Method and device for determining control parameters of human-robot collaborative embodied system, and storage medium

Abstract

The application discloses a human-computer collaborative embodied system control parameter determination method and device and a storage medium, and relates to the technical field of machine control. The method comprises the following steps: acquiring multi-modal perception data, and constructing a terrain semantic vector according to the multi-modal perception data; then, determining the similarity between the terrain semantic vector and the terrain vectors corresponding to each anchor point in a terrain base library, and determining the mixed weight of each anchor point according to the similarity; then, fusing the control parameter sets corresponding to each anchor point according to the mixed weight to generate a candidate parameter set corresponding to the current terrain; finally, determining a target filtering coefficient according to the change amplitude of the terrain semantic vector, and performing smooth filtering on the candidate parameter set based on the target filtering coefficient to obtain a target control parameter, so that the control parameter is continuously updated in a smooth manner without a step mutation phenomenon, the terrain adaptation accuracy of the exoskeleton is ensured, and the walking stability is improved.

Description

Technical Field

[0001] This application relates to the field of machine control technology, and in particular to a method, device and storage medium for determining control parameters of a human-machine collaborative embody system. Background Technology

[0002] Current control parameter generation schemes for wearable lower limb assistive exoskeletons mainly adopt a technical approach that combines discrete terrain classification with table lookup. That is, the terrain is divided into preset discrete categories, each category corresponds to a fixed parameter group; the corresponding parameters are directly retrieved from the pre-stored parameter table based on the real-time terrain recognition results, driving the exoskeleton joints to output assist torque.

[0003] However, in unstructured outdoor environments, the terrain and road conditions vary greatly within the same area, and there may be both potholes and continuous slopes. For example, when a wearer walks on a continuous slope with rapid changes in gradient and large gradient differences, assuming the gradient changes rapidly from 3° to 15°, the above scheme will switch parameter sets instantaneously at a certain threshold, causing a step change in joint torque. This can easily cause the wearer to feel a push or bump, disrupt their normal gait rhythm, or even lose their center of gravity, resulting in low walking stability.

[0004] The above content is only used to help understand the technical solution of this application and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] The main purpose of this application is to provide a method, device and storage medium for determining control parameters of a human-machine collaborative embody system, aiming to solve the technical problem of how to improve the stability of human-machine collaborative walking.

[0006] To achieve the above objectives, this application proposes a method for determining control parameters of a human-machine collaborative embody system, the method comprising: Acquire multimodal sensing data, and construct a terrain semantic vector based on the multimodal sensing data; Determine the similarity between the terrain semantic vector and the terrain vector corresponding to each anchor point in the terrain base library, and determine the mixing weight of each anchor point based on the similarity. The control parameter sets corresponding to each anchor point are fused according to the hybrid weight to generate a candidate parameter set corresponding to the current terrain. The target filtering coefficients are determined based on the change magnitude of the terrain semantic vector, and the candidate parameter set is smoothed based on the target filtering coefficients to obtain the target control parameters.

[0007] In one embodiment, the multimodal perception data includes ground point cloud data acquired by a visual camera and pressure sensing data acquired by a plantar pressure-sensitive unit, and the terrain semantic vector includes one or more of forward slope, lateral tilt angle, ground roughness, ground continuity value, lateral load offset, ground hardness, and slope time stability value.

[0008] In one embodiment, the step of constructing a terrain semantic vector based on the multimodal perception data includes one or more of the following: Select a target ground point cloud within a preset range, determine the normal vector of the fitting plane of the target ground point cloud, and determine the forward slope based on the angle between the normal vector and the vertical direction; The normal vector is projected onto the horizontal plane to obtain a horizontal projection vector. The lateral tilt angle is determined based on the angle between the horizontal projection vector and the current horizontal direction of travel. Determine the vertical distance from each of the target ground point clouds to the fitted plane, and determine the ground roughness based on the standard deviation of the residuals of the vertical distances; Determine the density dispersion of the target ground point cloud, and determine the ground continuity value based on the density dispersion; Based on the pressure values ​​and position coordinates collected by the pressure-sensitive units on the left and right soles, the pressure centers of the left and right feet and the lateral offset of the line connecting the pressure centers are determined. Based on the ratio of the lateral offset to the preset reference lateral distance between the left and right feet, the lateral load offset is determined. Determine the rate of pressure change of the pressure value within a preset time period before and after the pressure-sensitive unit touches the ground, and determine the ground hardness based on the rate of pressure change; The standard deviation of the forward slope within a preset time window is determined, and the ground roughness is weighted and fused with the standard deviation to obtain the slope time stability value.

[0009] In one embodiment, the step of determining the similarity between the terrain semantic vector and the terrain vector corresponding to each anchor point in the terrain base library, and determining the mixing weight of each anchor point based on the similarity, includes: The Euclidean distance between the terrain semantic vector and the terrain vector corresponding to each anchor point in the terrain base library is determined by a preset Gaussian kernel function, and the Euclidean distance is used as the similarity. The similarity corresponding to each anchor point is normalized and used as the mixed weight for each anchor point.

[0010] In one embodiment, the method for determining the control parameters of the human-machine collaborative embody system further includes the following steps: Determine the geometric distance between the historical terrain vector and the terrain vector corresponding to each anchor point; If the geometric distance between the historical terrain vector and the terrain vectors corresponding to all anchor points is greater than a preset distance threshold, then the ground point cloud corresponding to the historical terrain vector is taken as a blind zone terrain point. Clustering the terrain points in the blind area yields the blind area terrain vector; Construct new anchor points in the terrain base library, and use the blind zone terrain vector as the terrain vector corresponding to the new anchor points; The set of control parameters corresponding to the new anchor point is determined based on the terrain vector of the blind zone.

[0011] In one embodiment, the step of determining the control parameter set corresponding to the new anchor point based on the blind zone terrain vector includes: The first neighboring anchor point of the new anchor point is determined based on the geometric distance between the terrain vector of the blind zone and the terrain vector corresponding to each anchor point; The parameter weights of the first neighbor anchor points are determined based on the geometric distance corresponding to the first neighbor anchor point. The control parameter set corresponding to the first neighbor anchor point is fused according to the parameter weight to obtain the control parameter set corresponding to the new anchor point.

