; the two-channel LSTM neural network realizes motion intention online prediction; iron tower cross arm structure edge identification and safe climbing path generation; designing an asymmetric impedance function and optimizing joint stiffness damping parameters; solving a metabolic consumption evaluation model and a minimum energy consumption assistance curve; and real-time calculation is realized through three-stage control period setting and hardware acceleration. Based on the method, all the modules of the exoskeleton power assisting device operate cooperatively; data are accurately collected through the sensor, the power and execution module is flexibly adjusted according to instructions, efficient operation is ensured through the control and operation module, stable electric power is provided through the energy module, and therefore the safety and efficiency of iron tower climbing operation are remarkably improved, and the burden of operators is relieved.

Description

Technical Field

[0001] The present invention relates to the technical field of exoskeleton assist device control, and particularly to a control method and system for a lower limb exoskeleton assist device used for tower climbing. Background Art

[0002] With the growth of the maintenance requirements of industrial and power facilities, the safety and efficiency of tower climbing operations have received increasing attention. As an important equipment for enhancing the climbing operation ability, the technical development of lower limb exoskeleton assist devices is crucial.

[0003] Traditional lower limb exoskeletons mostly rely on single-type sensor data, such as only using surface electromyography sensors or plantar pressure sensors. Relying solely on surface electromyography signals is vulnerable to external electromagnetic interference and difficult to comprehensively reflect the motion state; relying solely on plantar pressure data cannot accurately capture the subtle changes in muscles, resulting in a low accuracy rate of motion intention recognition. This makes the exoskeleton unable to accurately provide assistance matching the user's actions during climbing operations, with unsmooth climbing actions and low efficiency, seriously affecting the operation progress.

[0004] Existing exoskeletons lack a comprehensive perception and effective response mechanism for complex tower environments. Most devices are not equipped with high-precision environmental perception devices such as lidar, making it difficult to accurately identify key environmental information such as the edges of tower cross arms, and unable to plan a safe climbing path in advance. In terms of joint control, a fixed control mode is adopted, and the joint stiffness and damping parameters cannot be adjusted in real time according to the forces in different directions and the motion stages during climbing. When encountering sudden situations or complex climbing conditions, the safety of the user cannot be guaranteed in a timely manner, increasing the operation risk.

[0005] Traditional exoskeletons lack an effective energy consumption optimization algorithm and cannot dynamically adjust the assistance output according to the user's real-time motion state and operation requirements. During long-term tower climbing operations, excessive energy consumption leads to insufficient battery life, and frequent charging or battery replacement seriously affects the continuity and efficiency of operations. Moreover, traditional devices lack a reasonable power distribution strategy for battery power, further exacerbating the energy consumption problem.

[0006] In view of the many deficiencies of traditional lower limb exoskeletons in motion intention recognition, safety guarantee, and energy consumption management, there is an urgent need for an innovative lower limb exoskeleton assist device and algorithm to improve the safety, efficiency, and battery life of tower climbing operations and meet the actual operation requirements. Summary of the Invention

[0007] In view of the above problems, the present invention is proposed to provide a control method and system for a lower limb exoskeleton assist device used for tower climbing that overcomes the above problems or at least partially solves the above problems.

[0008] To solve the above technical problems, the embodiments of the present application disclose the following technical solutions:

[0009] In a first aspect, an embodiment of the present invention discloses a control method for a lower limb exoskeleton assistive device for tower climbing, including:

[0010] Real-time collect the myoelectric signal and plantar pressure distribution data of the user through surface electromyography sensors and plantar pressure sensors;

[0011] Decompose the myoelectric signal, calculate the muscle activation index; construct a two-channel LSTM neural network, and realize the online prediction of the user's motion intention through the two-channel LSTM neural network;

[0012] Collect tower point cloud data, based on the tower point cloud data, identify the edges of the tower cross arm structure, generate a safe climbing path through cubic spline interpolation, calculate the minimum distance threshold between the user's foot end reachable workspace and obstacles, and establish a gradient sensitivity model to assist in motion safety decision-making;

[0013] Design an asymmetric impedance function, use a model predictive control algorithm to generate the desired joint trajectory in the upper layer of the gradient sensitivity model, use an improved particle swarm optimization algorithm to optimize the joint stiffness and damping parameters in real time in the lower layer of the gradient sensitivity model, and introduce a nonlinear disturbance observer to compensate for the hysteresis effect of the magnetorheological actuator;

[0014] Establish a metabolic consumption assessment model, use a rolling horizon optimization algorithm to solve the minimum energy consumption assistive curve, and set a safety threshold constraint to limit the overshoot of joint power;

[0015] Set a three-level control period, including the actuator current loop control in the first level, the impedance parameter adjustment in the second level, and the path planning update in the third level, and implement real-time calculation through the FPGA and GPU hardware acceleration modules.

[0016] Further, decompose the myoelectric signal, calculate the muscle activation index, and the specific method includes: collect the myoelectric signals of the main muscle groups of the lower limbs during the user's climbing through surface electromyography sensors, decompose the myoelectric signal x(t) to obtain the intrinsic mode function IMF i (t), i = 1, 2, 3 ···, n, select the first 6 intrinsic mode functions; use the improved Teager energy operator TKO(x[n]) = x 2 [n] - x[n - 1]x[n + 1], calculate the energy of each intrinsic mode, and obtain the muscle activation index by weighted summation, where x[n] is the myoelectric signal sampling sequence.