[0012] In one embodiment, the step of clustering the blind area terrain points to obtain the blind area terrain vector further includes: Traverse all anchor points in the terrain base library and determine the second neighbor anchor point of the current anchor point. Determine a first difference between the terrain vectors of the current anchor point and the second neighboring anchor point, and a second difference between the control parameter sets of the current anchor point and the second neighboring anchor point; The sensitivity of the current parameter region is determined based on the ratio of the first difference to the second difference, wherein the current parameter region is the parameter range of the control parameter set corresponding to the current anchor point and the second neighboring anchor point; Based on the sensitivity of the current parameter region, determine the corresponding anchor point density within the current parameter region; The number of clusters of the blind area terrain points is determined based on the anchor point density, and the step of clustering the blind area terrain points to obtain blind area terrain vectors is performed based on the number of clusters, wherein the number of blind area terrain vectors and the number of new anchor points are consistent with the number of clusters.

[0013] In one embodiment, after the steps of determining the target filtering coefficients based on the change magnitude of the terrain semantic vector and smoothing the candidate parameter set based on the target filtering coefficients to obtain the target control parameters, the method further includes: When the mixed weight of any target anchor point is greater than the preset weight threshold, the control deviation index corresponding to the target control parameter is determined. When the control deviation index is greater than the control deviation index of other anchor points, the recalibration of the control parameter set corresponding to the target anchor point is triggered, or the target anchor point is deleted. This is a first proprietary technical solution.

[0014] Furthermore, to achieve the above objectives, this application also proposes a control parameter determination device for a human-machine collaborative embody system. The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor. The computer program is configured to implement the steps of the control parameter determination method for the human-machine collaborative embody system as described above.

[0015] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the control parameter determination method for the human-machine collaborative embody system as described above.

[0016] This application provides a method for determining control parameters of a human-machine collaborative embody system. The method involves acquiring multimodal perception data and constructing a terrain semantic vector based on this data. Then, the similarity between the terrain semantic vector and the terrain vectors corresponding to each anchor point in the terrain base library is determined, and the mixing weight of each anchor point is determined based on the similarity. Next, the control parameter sets corresponding to each anchor point are fused according to the mixing weights to generate a candidate parameter set corresponding to the current terrain. Finally, the target filtering coefficient is determined based on the change amplitude of the terrain semantic vector, and the candidate parameter set is smoothed based on the target filtering coefficient to obtain the target control parameters.

[0017] The above method uses continuous high-dimensional terrain semantic vectors to fully represent the comprehensive state of unstructured composite terrain. It adapts to various mixed road conditions through a multi-standard terrain anchor point convex combination fusion strategy. At the same time, it adaptively adjusts the filtering coefficients according to the fluctuation amplitude of the terrain semantic vectors, which can achieve smooth and continuous updates of control parameters without abrupt changes. It can dynamically adapt to various working conditions such as road surface inclination, unevenness, softness and hardness, and surface coherence, ensuring the accuracy of exoskeleton-assisted terrain adaptation and thus improving walking stability. Attached Figure Description

[0018] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0019] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A flowchart illustrating the method for determining control parameters of the applicant's human-machine collaborative embodiment system, provided in Embodiment 1. Figure 2 A flowchart illustrating the second embodiment of the method for determining control parameters of the applicant's human-machine collaborative embodiment system; Figure 3 A flowchart illustrating the method for determining control parameters of the applicant's human-machine collaborative embodiment system, provided in Embodiment 3. Figure 4 A flowchart illustrating the method for determining control parameters of the applicant's human-machine collaborative embodiment system, provided in Embodiment 4. Figure 5 This is a schematic diagram of the equipment structure of the hardware operating environment involved in the method for determining control parameters of the human-machine collaborative embody system in the embodiments of this application.

[0021] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0022] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of this application and are not intended to limit this application.

[0023] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific embodiments. It should be noted that all actions involving the acquisition of signals, information, or data in this application are performed in accordance with the relevant data protection laws and regulations of the country where the application is located, and with authorization from the owner of the corresponding device.

[0024] Current control parameter generation schemes for wearable lower limb assistive exoskeletons mainly adopt a technical approach that combines discrete terrain classification with table lookup. That is, the terrain is divided into preset discrete categories, each category corresponds to a fixed parameter group; the corresponding parameters are directly retrieved from the pre-stored parameter table based on the real-time terrain recognition results, driving the exoskeleton joints to output assist torque.

[0025] However, in unstructured outdoor environments, the terrain and road conditions vary greatly within the same area, and there may be both potholes and continuous slopes. For example, when a wearer walks on a continuous slope with rapid changes in gradient and large gradient differences, assuming the gradient changes rapidly from 3° to 15°, the above scheme will switch parameter sets instantaneously at a certain threshold, causing a step change in joint torque. This can easily cause the wearer to feel a push or bump, disrupt their normal gait rhythm, or even lose their center of gravity, resulting in low walking stability.

[0026] In view of the above problems, this application proposes a method for determining control parameters of a human-machine collaborative embody system. The method involves acquiring multimodal perception data and constructing a terrain semantic vector based on this data. Then, the similarity between the terrain semantic vector and the terrain vectors corresponding to each anchor point in the terrain base library is determined, and the mixing weight of each anchor point is determined based on the similarity. Next, the control parameter sets corresponding to each anchor point are fused according to the mixing weights to generate a candidate parameter set corresponding to the current terrain. Finally, the target filtering coefficient is determined based on the change amplitude of the terrain semantic vector, and the candidate parameter set is smoothed based on the target filtering coefficient to obtain the target control parameters.

[0027] The above method uses continuous high-dimensional terrain semantic vectors to fully represent the comprehensive state of unstructured composite terrain. It adapts to various mixed road conditions through a multi-standard terrain anchor point convex combination fusion strategy. At the same time, it adaptively adjusts the filtering coefficients according to the fluctuation amplitude of the terrain semantic vectors, which can achieve smooth and continuous updates of control parameters without abrupt changes. It can dynamically adapt to various working conditions such as road surface inclination, unevenness, softness and hardness, and surface coherence, ensuring the accuracy of exoskeleton-assisted terrain adaptation and thus improving walking stability.

[0028] The following uses a human-machine collaborative embody system as an example to illustrate this embodiment and the following embodiments.

[0029] Based on this, the first embodiment of this application provides a method for determining control parameters of a human-machine collaborative embody system, referring to... Figure 1 In this embodiment, the method for determining the control parameters of the human-machine collaborative embody system includes steps S10 to S40: Step S10: Obtain multimodal perception data and construct a terrain semantic vector based on the multimodal perception data.

[0030] Multimodal perception data consists of heterogeneous environmental and motion data collected by different types of sensors, including ground point cloud data collected by visual cameras and pressure sensing data collected by plantar pressure-sensitive units. The terrain semantic vector is a vector composed of multiple continuous physical dimensions such as slope, tilt angle, and roughness, representing the comprehensive characteristics of the current terrain, including one or more of the following: forward slope, lateral tilt angle, ground roughness, ground continuity value, lateral load offset, ground hardness, and slope time stability value.

[0031] Furthermore, step S10 above includes one or more of steps S11 to S17: Step S11: Select a target ground point cloud within a preset range, determine the normal vector of the fitting plane of the target ground point cloud, and determine the forward slope based on the angle between the normal vector and the vertical direction.

[0032] As an alternative method for determining forward slope, a target ground point cloud is obtained by filtering a set of points within a preset range based on spatial coordinates from ground point cloud data acquired by a visual camera. The three-dimensional coordinates of all points in the target ground point cloud along the X, Y, and Z axes are extracted. A least-squares algorithm is then used to perform a plane fitting operation on the three-dimensional coordinates of all points, solving for the plane equation corresponding to the unique fitted plane, and thus determining the normal vector corresponding to this plane equation. The spatial angle between the normal vector corresponding to the plane equation and the vertical direction, i.e., the Z-axis direction, is calculated and used as the forward slope.

[0033] As an alternative approach to determining the forward slope, after obtaining the target ground point cloud, the three-dimensional coordinates of all points within the target ground point cloud on the X, Y, and Z axes are extracted to construct a coordinate dataset. This dataset is then preprocessed to decentralize it, eliminating computational interference caused by coordinate reference offset. Principal component analysis is performed on the decentralized coordinate dataset to extract the principal direction of the dataset, thereby determining the fitting plane. The normal vector of the fitting plane is then calculated, and the spatial angle between the normal vector of the fitting plane and the vertical direction, i.e., the Z-axis direction, is taken as the forward slope.

[0034] For example, the steps described above, such as performing principal component analysis on the decentralized coordinate dataset to extract the principal direction of the coordinate dataset, thereby determining the fitting plane, and calculating the normal vector of the fitting plane, can specifically include: selecting the three-dimensional coordinates (x, y, z) of all points within the target ground point cloud and integrating them to form a coordinate dataset containing the spatial location information of all points; performing decentralization processing on the coordinate dataset by calculating the average values ​​of the x, y, and z coordinates of all points respectively; then subtracting the average values ​​of the corresponding dimensions from the original x, y, and z coordinates of each point to obtain the decentralized coordinate dataset; constructing a three-dimensional covariance matrix based on the decentralized coordinate dataset, which is used to characterize the data dispersion and correlation features between the three coordinate dimensions of the X, Y, and Z axes; performing eigenvalue decomposition on the three-dimensional covariance matrix to obtain the eigenvalues ​​and corresponding eigenvectors of the three-dimensional covariance matrix. The eigenvalues ​​represent the degree of dispersion of the data in the direction of the corresponding eigenvector, and the eigenvectors represent the spatial direction of the data distribution. All eigenvalues ​​are sorted from largest to smallest. The eigenvalue with the largest value corresponds to the first principal direction, the second largest to the second principal direction, and the smallest to the third principal direction. The eigenvector corresponding to the smallest eigenvalue is selected as the direction vector perpendicular to the point cloud distribution plane, which is also the normal vector of the fitting plane.

[0035] Step S12: Project the normal vector onto the horizontal plane to obtain a horizontal projection vector, and determine the lateral tilt angle based on the angle between the horizontal projection vector and the current horizontal travel direction.

[0036] The lateral tilt angle is the angle between the horizontal projection of the normal vector of the fitted plane and the current horizontal direction of travel in the horizontal plane. It intuitively reflects the degree of left-right tilt or lateral deviation of the road surface relative to the travel route, indicating the presence of a lateral slope in the terrain. The lateral tilt angle quantifies the lateral slope or tilt of the road surface, distinguishes between flat ground, left-sloping roads, and right-sloping roads, and compensates for the deficiency of being unable to identify cross-slope terrain based solely on forward slope, thus reconstructing the characteristics of unstructured terrain.

[0037] For example, the normal vector of the fitted plane obtained in step S11 is acquired. Using the horizontal plane, i.e., the plane formed by the X-axis and Y-axis, as the projection reference plane, the normal vector in three-dimensional space is orthogonally projected onto the horizontal plane, retaining the two dimensional components of the normal vector on the horizontal plane, to obtain the aforementioned horizontal projection vector. Next, the current horizontal direction of travel of the wearable lower limb assistive exoskeleton is acquired and converted into a travel direction vector in the horizontal plane. The angle between the horizontal projection vector and the travel direction vector in the two-dimensional space of the horizontal plane is calculated as the aforementioned lateral tilt angle.

[0038] Step S13: Determine the vertical distance from each of the target ground point clouds to the fitting plane, and determine the ground roughness based on the standard deviation of the residuals of the vertical distances.

[0039] Ground roughness measures the degree of deviation between the actual ground and the fitted plane. A larger standard deviation of the vertical distance residual indicates more significant fluctuations in the vertical distance of the target ground point cloud, and a higher degree of unevenness, undulation, or roughness in the ground; a smaller standard deviation indicates a smoother ground. Ground roughness, as one dimension of the terrain semantic vector, is used to distinguish between smooth roads, potholed roads, and gravel roads. Wearable lower limb assistive exoskeletons can dynamically adjust joint assist flexibility, torque output rate, and gait cushioning parameters based on ground roughness. For example, on high-roughness roads, joint response is appropriately slowed down and cushioning is increased to reduce vibration and impact; on smooth roads, conventional assist output is maintained to improve walking efficiency, while other terrain parameters are used to achieve a smooth transition of control parameters, ensuring stable walking.

[0040] For example, the target ground point cloud and its corresponding fitting plane, determined in step S11, are obtained. Each point in the target ground point cloud is traversed, and the vertical distance from that point to the fitting plane is calculated. This process is repeated for all points to obtain a vertical distance dataset. Next, the average value of the vertical distance dataset is calculated. The average value is then subtracted from the vertical distance of each point to obtain the distance residuals for each point, forming a residual dataset. The standard deviation of the residual dataset is then used to obtain the aforementioned ground roughness.

[0041] Step S14: Determine the density dispersion of the target ground point cloud, and determine the ground continuity value based on the density dispersion.

[0042] Ground continuity values ​​characterize the integrity and continuity of the ground surface, reflecting the presence of voids, faults, depressions, obstructions, or fractures in the road surface. Low density dispersion indicates a uniform point cloud distribution, meaning a high ground continuity value, representing a complete and continuous road surface without obvious breaks. High density dispersion indicates sparse or even absent point clouds in some areas, meaning a low ground continuity value, representing discontinuous areas such as potholes, fractures, or defects in the road surface.