[0017] Furthermore, a dual-channel LSTM neural network is constructed to achieve online prediction of the user's motion intention through the dual-channel LSTM neural network. The specific method includes: constructing a dual-channel LSTM neural network, where the first channel processes muscle activation metrics to predict the motion direction of the user's lower limb joints, and the second channel analyzes the plantar pressure distribution data to determine the climbing stage through the plantar pressure distribution data, and dynamically weights multi-sensor data through an attention mechanism to achieve online prediction of the user's motion intention.

[0018] Furthermore, dynamically weighting multi-sensor data through an attention mechanism, the specific method includes: applying an attention mechanism, where the output feature of the first channel is F1, the output feature of the second channel is F2, and the weights α1 and α2 are calculated by the attention model, α1 + α2 = 1, to obtain the fused feature F = α1F1 + α2F2.

[0019] Furthermore, lidar installed on the exoskeleton is used to collect tower point cloud data. By improving the RANSAC algorithm, a straight-line model is randomly sampled and hypothesized in the point cloud data, the distance from other points to the model is calculated to determine inliers, and the edge of the tower cross-arm structure is iteratively found; using the starting point and the target climbing point as endpoints, a safe climbing path is generated by cubic spline interpolation, and based on the exoskeleton structure and kinematic parameters, the minimum distance threshold between the foot end reachable workspace and obstacles is calculated, and a gradient sensitivity model is established to assist in safety decision-making.

[0020] Furthermore, the expression of the asymmetric impedance function is K d 、B d are asymmetric stiffness and damping coefficients, and K d 、B d are adjusted according to different motion directions and climbing stages. The model predictive control algorithm predicts the system response based on the exoskeleton dynamics model and the current state, and solves the optimization problem to generate the desired joint trajectory; the improved particle swarm algorithm iteratively optimizes the joint stiffness and damping parameters as particle positions, and at the same time introduces a nonlinear disturbance observer to compensate for the hysteresis effect of the magnetorheological actuator.

[0021] Furthermore, the expression of the metabolic consumption evaluation model is P m =f(v,a,F m ,A), where v is the user's motion speed v, a is the acceleration a, F m is the exoskeleton assistance F m 、A is the muscle activation metric, and P m is the energy consumption; the rolling horizon optimization algorithm is used to predict the future system state starting from the current moment, solve the exoskeleton assistance curve with the minimum metabolic consumption, and at the same time set safety thresholds to constrain the joint power.

[0022] Furthermore, the actuator current loop control of the first stage is the actuator current loop control at 1 kHz, and the magnetic rheological actuator current is precisely controlled through the 1 kHz frequency control to regulate the exoskeleton output force; the impedance parameter adjustment of the second stage is the impedance parameter adjustment at 100 Hz, and the impedance parameter adjustment at 100 Hz adjusts the joint stiffness damping parameters according to the motion state and intention; the path planning of the third stage is updated to the path planning update at 10 Hz, and the path planning update at 10 Hz updates the safe climbing path according to the lidar data and the user state; the FPGA is used to implement the sensor data acquisition logic operation, and the GPU is used to process the large-scale matrix operation of the LSTM neural network calculation.

[0023] In a second aspect, an embodiment of the present invention discloses a control system for a lower limb exoskeleton assist device for tower climbing, including: an electromyogram signal and plantar pressure distribution data acquisition unit, an online prediction unit for motion intention, a tower cross-arm structure edge and safe climbing path generation unit, an asymmetric impedance function design and joint stiffness damping parameter optimization unit, a metabolic consumption evaluation model and minimum energy consumption assist curve solving unit, and a three-level control period setting and hardware acceleration unit; wherein:

[0024] The electromyogram signal and plantar pressure distribution data acquisition unit is used to collect the user's electromyogram signal and plantar pressure distribution data in real time through surface electromyogram sensors and plantar pressure sensors;

[0025] The online prediction unit for motion intention is used to decompose the electromyogram signal to calculate the muscle activation degree index; a two-channel LSTM neural network is constructed, and the online prediction of the user's motion intention is realized through the two-channel LSTM neural network;

[0026] The tower cross-arm structure edge and safe climbing path generation unit is used to decompose the electromyogram signal to calculate the muscle activation degree index; a two-channel LSTM neural network is constructed, and the online prediction of the user's motion intention is realized through the two-channel LSTM neural network;

[0027] The asymmetric impedance function design and joint stiffness damping parameter optimization unit is used to design an asymmetric impedance function, adopt a model predictive control algorithm to generate a desired joint trajectory in the upper layer of the gradient sensitivity model, and use an improved particle swarm algorithm to optimize the joint stiffness damping parameters in real time in the lower layer of the gradient sensitivity model, and introduce a nonlinear disturbance observer to compensate for the hysteresis effect of the magnetic rheological actuator;

[0028] The metabolic consumption evaluation model and minimum energy consumption assist curve solving unit is used to establish a metabolic consumption evaluation model, adopt a rolling horizon optimization algorithm to solve the minimum energy consumption assist curve, and set a safety threshold constraint to limit the joint power overshoot;

[0029] Three - level control cycle setting and hardware acceleration unit, which is used to set the three - level control cycle, including the actuator current loop control at the first level, the impedance parameter adjustment at the second level, and the path planning update at the third level, and realizes real - time calculation through the FPGA and GPU hardware acceleration modules.

[0030] In a third aspect, embodiments of the present invention disclose an electronic device, including:

[0031] One or more processors;

[0032] A memory for storing one or more programs;

[0033] When the one or more programs are executed by the one or more processors, the one or more processors implement the control method of the lower - limb exoskeleton assist device.