[0043] For example, using the previously selected target ground point cloud, the spatial region containing the target ground point cloud is divided into several uniform grid cells of a uniform size. The number of point clouds contained in each grid cell is counted to obtain a point cloud quantity dataset corresponding to all grid cells, thus reflecting the density of point cloud distribution in different regions. The density dispersion is determined based on the variance, range, or coefficient of variation of the point cloud quantity dataset. A higher density dispersion indicates a more significant difference in point cloud density across regions; a lower dispersion indicates a more uniform point cloud distribution. According to a preset mapping relationship, the quantified results of the density dispersion are converted into corresponding values ​​to obtain the aforementioned ground continuity value.

[0044] Step S15: Based on the pressure values ​​and position coordinates collected by the pressure-sensitive units on the left and right soles, determine the pressure center of the left and right feet and the lateral offset of the line connecting the pressure centers. Based on the ratio of the lateral offset to the preset reference lateral distance between the left and right feet, determine the lateral load offset.

[0045] Lateral load offset is a dimensionless ratio used to quantify the lateral displacement of the center of force on the sole of the foot relative to a standard standing or walking posture, reflecting the degree to which the body's center of gravity and foot load shift to the left or right. The closer the lateral load offset is to 0, the more symmetrical the load distribution; the larger the absolute value of the lateral load offset, the higher the proportion of load on one side and the more obvious the center of gravity tilt.

[0046] For example, the human-machine collaborative embody system reads the sensing data from the pressure-sensitive units on the soles of the left and right feet in real time, including the pressure value and position coordinates of each sensing point. Then, it calculates the pressure center coordinates of each sole: using the pressure value collected by the pressure-sensitive unit at each sensing point on the sole as a weight, it performs a weighted average calculation on the position coordinates of that sensing point, sequentially calculating the pressure centers of the left and right feet, obtaining the pressure center coordinates of the left and right feet. Connecting the left and right foot pressure center coordinates yields a line, and the horizontal deviation of this line relative to the reference line in the standard posture is calculated, i.e., the positional difference between this line and the reference line on the horizontal plane, obtaining the lateral offset. The reference line in the standard posture refers to the straight line connecting the pressure centers of the left and right feet in a standard walking posture of normal upright or uniform speed walking on a level, flat surface; it is a preset reference baseline. Next, the preset reference lateral distance between the left and right feet is obtained from the human-machine collaborative body system. That is, when the wearer maintains a standard upright or uniform walking posture on a level and flat road, the standard distance between the pressure centers of the left and right feet is obtained. The lateral offset is divided by the preset reference lateral distance to obtain the lateral load offset.

[0047] Step S16: Determine the rate of pressure change of the pressure value within a preset time period before and after the pressure-sensitive unit touches the ground, and determine the ground hardness based on the rate of pressure change.

[0048] Ground hardness is used to characterize the softness, elasticity, and cushioning capacity of a contact surface. Hard surfaces such as cement or stone deform very little, causing foot pressure to rise rapidly and changing at a large rate; soft surfaces such as soil, grass, or mats exhibit cushioning deformation, resulting in a gradual increase in pressure and a smaller rate of pressure change.

[0049] For example, pressure data within a preset time period before and after the pressure-sensitive units on the left and right soles are first extracted from the sensing data of the pressure-sensitive units. Then, the rate of change of the pressure data with time within the preset time period is determined. Finally, the rate of change of pressure is converted into the aforementioned ground hardness according to the mapping rules preset by the human-machine collaborative embodied system.

[0050] As an optional approach to extracting pressure data within a preset time period before and after the pressure-sensitive unit touches the ground, the human-machine collaborative embody system continuously collects pressure values ​​from the pressure-sensitive units at each sensing point on the sole of the foot at a fixed sampling frequency, forming time-series pressure data. It also obtains a preset foot contact pressure threshold, which is the critical pressure value used to distinguish between foot suspension and foot contact with the ground. The system iterates through the time-series pressure data frame by frame. When the real-time pressure value detected by the pressure-sensitive unit changes from less than the foot contact pressure threshold to greater than or equal to the foot contact pressure threshold, the current time is recorded and determined as the foot contact start time. The system continues to iterate through the time-series pressure data. When the real-time pressure value changes from greater than or equal to the foot contact pressure threshold to less than the foot contact pressure threshold, the current time is recorded and determined as the foot lift-off time. The time-series pressure data between the foot contact start time and the foot lift-off time is used as the pressure data within the preset time period before and after the pressure-sensitive unit touches the ground.

[0051] Step S17: Determine the standard deviation of the forward slope within a preset time window, and weight and fuse the ground roughness with the standard deviation to obtain the slope time stability value.

[0052] Slope temporal stability value is used to comprehensively evaluate the dynamic stability of slope morphology and road conditions over a short period of time. The smaller the slope temporal stability value, the more drastic the dynamic changes of the slope and the worse the stability, the larger the standard deviation of the forward slope and the greater the surface roughness. When the slope temporal stability is relatively small, it is determined that the road surface is continuously undulating and its condition is changeable. The human-machine collaborative embody system can slow down the rate of change of joint torque to improve the motion buffering capacity.

[0053] For example, the human-machine collaborative embody system defines a preset time window of fixed duration. Within this preset time window, multiple forward slopes are continuously collected to form a forward slope time-series dataset. The standard deviation of this forward slope time-series dataset is determined to reflect the fluctuation range of the forward slope over a short period of time. Simultaneously, the ground roughness calculated in step S13 is obtained. The standard deviations of the ground roughness and the forward slope are multiplied by their corresponding weights and then summed to obtain the slope time stability value.

[0054] The above steps calculate the continuous physical dimension features of forward slope, lateral tilt angle, ground roughness, ground continuity value, lateral load offset, ground hardness, and slope time stability value from point cloud and pressure sensor multimodal data, and construct a terrain semantic vector. This differs from traditional methods that classify terrain into a single discrete category. It fully preserves multi-dimensional terrain information such as slope, tilt, unevenness, road surface integrity, load distribution, ground hardness, and slope dynamic fluctuations, without losing key features of complex terrain. Furthermore, it constructs a feature space with objective geometric distance based on continuous vectors, allowing the feature vectors to transition smoothly and continuously when the terrain state changes. This avoids the information fragmentation and parameter step switching problems caused by the forced mapping of labels in traditional discrete classification methods.

[0055] Step S20: Determine the similarity between the terrain semantic vector and the terrain vector corresponding to each anchor point in the terrain base library, and determine the mixing weight of each anchor point based on the similarity.

[0056] The terrain base library is a database of multiple pre-stored terrain vectors. Anchor points are pre-calibrated terrain sample points in the terrain base library. Each anchor point is bound to a unique terrain vector and a corresponding set of control parameters.

[0057] Furthermore, step S20 above also includes steps S21 to S22: Step S21: Determine the Euclidean distance between the terrain semantic vector and the terrain vector corresponding to each anchor point in the terrain base library using a preset Gaussian kernel function, and use the Euclidean distance as the similarity.

[0058] Step S22: Normalize the similarity corresponding to each anchor point, and use it as the mixed weight for each anchor point.

[0059] For example, the terrain vectors corresponding to all anchor points in the terrain base library are read one by one. The current terrain semantic vector is paired with the terrain vector of a single anchor point, and the Euclidean distance between the two vectors is calculated to obtain the Euclidean distance value for each pair of vectors. Each Euclidean distance value is input into a preset Gaussian kernel function for calculation. The output of the Gaussian kernel function is the similarity of the corresponding anchor point. The similarity of all anchor points is globally summed to obtain the total similarity. The similarity of each individual anchor point is then divided by the total similarity to complete the normalization operation, resulting in the mixed weight corresponding to each anchor point.

[0060] Step S30: The control parameter sets corresponding to each anchor point are fused according to the hybrid weight to generate a candidate parameter set corresponding to the current terrain.

[0061] The control parameter set is a set of control parameters that drive the joints of the wearable lower limb assistive exoskeleton to output assist torque. Each anchor point corresponds to an independent control parameter set. The control parameter set consists of multiple independent, each with its own function, and each independent variable is a parameter dimension. Parameter dimensions may include, but are not limited to: joint assist torque dimension, corresponding to the output torque magnitude of the hip, knee, and ankle joints; joint response speed dimension, used to control the speed of torque output; gait stride limitation dimension, used to adjust the maximum allowable stride based on ground continuity and roughness; landing cushioning stiffness dimension, used to adjust the degree of joint flexibility and cushioning when the foot strikes the ground; left-right torque distribution coefficient dimension, used to balance the torque on both sides in conjunction with lateral tilt angle and lateral load offset; and center of gravity offset correction coefficient dimension, used to counteract the posture tilt caused by cross slope or off-center loading.

[0062] The magnitude of the aforementioned mixing weight represents the degree of fit between the current terrain and the terrain corresponding to the anchor point. The higher the mixing weight, the greater the proportion of the control parameter set of that anchor point in the fusion process. Based on the mixing weight, a convex combination operation is performed on the control parameter sets corresponding to all anchor points, merging multiple sets of standard control parameter sets into a candidate parameter set that adapts to the current complex terrain. This solves the problem that the control parameter set of a single anchor point cannot match complex terrain.

[0063] As an alternative approach to generating a candidate parameter set for the current terrain, the control parameter set corresponding to all anchor points is read from the terrain base library. For each parameter dimension in the control parameter set, the parameter value of each anchor point under that parameter dimension is extracted sequentially, and multiplied with the corresponding anchor point's mixed weight. All multiplication results under that parameter dimension are summed to obtain the fused parameter value for that parameter dimension. The fused parameter values ​​from all parameter dimensions are then combined in their original order to generate the candidate parameter set for the current terrain.

[0064] As an alternative approach to generating the candidate parameter set corresponding to the current terrain, the control parameter sets of all anchor points are extracted, and each set of control parameter sets is treated as a single parameter vector. Based on the determined mixing weights of each anchor point, these weights are used as weighting coefficients for the parameter vector corresponding to that anchor point. A scalar multiplication operation is then performed on the parameter vector of each anchor point with its own mixing weight to obtain a weighted parameter vector. All weighted parameter vectors are then superimposed dimension-by-dimensional using vector addition to obtain the fused overall parameter vector, which serves as the candidate parameter set for the current terrain.

[0065] Step S40: Determine the target filtering coefficients based on the change amplitude of the terrain semantic vector, and perform smoothing filtering on the candidate parameter set based on the target filtering coefficients to obtain the target control parameters.

[0066] The change amplitude of the terrain semantic vector is a quantified value of the overall fluctuation difference of the terrain semantic vector between adjacent sampling times and between two consecutive frames, representing the speed of terrain state change.

[0067] For example, the Euclidean distance between the historical terrain semantic vector stored in the previous sampling period and the terrain semantic vector of the current frame is obtained as the change amplitude of the terrain semantic vector. Then, according to the pre-defined segmented mapping relationship of the human-machine collaborative embody system, the filtering coefficient corresponding to the change amplitude is determined. The smaller the change amplitude, the more stable the terrain, and the larger the filtering coefficient; the larger the change amplitude, the more abrupt the terrain, and the smaller the filtering coefficient. Afterwards, the historical target control parameter set output in the previous sampling period and the candidate parameter set generated in this sampling period are obtained. A first-order low-pass filtering operation is performed on each parameter dimension in the candidate parameter set: the output parameter of this sampling period = historical target control parameter × target filtering coefficient + candidate parameter set × (1 (Target filter coefficients). After filtering calculations are completed for all parameter dimensions, all output parameters are integrated to obtain the smoothed target control parameters.

[0068] In this embodiment, point cloud and plantar pressure multimodal sensing data are first collected. Then, forward slope, lateral tilt angle, ground roughness, ground continuity value, lateral load offset, ground hardness, and slope time stability value are calculated sequentially to construct a terrain semantic vector. Next, based on a terrain base library with pre-stored exclusive control parameter anchor points, the similarity of each anchor point is obtained by converting Euclidean distance using Gaussian kernel and normalizing it to generate a hybrid weight. The hybrid weight is used to convexly combine and fuse the control parameter sets of each anchor point dimension by dimension to generate a candidate parameter set. Finally, based on the fluctuation amplitude of the terrain semantic vectors of the previous and next frames, adaptive matching filtering coefficients are used to smooth the candidate parameter set to obtain the final target control parameters to be issued for execution. The above method uses continuous high-dimensional terrain semantic vectors to fully represent the comprehensive state of unstructured composite terrain. It adapts to various mixed road conditions through a multi-standard terrain anchor point convex combination fusion strategy. At the same time, it adaptively adjusts the filtering coefficients according to the fluctuation amplitude of the terrain semantic vectors, taking into account both the terrain response speed of the control parameters and the smooth operation of the human-machine collaborative embody system. It can achieve smooth and continuous updates of control parameters without abrupt changes. It dynamically adapts to various working conditions such as road surface inclination, unevenness, softness and hardness, and surface coherence. While ensuring the accuracy of terrain adaptation of the exoskeleton, it improves walking stability and reduces the risk of gait imbalance or unilateral limb overload. It significantly improves the smoothness of the exoskeleton's assistance output and terrain adaptation ability in complex outdoor environments.