[0034] The beneficial effects of the above - mentioned technical solutions provided by the embodiments of the present invention at least include:

[0035] A control method and system for a lower - limb exoskeleton assist device for tower climbing disclosed by the present invention, by integrating surface electromyogram signals, plantar pressure data and lidar point cloud data, adaptive noise complete ensemble empirical mode decomposition algorithm, improved Teager energy operator, and dual - channel LSTM neural network with attention mechanism, comprehensively and accurately perceives the user's motion intention and environmental information, breaks through the limitation of single - data processing, makes the motion intention recognition more accurate, and makes the climbing action smoother, more efficient and more comfortable. In the dynamic adaptive control algorithm, an asymmetric impedance function is designed. The upper - layer MPC generates the desired joint trajectory and adjusts the strategy in real - time according to the actual situation. The lower - layer improved particle swarm algorithm optimizes the joint stiffness and damping parameters, realizes deep coordination with the hardware power output, flexibly adapts to complex climbing working conditions, and ensures safety. The hardware acceleration and real - time optimization algorithm uses the FPGA and GPU hardware acceleration modules, sets a three - level control cycle, and uses the RHO algorithm to solve the minimum - energy assist curve, improving the real - time performance and energy - consumption management efficiency of the system, and extending the battery life to meet the long - time operation requirements.

[0036] The following further describes the technical solutions of the present invention in detail through the drawings and embodiments. Description of the Drawings

[0037] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention, and do not constitute a limitation to the present invention. In the drawings:

[0038] Figure 1 It is a flowchart of a control method for a lower - limb exoskeleton assist device for tower climbing in Embodiment 1 of the present invention;

[0039] Figure 2 This is the structural diagram of a control system for a lower limb exoskeleton assistive device for tower climbing in Embodiment 3 of the present invention;

[0040] Figure 3 This is the schematic structural diagram of an electronic device in Embodiment 4 of the present invention. Detailed implementation manners

[0041] Hereinafter, exemplary embodiments of the present disclosure will be described in more detail with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be fully communicated to those skilled in the art.

[0042] In order to solve the problems existing in the prior art, the embodiments of the present invention provide a control method and system for a lower limb exoskeleton assistive device for tower climbing.

[0043] Embodiment 1

[0044] The present invention discloses a control method for a lower limb exoskeleton assistive device for tower climbing, as Figure 1 , including:

[0045] S100. Real-time collect the myoelectric signal and plantar pressure distribution data of the user through surface electromyography sensors and plantar pressure sensors;

[0046] The surface electromyography sensors and plantar pressure sensors provide key data support for motion analysis, rehabilitation evaluation, and human-computer interaction by real-time collecting biomechanical signals. Among them, the surface electromyography sensors record the potential changes generated during muscle contraction through electrodes attached to the skin, reflecting the neuromuscular activity state. The plantar pressure sensors measure the pressure values of each area of the sole using piezoelectric films or capacitive sensor arrays to generate a dynamic pressure distribution map.

[0047] S200. Decompose the myoelectric signal, calculate the muscle activation degree index; construct a dual-channel LSTM neural network, and realize the online prediction of the user's motion intention through the dual-channel LSTM neural network.

[0048] In this embodiment, when decomposing the myoelectric signal to calculate the muscle activation degree index, the specific method includes: collecting the myoelectric signals of the main muscle groups of the lower limbs during the user's climbing through surface electromyography sensors, decomposing the myoelectric signal x(t) to obtain the intrinsic mode functions IMF i (t), i = 1, 2, 3 ···, n, and selecting the first 6 intrinsic mode functions; adopting the improved Teager energy operator TKO(x[n]) = x 2[n] - x[n - 1]x[n + 1], calculate the energy of each eigenmode, and obtain the muscle activation index through weighted summation, where x[n] is the sampling sequence of the EMG signal. It can be understood that the present invention collects the EMG signal and the plantar pressure distribution data of the user in real time, and realizes the prediction of the user's motion intention through a dual-channel LSTM neural network. Specifically, during the process of climbing the iron tower, the user's climbing operation process mainly includes the climbing start stage, the climbing process stage, and the climbing rest and adjustment stage.

[0049] In this embodiment, the multi-sensor data is dynamically weighted through an attention mechanism. The specific method includes: applying the attention mechanism, where the output feature of the first channel is F1, and the output feature of the second channel is F2. The weights α1 and α2 are calculated by the attention model, and α1 + α2 = 1, and the fused feature F = α1F1 + α2F2 is obtained.

[0050] S300. Collect the point cloud data of the iron tower. Based on the point cloud data of the iron tower, identify the edge of the cross-arm structure of the iron tower, generate a safe climbing path through cubic spline interpolation, and calculate the minimum distance threshold between the reachable working space of the user's foot end and the obstacle, and establish a gradient sensitivity model to assist in the motion safety decision-making;

[0051] In S300 of this embodiment, the lidar installed on the exoskeleton is used to collect the point cloud data of the iron tower. Through an improved RANSAC algorithm, a straight line model is randomly sampled and hypothesized in the point cloud data, and the inliers are determined by calculating the distance from other points to the model. Iteratively find the edge of the cross-arm structure of the iron tower; use cubic spline interpolation with the starting point and the target climbing point as endpoints to generate a safe climbing path. According to the exoskeleton structure and kinematic parameters, calculate the minimum distance threshold between the reachable working space of the foot end and the obstacle, and establish a gradient sensitivity model to assist in the safety decision-making.