[0069] Example 2 Based on the above embodiment one, in this embodiment, referring to... Figure 2 The method for determining the control parameters of the aforementioned human-machine collaborative embody system further includes steps S50 to S90: Step S50: Determine the geometric distance between the historical terrain vector and the terrain vector corresponding to each anchor point.

[0070] Step S60: If the geometric distance between the historical terrain vector and the terrain vectors corresponding to all anchor points is greater than a preset distance threshold, then the ground point cloud corresponding to the historical terrain vector is taken as a blind zone terrain point.

[0071] Historical terrain vectors are terrain semantic vectors that have been calculated, generated, and stored at previous sampling times. Blind zone terrain points refer to terrain sampling points that cannot be successfully matched with any existing anchor point in the terrain base database, meaning that the base database does not have corresponding sample coverage.

[0072] For example, read the terrain vectors corresponding to all original anchor points in the terrain base library, and calculate the geometric distance between each historical terrain vector and the terrain vector of each anchor point in turn. If the geometric distance between the historical terrain vector and the terrain vectors of all anchor points is greater than a preset distance threshold, then mark the ground point cloud corresponding to the historical terrain vector as a blind zone terrain point.

[0073] Step S70: Cluster the blind area terrain points to obtain the blind area terrain vector.

[0074] The blind zone terrain vector is a representative vector obtained after clustering multiple blind zone terrain points, representing a new type of terrain that has not been included in the data.

[0075] For example, the historical terrain semantic vectors corresponding to all blind area terrain points after filtering are obtained. A clustering algorithm is used to perform grouping clustering operations on the historical terrain semantic vectors corresponding to all blind area terrain points. The historical terrain semantic vectors corresponding to blind area terrain points with similar features are divided into the same cluster. The mean operation is performed on all historical terrain semantic vectors in each cluster to calculate the center vector corresponding to each cluster. This center vector is the aforementioned blind area terrain vector.

[0076] Step S80: Construct a new anchor point in the terrain base library and use the blind zone terrain vector as the terrain vector corresponding to the new anchor point.

[0077] For example, after determining the blind zone terrain vector, a write creation command is sent to the terrain base library to create a new storage entry in the storage structure of the terrain base library, that is, to create a new anchor point, and the blind zone terrain vectors calculated by each cluster are written into the new anchor point entry and bound as the terrain vectors corresponding to the new anchor point.

[0078] Step S90: Determine the set of control parameters corresponding to the new anchor point based on the blind zone terrain vector.

[0079] Furthermore, step S90 above includes steps S91 to S93: Step S91: Determine the first neighboring anchor point of the new anchor point based on the geometric distance between the blind zone terrain vector and the terrain vector corresponding to each anchor point.

[0080] Step S92: Determine the parameter weight of the first neighbor anchor point based on the geometric distance corresponding to the first neighbor anchor point.

[0081] Step S93: The control parameter set corresponding to the first neighbor anchor point is fused according to the parameter weight to obtain the control parameter set corresponding to the new anchor point.

[0082] For example, the blind zone terrain vector corresponding to the new anchor point is read, and the terrain vectors of all existing anchor points in the terrain base library are traversed. The geometric distance between the blind zone terrain vector and the terrain vectors of each existing anchor point is calculated one by one, and the first neighbor anchor point whose geometric distance is less than a preset geometric distance threshold is selected. The geometric distances corresponding to each first neighbor anchor point are converted into similarity values ​​using a Gaussian kernel function, and the similarity values ​​are normalized to obtain the parameter weights of each first neighbor anchor point. For each parameter dimension, the control parameter of that parameter dimension of the first neighbor anchor point is multiplied by the corresponding parameter weight and then summed to obtain the interpolation parameter of each parameter dimension. Then, the interpolation parameters of all parameter dimensions are combined to form the control parameter set of the new anchor point.

[0083] This second embodiment adds an autonomous terrain update and expansion process to the first embodiment. First, it calculates the geometric distance between the historical terrain vector and the terrain vectors of all anchor points in the terrain base library. Then, it identifies blind area terrain points in the terrain base library that have no matching samples by using a preset distance threshold. Next, it obtains the blind area terrain vector through clustering operations. In this way, it creates new anchor points in the terrain base library and binds them to the blind area terrain vector. Finally, it generates a new set of control parameters for the new anchor points by interpolating the neighboring anchor points. This realizes the autonomous online expansion of the terrain base library. It can actively identify unfamiliar terrain conditions not covered by the initial sample library and supplement anchor point samples, continuously enriching the terrain matching sample system. With long-term use, it can continuously improve the completeness of terrain coverage and matching accuracy, reduce parameter adaptation deviation under unfamiliar terrain, and further enhance the exoskeleton's adaptive adjustment capability and long-term operational stability in the face of diverse and unknown outdoor road conditions.

[0084] Example 3 Based on Embodiment 2 above, in the method for determining control parameters of the human-machine collaborative embody system proposed in this embodiment, referring to... Figure 3 Before step S70, steps S100 to S140 are also included: Step S100: Traverse all anchor points in the terrain base library and determine the second neighbor anchor point of the current anchor point.

[0085] Step S110: Determine the first difference between the terrain vectors of the current anchor point and the second neighboring anchor point, and the second difference between the control parameter sets of the current anchor point and the second neighboring anchor point.

[0086] Step S120: Determine the sensitivity of the current parameter region based on the ratio of the first difference to the second difference, wherein the current parameter region is the parameter range of the control parameter set corresponding to the current anchor point and the second neighboring anchor point.

[0087] For example, all anchor points in the terrain base database are read, and a single anchor point is selected sequentially as the current anchor point. The geometric distance between the terrain vector of the current anchor point and the terrain vectors of other anchor points in the terrain base database is calculated. Anchor points whose geometric distance is less than a preset proximity threshold are designated as the second neighbor anchor points corresponding to the current anchor point. After determining the second neighbor anchor points corresponding to the current anchor point, a first difference between the terrain vectors of the current anchor point and the second neighbor anchor point, and a second difference between the control parameter sets of the current anchor point and the second neighbor anchor point are determined. The ratio of the second difference to the first difference is used as the sensitivity corresponding to the current parameter region. Here, the current parameter region refers to the range of control parameter values ​​enclosed by the current anchor point and the second neighbor anchor point, and the sensitivity is the change in control parameters corresponding to a unit change in the terrain vector, reflecting the sensitivity of the parameters to changes in terrain features.

[0088] Step S130: Determine the anchor point density corresponding to the current parameter region based on the sensitivity of the current parameter region.

[0089] Step S140: Determine the number of clusters for the blind area terrain points based on the anchor point density, and perform the step of clustering the blind area terrain points to obtain blind area terrain vectors based on the number of clusters, wherein the number of blind area terrain vectors and the number of new anchor points are consistent with the number of clusters.

[0090] For example, after determining the sensitivity of the current parameter region, the anchor point density corresponding to the sensitivity of the current parameter region is determined based on a preset sensitivity and anchor point density mapping rule. A higher sensitivity of the current parameter region means that even subtle terrain differences in this terrain space can cause significant changes in the assistive parameters required by the wearable lower limb assistive exoskeleton. If the anchor points are sparsely distributed, relying solely on interpolation and fusion using only a few anchor points can easily lead to parameter calculation deviations and inaccurate adaptation. Therefore, increasing the anchor point density corresponding to the current parameter region and arranging more terrain sample anchor points can reduce the terrain span between adjacent anchor points, limiting the interpolation calculation error to a smaller range and ensuring that the control parameters fused under complex terrain conform to real-world working conditions.

[0091] Furthermore, after determining the anchor point density corresponding to the current parameter region, the spatial location of the terrain features of the blind area terrain points is identified, that is, the historical terrain vector corresponding to the blind area terrain points is determined to fall within the target parameter region enclosed by the original anchor points. Then, based on the anchor point density corresponding to the target parameter region, the number of clusters required for the clustering operation of the blind area terrain points is determined. The higher the anchor point density, the larger the number of clusters. A group of blind area terrain points with slightly different features are split into more subdivided clusters. Each cluster generates a blind area terrain vector and creates an independent new anchor point.

[0092] This third embodiment, building upon the ability of the second embodiment to autonomously identify blind zone terrain and expand the anchor point sample library, first traverses all anchor points in the terrain base library and matches the second neighbor anchor points of each anchor point. The sensitivity of each parameter region is calculated by the ratio of the terrain vector difference between the anchor point and its second neighbor anchor point to the control parameter difference. Then, anchor point density within the parameter region is allocated according to the sensitivity. Finally, after locating the blind zone terrain points in the parameter region of the feature space and reading the corresponding anchor point density, the cluster number is adaptively set. Based on this cluster number, the blind zone terrain points are clustered, generating a corresponding number of blind zone terrain vectors and new anchor points. This method can adaptively adjust the subdivision granularity of blind zone samples according to the sensitivity of terrain parameters. For highly sensitive terrain, more anchor points are added to ensure interpolation and control parameter accuracy, while for low-sensitivity areas, the number of anchor points is reduced to save storage and real-time computing power. While continuously and autonomously expanding the terrain base library, it dynamically balances control accuracy, walking stability, and system computing power overhead, further optimizing the adaptive control performance of the exoskeleton in various unfamiliar and complex terrains.

[0093] Example 4 Based on the above embodiments, in the method for determining control parameters of the human-machine collaborative embody system proposed in this embodiment, refer to Figure 4 Following step S40, steps S150 to S160 are also included: Step S150: When the mixed weight of any target anchor point is greater than the preset weight threshold, determine the control deviation index corresponding to the target control parameter.

[0094] Step S160: When the control deviation index is greater than the control deviation index of other anchor points, the recalibration of the control parameter set corresponding to the target anchor point is triggered or the target anchor point is deleted.

[0095] The control deviation index is a comprehensive evaluation value that quantifies the difference between the actual operating state of a wearable lower limb assistive exoskeleton and the ideal coordinated walking state. It is used to measure the fit of the current target control parameters. The larger the control deviation index, the worse the assistive coordination effect and the lower the fit.

[0096] The calculation methods for determining the control deviation index can include: collecting sensor data from the pressure-sensitive units of the left and right soles, extracting the line connecting the real-time pressure center coordinates of the left and right feet respectively, and obtaining the offset difference between this line and the reference line under the standard posture as the control deviation index; or calculating the real-time center of gravity position of the human body through the posture sensor, comparing it with the ideal center of gravity trajectory under the standard posture, and obtaining the average offset of the center of gravity position as the control deviation index; or separately calculating the total force value collected by the pressure-sensitive units of the left and right soles, and then calculating the absolute value of the force difference between the left and right soles as the control deviation index, the larger the force difference between the left and right soles, the more serious the force imbalance; or multiplying the above offset difference, the average offset of the center of gravity position, and the force difference between the left and right soles by a preset weighting coefficient, and summing them to obtain a comprehensive control deviation index, which is suitable for the accuracy verification requirements of all scenarios.

[0097] When the mixed weight of any target anchor point exceeds a preset weight threshold, it indicates that the target control parameters are mainly dominated by the control parameter set corresponding to that target anchor point. The actual operating state of the wearable lower limb assistive exoskeleton can reflect the adaptation performance of the control parameter set corresponding to that target anchor point. Under this condition, performing control deviation detection can identify whether the original control parameters of the target anchor point have problems such as aging, numerical deviation, or adaptation performance degradation.

[0098] For example, the human-machine collaborative embody system calculates the control deviation index of the target control parameters in real time under the current working state. Then, it obtains the historical control deviation indexes of all other anchor points in the terrain base database besides the target anchor point, and compares the control deviation index of the target anchor point with the historical control deviation indexes of other anchor points one by one. If the control deviation index of the target anchor point is higher than the historical control deviation of other anchor points, it is determined that the original control parameter adaptation performance of the target anchor point has deteriorated. Then, an instruction is issued to start the recalibration process of the control parameter set of the anchor point, readjust the control parameters adapted to the terrain vector corresponding to the target anchor point, or delete the target anchor point.

[0099] Based on the previous embodiments, this fourth embodiment adds a parameter self-checking and correction mechanism after the candidate parameter smoothing filter generates the target control parameters. When the human-machine collaborative embody system identifies a dominant anchor point whose mixed weight exceeds a preset threshold, it compares the real-time control deviation index of that anchor point with the historical control deviation index of other anchor points in the database. If the real-time control deviation index of that anchor point is higher, it automatically starts the recalibration of the control parameters of that anchor point, avoiding the problem of chaotic deviation tracing caused by multi-anchor point fusion, ensuring accurate and reliable detection results, and continuously automatically verifying and correcting aging and adaptability-degraded control parameters. This maintains the accuracy of the control parameters of each anchor point in the terrain base database for a long time, and steadily improves the balance and overall control effect of human-machine collaborative walking.