[0052] Specifically, the improved RANSAC algorithm is used for straight line model detection. The clustering algorithm (such as DBSCAN or K-means) is used to cluster the point cloud data to reduce the influence of noise and outliers; sample points are randomly selected in each cluster for model estimation to increase the probability of finding inliers; the least squares method is used to fit a straight line to the selected sample points; calculate the distance from other points to the fitted straight line, and determine the inliers according to the set threshold; iterate and optimize, repeat the above process, and select the model with the most inliers as the best model. During the iteration process, the distance threshold is dynamically adjusted according to the proportion of the current inliers to improve the robustness and efficiency of the algorithm.

[0053] In this embodiment, cubic spline interpolation is used to generate a safe climbing path. The specific method includes: determining the end points, taking the starting point and the target climbing point as the end points of the cubic spline interpolation; performing interpolation calculation, performing cubic spline interpolation between the end points to generate a smooth climbing path; optimizing the path, and optimizing the path according to the kinematic parameters of the exoskeleton to ensure its reachability and safety.

[0054] By improving the RANSAC algorithm, the straight edges of the tower cross-arm structure can be detected more efficiently and robustly. Combining the safe climbing path generated by cubic spline interpolation and the calculation of the minimum distance threshold between the reachable workspace of the exoskeleton foot end and the obstacle can ensure the safe movement of the exoskeleton in a complex environment. This method not only improves the efficiency and safety of path planning, but also provides reliable technical support for the intelligent navigation of the exoskeleton.

[0055] S400. Design an asymmetric impedance function, use a model predictive control algorithm to generate a desired joint trajectory in the upper layer of the gradient sensitivity model, and use an improved particle swarm algorithm to optimize the joint stiffness and damping parameters in real time in the lower layer of the gradient sensitivity model, and introduce a nonlinear disturbance observer to compensate for the hysteresis effect of the magnetorheological actuator;

[0056] In S400 of this embodiment, the expression of the asymmetric impedance function is K d 、B d are asymmetric stiffness and damping coefficients, and K d 、B d are adjusted according to different motion directions and climbing stages. The model predictive control algorithm predicts the system response based on the exoskeleton dynamics model and the current state, and solves the optimization problem to generate a desired joint trajectory; the improved particle swarm algorithm iteratively optimizes the joint stiffness and damping parameters as the particle positions, and at the same time introduces a nonlinear disturbance observer to compensate for the hysteresis effect of the magnetorheological actuator.

[0057] In this embodiment, the model predictive control generates a desired joint trajectory. The specific method includes:

[0058] Construct a state space model; establish the kinematic and dynamic models of the joints and represent them in the state space form. Use a linear or nonlinear model and select an appropriate model form according to the system characteristics.

[0059] Predict and control the model; at each time step, predict the system state in the future for a period of time. Solve the optimal control input through an optimization algorithm (such as quadratic programming) to minimize the error between the predicted state and the desired state.

[0060] Generate the desired trajectory, and generate the desired joint trajectory through the MPC algorithm to ensure that the trajectory meets the dynamic constraints and performance requirements.

[0061] In this embodiment, the improved particle swarm optimization algorithm is used to optimize the joint stiffness and damping parameters. The specific method includes:

[0062] Improve the particle swarm optimization algorithm; adjust the adaptive inertia weight, introduce the adaptive inertia weight, and dynamically adjust the inertia weight according to the convergence of the particle swarm to improve the global and local search capabilities of the algorithm. Improve the particle update strategy, introduce the crossover and mutation operations in the chaotic mapping or genetic algorithm to enhance the exploration ability of the algorithm.

[0063] Real-time optimize the joint stiffness and damping parameters; in each control cycle, use the improved particle swarm optimization algorithm to real-time optimize the joint stiffness and damping parameters. The optimization objectives include minimizing the tracking error, energy consumption or other performance indicators.

[0064] By designing an asymmetric impedance function, combining MPC to generate the desired joint trajectory, using the improved PSO algorithm to real-time optimize the joint parameters, and introducing NDO to compensate for the hysteresis effect, the dynamic performance and robustness of the system can be significantly improved. This method not only optimizes the control strategy but also improves the adaptability and stability of the system in complex environments.

[0065] S500. Establish a metabolic consumption evaluation model, use the rolling horizon optimization algorithm to solve the minimum energy consumption assistance curve, and set safety threshold constraints to limit the joint power overshoot;

[0066] In S500 of this embodiment, the expression of the metabolic consumption evaluation model is P m = f(v, a, F m , A), where v is the user's movement speed v, a is the acceleration a, F m is the exoskeleton assistance F m , A is the muscle activation index, and P m is the energy consumption; use the rolling horizon optimization algorithm to predict the future system state starting from the current moment, solve the exoskeleton assistance curve with the minimum metabolic consumption, and at the same time set safety threshold constraints for the joint power.

[0067] By establishing a metabolic consumption evaluation model, combining the rolling horizon optimization algorithm to solve the minimum energy consumption assistance curve, and setting safety threshold constraints to limit the joint power overshoot, efficient and safe motion control can be achieved. This method not only optimizes the energy consumption but also improves the stability and reliability of the system, and is applicable to various application scenarios such as exoskeleton robots.

[0068] S600. Set three-level control cycles, including the actuator current loop control at the first level, the impedance parameter adjustment at the second level, and the path planning update at the third level, and implement real-time calculation through the FPGA and GPU hardware acceleration modules.

[0069] In S600 of this embodiment, the actuator current loop control of the first stage is the actuator current loop control at 1 kHz, and the magnetic rheological actuator current is precisely regulated through the 1 kHz frequency control to regulate the exoskeleton output force; the impedance parameter adjustment of the second stage is the impedance parameter adjustment at 100 Hz, and the impedance parameter adjustment at 100 Hz adjusts the joint stiffness damping parameters according to the motion state and intention; the path planning of the third stage is updated to the path planning update at 10 Hz, and the path planning update at 10 Hz updates the safe climbing path according to the lidar data and the user state; the FPGA is used to implement the logical operation of sensor data acquisition, and the GPU is used to process the large-scale matrix operation of the LSTM neural network calculation.

[0070] By setting three control periods and using the FPGA and GPU hardware acceleration modules to achieve real-time calculation, the response speed and control accuracy of the system can be significantly improved. This design not only optimizes the performance of the control system, but also improves the reliability and stability of the system, and is suitable for application scenarios that require high precision and real-time response.

[0071] A control method for a lower limb exoskeleton assist device for tower climbing disclosed in this embodiment comprehensively and accurately perceives the user's motion intention and environmental information by integrating surface electromyography signals, plantar pressure data, and lidar point cloud data, the adaptive noise complete ensemble empirical mode decomposition algorithm, the improved Teager energy operator, and the dual-channel LSTM neural network with an attention mechanism, breaks through the limitation of single data processing, makes the motion intention recognition more accurate, and makes the climbing action smoother, more efficient, and more comfortable. In the dynamic adaptive control algorithm, an asymmetric impedance function is designed. The upper-layer MPC generates the desired joint trajectory and adjusts the strategy in real time according to the actual situation. The lower-layer improved particle swarm algorithm optimizes the joint stiffness damping parameters to achieve deep coordination with the hardware power output, flexibly adapts to complex climbing conditions, and ensures safety. The hardware acceleration and real-time optimization algorithm uses the FPGA and GPU hardware acceleration modules, sets three control periods, and uses the RHO algorithm to solve the minimum energy consumption assist curve, improving the real-time performance and energy consumption management efficiency of the system and extending the battery life to meet the long-time operation requirements.

[0072] Embodiment 2

[0073] To better understand the control method for a lower limb exoskeleton assist device for tower climbing disclosed in Embodiment 1, this embodiment publicly applies the method disclosed in Embodiment 1 to a specific tower maintenance and repair scenario.

[0074] First, the application scenario and user profile are introduced. Assume that a power company needs to perform maintenance and repair work on an old tower. The tower is 50 meters high and has a complex structure. An operator is responsible for this climbing operation and is wearing the lower limb exoskeleton assist device of the present invention.

[0075] The operator is then introduced to the donning and initial setup of the device. The donning process begins with the operator standing next to a specially designed donning frame. The operator first aligns the exoskeleton's legs with their lower limbs, adjusting the straps to ensure a snug and comfortable fit at the hip, knee, and ankle joints. The straps are elastic and adjustable to accommodate various body types. Next, the waist support is secured to ensure a secure fit and overall support. The entire donning process takes approximately three minutes.

[0076] Initial parameter setup: Power on the device and perform initial setup using the handheld control terminal. Based on the operator's physical parameters (assuming a height of 175 cm and a weight of 75 kg), set the basic dimensions and adaptation parameters of the exoskeleton. Simultaneously, calibrate the surface electromyography (EMG) sensors and plantar pressure sensors to ensure accurate data collection. For example, the surface electromyography (EMG) sensors must be tightly attached to the surface of key muscle groups, such as the quadriceps and hamstrings, and signal transmission is enhanced by applying a special conductive gel.

[0077] Next, the method disclosed in Example 1 is used to introduce the climbing operation process of the operator; the entire climbing operation process is divided into the starting stage, the climbing process stage, and the rest and adjustment stage; specifically:

[0078] Starting phase: The worker approaches the tower and grasps the climbing ladder with both hands. At this point, surface electromyography sensors detect the initial contraction signals of the quadriceps and calf muscles. The CEEMDAN algorithm decomposes these signals to extract key IMF features, which are then combined with TKO to calculate muscle activation metrics. Simultaneously, plantar pressure sensors detect the pressure distribution of both feet. Channel 1 of the dual-channel LSTM neural network processes the electromyographic features to predict the worker's intention to climb upward, while channel 2 analyzes the trajectory of the plantar pressure center to determine the worker is in the starting phase. The attention mechanism integrates this information, allowing the device to determine that the worker is ready to start climbing. The upper-layer MPC algorithm generates the initial expected joint trajectory based on the pre-set safe climbing strategy. The lower-layer improved particle swarm algorithm rapidly optimizes the joint stiffness and damping parameters, enabling the exoskeleton to gently assist the worker in lifting their leg and taking the first step.

[0079] During the climbing process: When the operator climbs to about 10 meters, he encounters an area with a relatively large distance between cross arms. The lidar scans the surrounding environment in real time, and the point cloud data quickly identifies the edges of the tower cross arm structure through an improved RANSAC algorithm. Based on this, the device generates a safe climbing path using cubic spline interpolation. The safe climbing path is the foot contact point for each step of the operator during the climbing process along the tower. A number of consecutive foot contact points form the safe climbing path. During this process, the body posture of the operator changes, and the surface electromyogram signal and plantar pressure data also change accordingly. The multimodal data fusion algorithm continuously and accurately perceives these changes, and the dynamic adaptive control algorithm adjusts the joint stiffness and damping parameters in a timely manner. For example, when the operator needs to bypass an obstacle sideways, the hip and knee joints of the exoskeleton dynamically adjust the stiffness and damping according to the algorithm instructions, providing appropriate assistance and support to ensure smooth and safe movements. At the same time, the hardware acceleration module ensures the rapid operation of the algorithm. The 1kHz actuator current loop control enables the magnetorheological actuator to respond quickly, the 100Hz impedance parameter adjustment ensures comfortable joint movement, and the 10Hz path planning update fine-tunes the climbing path according to the real-time environmental changes.

[0080] Rest and adjustment stage: When climbing to 25 meters, the operator takes a short rest. At this time, the metabolic consumption assessment model estimates the current energy consumption based on the operator's motion state (speed, acceleration, etc.), the assistance output of the exoskeleton, and the muscle activation index. The rolling horizon optimization algorithm adjusts the subsequent assistance curve according to this assessment result to achieve minimum energy consumption assistance. At the same time, the battery management system monitors the battery power in real time and distributes the electric energy reasonably to ensure sufficient power for the subsequent climbing. For example, if it is found that the power of a certain battery is consumed quickly, the system will automatically adjust the current distribution and give priority to using the battery with sufficient power to extend the overall endurance time.

[0081] After the operator climbs onto the tower, the climbing operation ends, and the operator enters the descending process. After the operator completes the maintenance work at the top of the tower, he starts to descend. The device judges the descending intention based on the surface electromyogram signal and plantar pressure data and dynamically adjusts the control strategy. For example, the joint stiffness and damping parameters are adjusted to a mode more suitable for descending to provide stable support and buffering. During the descending process, the lidar continuously monitors the surrounding environment to ensure safety.

[0082] After returning to the ground, the operator unties the straps of the exoskeleton device and completes the unloading. The device is placed in a dedicated maintenance area, and technicians check it. The inspection contents include whether the sensors are damaged, whether the battery power is sufficient, and whether there is wear on each joint component. For example, the signal acquisition accuracy of the surface electromyogram sensor is checked through professional detection equipment. If the accuracy is found to decrease, the sensor is replaced or calibrated in a timely manner. The battery is charged and maintained according to the usage situation to ensure that the device can operate normally the next time it is used.

[0083] Example 3

[0084] Based on the same inventive concept, an embodiment of the present disclosure also provides a control system for a lower limb exoskeleton assistive device for tower climbing, as Figure 2 , including: an electromyogram signal and plantar pressure distribution data acquisition unit, an online prediction unit of motion intention, a tower cross-arm structure edge and safe climbing path generation unit, an asymmetric impedance function design and joint stiffness damping parameter optimization unit, a metabolic consumption evaluation model and minimum energy consumption assist curve solving unit, and a three-level control period setting and hardware acceleration unit; where:

[0085] The electromyogram signal and plantar pressure distribution data acquisition unit is used to collect the user's electromyogram signal and plantar pressure distribution data in real time through surface electromyogram sensors and plantar pressure sensors;

[0086] The online prediction unit of motion intention is used to decompose the electromyogram signal, calculate the muscle activation degree index; construct a two-channel LSTM neural network, and realize the online prediction of the user's motion intention through the two-channel LSTM neural network;

[0087] The tower cross-arm structure edge and safe climbing path generation unit is used to decompose the electromyogram signal, calculate the muscle activation degree index; construct a two-channel LSTM neural network, and realize the online prediction of the user's motion intention through the two-channel LSTM neural network;

[0088] The asymmetric impedance function design and joint stiffness damping parameter optimization unit is used to design an asymmetric impedance function, adopt a model predictive control algorithm to generate an expected joint trajectory in the upper layer of the gradient sensitivity model, and optimize the joint stiffness damping parameters in real time through an improved particle swarm algorithm in the lower layer of the gradient sensitivity model, and introduce a nonlinear disturbance observer to compensate for the hysteresis effect of the magnetorheological actuator;

[0089] The metabolic consumption evaluation model and minimum energy consumption assist curve solving unit is used to establish a metabolic consumption evaluation model, solve the minimum energy consumption assist curve by using a rolling horizon optimization algorithm, and set a safety threshold constraint to limit the joint power overshoot;

[0090] The three-level control period setting and hardware acceleration unit is used to set a three-level control period, including the actuator current loop control in the first level, the impedance parameter adjustment in the second level, and the path planning update in the third level, and realize real-time calculation through the FPGA and GPU hardware acceleration modules.

[0091] Example 4

[0092] Based on the same inventive concept, an embodiment of the present disclosure also provides an electronic device. Figure 3The following is a schematic structural diagram of an electronic device according to an embodiment of the present disclosure. As Figure 3 shown, an electronic device provided by an embodiment of the present disclosure includes: one or more processors 101, a memory 102, and one or more I / O interfaces 103. One or more programs are stored on the memory 102. When the one or more programs are executed by the one or more processors, the one or more processors implement any of the optimization methods in the above embodiments; one or more I / O interfaces 103 are connected between the processor and the memory and are configured to implement information interaction between the processor and the memory.

[0093] Among them, the processor 101 is a device with data processing capabilities, including but not limited to a central processing unit (CPU), etc.; the memory 102 is a device with data storage capabilities, including but not limited to a random access memory (RAM, more specifically SDRAM, DDR, etc.), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), and a flash memory (FLASH); the I / O interface (read / write interface) 103 is connected between the processor 101 and the memory 102 and can implement information interaction between the processor 101 and the memory 102, including but not limited to a data bus (Bus), etc.

[0094] In some embodiments, the processor 101, the memory 102, and the I / O interface 103 are interconnected through a bus 104 and are further connected to other components of the computing device.

[0095] In some embodiments, the one or more processors 101 include a field programmable gate array.

[0096] According to an embodiment of the present disclosure, a computer-readable medium is also provided. A computer program is stored on the computer-readable medium. When the program is executed by a processor, the steps in any of the optimization methods in the above embodiments are implemented.

[0097] It should be understood that the specific order or hierarchy of steps in the disclosed process is an example of an exemplary method. Based on design preferences, it should be understood that the specific order or hierarchy of steps in the process can be rearranged without departing from the protection scope of the present disclosure. The appended method claims present the elements of various steps in an exemplary order and are not intended to be limited to the specific order or hierarchy described.

[0098] In the foregoing detailed description, various features are combined in a single embodiment to simplify the disclosure. This method of disclosure should not be interpreted as reflecting an intention that the embodiments of the claimed subject matter require more features than are expressly recited in each claim. Rather, as reflected in the appended claims, the invention lies in less than all of the features of a single disclosed embodiment. Accordingly, the appended claims are hereby expressly incorporated into the detailed description, with each claim standing on its own as a separate preferred embodiment of the invention.

[0099] Those skilled in the art should also understand that the various illustrative logical blocks, modules, circuits, and algorithm steps described in connection with the embodiments herein can be implemented as electronic hardware, computer software, or combinations thereof. To clearly illustrate the interchangeability of hardware and software, the various illustrative components, blocks, modules, circuits, and steps have been generally described in terms of their functionality. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the overall system. Skilled artisans may implement the described functionality in a variable manner for each particular application, but such implementation decisions should not be interpreted as departing from the scope of the present disclosure.

[0100] The steps of a method or algorithm described in connection with the embodiments herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination thereof. The software modules may be located in RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium well known in the art. An exemplary storage medium is coupled to the processor such that the processor can read information from, and write information to, the storage medium. Of course, the storage medium may also be integral to the processor. The processor and the storage medium may be located in an ASIC. The ASIC may be located in a user terminal. Of course, the processor and the storage medium may also exist as discrete components in a user terminal.

[0101] For a software implementation, the techniques described in this application can be implemented using modules (e.g., procedures, functions, etc.) that perform the functions described in this application. These software codes can be stored in a memory unit and executed by a processor. The memory unit may be implemented within the processor or outside the processor, and in the latter case, it is coupled to the processor in a communicative manner by various means, which are well known in the art.

[0102] The foregoing description includes examples of one or more embodiments. Of course, it is not possible to describe all possible combinations of components or methods for the purpose of describing the above embodiments, but those of ordinary skill in the art should recognize that the various embodiments can be further combined and arranged. Accordingly, the embodiments described herein are intended to embrace all such changes, modifications, and variations that fall within the scope of the appended claims. In addition, with respect to the term "comprising" as used in the specification or claims, the term is inclusive in a manner similar to the term "including" as interpreted when used as a transitional word in a claim. Further, any use of the term "or" in the specification or claims is to be meant "non-exclusive or".

Claims

1. A control method for a lower limb exoskeleton assist device for climbing a steel tower, characterized in that, Including: Real-time collecting the myoelectric signal and plantar pressure distribution data of the user through a surface electromyography sensor and a plantar pressure sensor; Decomposing the myoelectric signal, calculating to obtain a muscle activation index; constructing a dual-channel LSTM neural network, and realizing the online prediction of the user's motion intention through the dual-channel LSTM neural network; Collecting tower point cloud data, based on the tower point cloud data, identifying the edge of the tower cross-arm structure, generating a safe climbing path through cubic spline interpolation, and calculating the minimum distance threshold between the reachable workspace of the user's foot end and the obstacle, and establishing a gradient sensitivity model to assist in motion safety decision-making; Designing an asymmetric impedance function, adopting a model predictive control algorithm in the upper layer of the gradient sensitivity model to generate an expected joint trajectory, and in the lower layer of the gradient sensitivity model, optimizing the joint stiffness and damping parameters in real time through an improved particle swarm algorithm, and introducing a nonlinear disturbance observer to compensate for the hysteresis effect of the magnetorheological actuator; Establishing a metabolic consumption evaluation model, using a rolling horizon optimization algorithm to solve the minimum energy consumption assistance curve, and setting a safety threshold constraint to limit the overshoot of joint power; Setting a three-level control period, including the actuator current loop control in the first level, the impedance parameter adjustment in the second level, and the path planning update in the third level, and realizing real-time calculation through an FPGA and a GPU hardware acceleration module.

2. The control method of the lower limb exoskeleton assist device according to claim 1, characterized in that, Decompose the electromyogram signal and calculate the muscle activation index. The specific method includes: collecting the electromyogram signals of the main muscle groups of the lower limbs during the user's climbing through a surface electromyogram sensor, decomposing the electromyogram signal x(t) to obtain the intrinsic mode function IMF i (t), i = 1, 2, 3 ···, n, and selecting the first 6 intrinsic mode functions; using the improved Teager energy operator TKO(x[n]) = x 2 [n] - x[n - 1]x[n + 1], calculating the energy of each intrinsic mode, and obtaining the muscle activation index by weighted summation, where x[n] is the sampling sequence of the electromyogram signal.

3. The control method of the lower limb exoskeleton assist device according to claim 1, characterized in that Constructing a dual-channel LSTM neural network, and realizing the online prediction of the user's motion intention through the dual-channel LSTM neural network. The specific method includes: constructing a dual-channel LSTM neural network, where the first channel processes the muscle activation index to predict the motion direction of the user's lower limb joints, the second channel analyzes the plantar pressure distribution data, judges the climbing stage through the plantar pressure distribution data, and dynamically weights multi-sensor data through an attention mechanism to realize the online prediction of the user's motion intention.

4. The control method of the lower limb exoskeleton assist device according to claim 3, characterized in that Dynamically weighting multi-sensor data through an attention mechanism. The specific method includes: applying the attention mechanism, where the output feature of the first channel is F1, the output feature of the second channel is F2, and the weights α1 and α2 are calculated by the attention model, α1 + α2 = 1, to obtain the fused feature F = α1F1 + α2F2.

5. The control method of the lower limb exoskeleton assistive device according to claim 1, characterized in that, Using a lidar installed on the exoskeleton to collect tower point cloud data, and through an improved RANSAC algorithm, randomly sampling and hypothesizing a straight-line model in the point cloud data, calculating the distance from other points to the model to determine the inliers, and iteratively finding the edge of the tower cross-arm structure; Taking the starting point and the target climbing point as endpoints, generating a safe climbing path through cubic spline interpolation, calculating the minimum distance threshold between the reachable workspace of the foot end and the obstacle according to the exoskeleton structure and kinematic parameters, and establishing a gradient sensitivity model to assist in safety decision-making.

6. The control method of the lower limb exoskeleton assist device according to claim 1, characterized in that The expression of the asymmetric impedance function is K d , B d are the asymmetric stiffness and damping coefficients, and K d , B d are adjusted according to different motion directions and climbing stages. The model predictive control algorithm predicts the system response based on the exoskeleton dynamics model and the current state, and solves the optimization problem to generate the desired joint trajectory; the improved particle swarm optimization algorithm iteratively optimizes the joint stiffness and damping parameters as the particle positions, and at the same time introduces a nonlinear disturbance observer to compensate for the hysteresis effect of the magnetorheological actuator.

7. The control method of the lower limb exoskeleton assist device according to claim 1, characterized in that, The expression of the metabolic consumption evaluation model is P m = f(v, a, F m , A), where v is the user's movement speed v, a is the acceleration a, F m is the exoskeleton assistance F m , and A is the muscle activation index. P m is the energy consumption; the rolling horizon optimization algorithm is used to predict the future system state starting from the current moment, solve the exoskeleton assistance curve with the minimum metabolic consumption, and at the same time set the safety threshold to constrain the joint power.

8. The control method of the lower limb exoskeleton assist device according to claim 1, characterized in that, The actuator current loop control of the first stage is the actuator current loop control at 1 kHz, which precisely regulates the output force of the exoskeleton by controlling the current of the magnetorheological actuator at a frequency of 1 kHz; the impedance parameter adjustment of the second stage is the impedance parameter adjustment at 100 Hz, and the impedance parameter adjustment at 100 Hz adjusts the joint stiffness and damping parameters according to the motion state and intention; the path planning of the third stage is updated to the path planning update at 10 Hz, and the path planning update at 10 Hz updates the safe climbing path according to the lidar data and the user state; the FPGA is used to implement the logical operation of sensor data acquisition, and the GPU is used to process the large-scale matrix operation of the LSTM neural network calculation.

9. A control system for a lower limb exoskeleton assistive device for tower climbing, which adopts the control method of any one of the lower limb exoskeleton assistive devices in the above claims 1-8, is characterized in that, Including: The electromyogram signal and plantar pressure distribution data acquisition unit, the online prediction unit of motion intention, the tower cross-arm structure edge and safe climbing path generation unit, the asymmetric impedance function design and joint stiffness and damping parameter optimization unit, the metabolic consumption evaluation model and minimum energy consumption assistance curve solving unit, and the three-level control period setting and hardware acceleration unit; among them: The electromyogram signal and plantar pressure distribution data acquisition unit is used to collect the electromyogram signal and plantar pressure distribution data of the user in real time through surface electromyogram sensors and plantar pressure sensors; The online prediction unit of motion intention is used to decompose the electromyogram signal and calculate the muscle activation index; a two-channel LSTM neural network is constructed, and the online prediction of the user's motion intention is realized through the two-channel LSTM neural network; The tower cross-arm structure edge and safe climbing path generation unit is used to decompose the electromyogram signal and calculate the muscle activation index; a two-channel LSTM neural network is constructed, and the online prediction of the user's motion intention is realized through the two-channel LSTM neural network; The asymmetric impedance function design and joint stiffness and damping parameter optimization unit is used to design an asymmetric impedance function, adopt a model predictive control algorithm to generate the desired joint trajectory in the upper layer of the gradient sensitivity model, and optimize the joint stiffness and damping parameters in real time through an improved particle swarm algorithm in the lower layer of the gradient sensitivity model, and introduce a nonlinear disturbance observer to compensate for the hysteresis effect of the magnetorheological actuator; The metabolic consumption evaluation model and minimum energy consumption assistance curve solving unit is used to establish a metabolic consumption evaluation model, adopt a rolling horizon optimization algorithm to solve the minimum energy consumption assistance curve, and set a safety threshold constraint to limit the joint power overshoot; The three-level control period setting and hardware acceleration unit is used to set the three-level control period, including the actuator current loop control of the first stage, the impedance parameter adjustment of the second stage, and the path planning update of the third stage, and realize real-time calculation through the FPGA and GPU hardware acceleration modules.

10. An electronic device, characterized in that, Including: One or more processors; A memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the control method of the lower limb exoskeleton assistance device according to any one of the above claims 1-8.

Images