[0100] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the method for determining the control parameters of the applicant's machine-cooperative embodiment system. Any simple modifications based on this technical concept are within the scope of protection of this application.

[0101] This application provides a control parameter determination device for a human-machine collaborative embody system. The control parameter determination device for a human-machine collaborative embody system includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the control parameter determination method for the human-machine collaborative embody system in the first embodiment described above.

[0102] The following is for reference. Figure 5 This document illustrates a schematic diagram of a control parameter determination device suitable for implementing the human-machine collaborative embody system of the embodiments of this application. The control parameter determination device for the human-machine collaborative embody system in the embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, personal digital assistants (PDAs), tablet computers (PADs), portable media players (PMPs), and in-vehicle terminals (e.g., in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. Figure 5 The control parameter determination device for the human-machine collaborative embody system shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0103] like Figure 5As shown, the control parameter determination device of the human-machine collaborative embody system may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. The random access memory 1004 also stores various programs and data required for the operation of the control parameter determination device of the human-machine collaborative embody system. The processing unit 1001, the read-only memory 1002, and the random access memory 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the control parameter determination device of the human-machine collaborative embody system to exchange data wirelessly or via wired communication with other devices. Although the figure shows a control parameter determination device for a human-machine collaborative embody system with various systems, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems can be implemented or possessed alternatively.

[0104] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from read-only memory 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0105] The control parameter determination device for the human-machine collaborative embody system provided in this application, employing the control parameter determination method for the human-machine collaborative embody system described in the above embodiments, can solve the technical problem of improving the stability of human-machine collaborative walking. Compared with the prior art, the beneficial effects of the control parameter determination device for the human-machine collaborative embody system provided in this application are the same as the beneficial effects of the control parameter determination method for the human-machine collaborative embody system provided in the above embodiments, and other technical features in the control parameter determination device for the human-machine collaborative embody system are the same as those disclosed in the method of the previous embodiment, and will not be repeated here.

[0106] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0107] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0108] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the control parameter determination method of the human-machine collaborative embody system in the above embodiments.

[0109] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), or flash memory, optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, radio frequency (RF), etc., or any suitable combination thereof.

[0110] The aforementioned computer-readable storage medium may be included in the control parameter determination device of the human-machine collaborative embody system; or it may exist independently and not be assembled into the control parameter determination device of the human-machine collaborative embody system.

[0111] The aforementioned computer-readable storage medium carries one or more programs that, when executed by the control parameter determining device of the human-machine collaborative embody system, enable the control parameter determining device to write computer program code for performing the operations of this application in one or more programming languages ​​or a combination thereof. These programming languages ​​include object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, or as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (e.g., using an Internet connection provided by an Internet service).

[0112] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation that may be implemented in systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0113] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0114] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the control parameter determination method of the above-described human-machine collaborative embody system, thereby solving the technical problem of improving the stability of human-machine collaborative walking. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as the beneficial effects of the control parameter determination method of the human-machine collaborative embody system provided in the above embodiments, and will not be repeated here.

[0115] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the control parameter determination method for the human-machine collaborative embody system described above.

[0116] The computer program product provided in this application can solve the technical problem of improving the stability of human-machine cooperative walking. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as the beneficial effects of the control parameter determination method of the human-machine cooperative embody system provided in the above embodiments, and will not be repeated here.

[0117] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A method for determining control parameters of a human-machine collaborative embody system, characterized in that, The method for determining the control parameters of the human-machine collaborative embody system includes: Acquire multimodal sensing data, and construct a terrain semantic vector based on the multimodal sensing data; Determine the similarity between the terrain semantic vector and the terrain vector corresponding to each anchor point in the terrain base library, and determine the mixing weight of each anchor point based on the similarity. The control parameter sets corresponding to each anchor point are fused according to the hybrid weight to generate a candidate parameter set corresponding to the current terrain. The target filtering coefficients are determined based on the change magnitude of the terrain semantic vector, and the candidate parameter set is smoothed based on the target filtering coefficients to obtain the target control parameters.

2. The method for determining control parameters of the human-machine collaborative embody system as described in claim 1, characterized in that, The multimodal perception data includes ground point cloud data collected by a visual camera and pressure sensing data collected by a foot pressure-sensitive unit. The terrain semantic vector includes one or more of the following: forward slope, lateral tilt angle, ground roughness, ground continuity value, lateral load offset, ground hardness, and slope time stability value.

3. The method for determining control parameters of the human-machine collaborative embody system as described in claim 2, characterized in that, The step of constructing a terrain semantic vector based on the multimodal sensing data includes one or more of the following: Select a target ground point cloud within a preset range, determine the normal vector of the fitting plane of the target ground point cloud, and determine the forward slope based on the angle between the normal vector and the vertical direction; The normal vector is projected onto the horizontal plane to obtain a horizontal projection vector. The lateral tilt angle is determined based on the angle between the horizontal projection vector and the current horizontal direction of travel. Determine the vertical distance from each of the target ground point clouds to the fitted plane, and determine the ground roughness based on the standard deviation of the residuals of the vertical distances; Determine the density dispersion of the target ground point cloud, and determine the ground continuity value based on the density dispersion; Based on the pressure values ​​and position coordinates collected by the pressure-sensitive units on the left and right soles, the pressure centers of the left and right feet and the lateral offset of the line connecting the pressure centers are determined. Based on the ratio of the lateral offset to the preset reference lateral distance between the left and right feet, the lateral load offset is determined. Determine the rate of pressure change of the pressure value within a preset time period before and after the pressure-sensitive unit touches the ground, and determine the ground hardness based on the rate of pressure change; The standard deviation of the forward slope within a preset time window is determined, and the ground roughness is weighted and fused with the standard deviation to obtain the slope time stability value.

4. The method for determining control parameters of the human-machine collaborative embody system as described in claim 1, characterized in that, The step of determining the similarity between the terrain semantic vector and the terrain vector corresponding to each anchor point in the terrain base library, and determining the mixing weight of each anchor point based on the similarity, includes: The Euclidean distance between the terrain semantic vector and the terrain vector corresponding to each anchor point in the terrain base library is determined by a preset Gaussian kernel function, and the Euclidean distance is used as the similarity. The similarity corresponding to each anchor point is normalized and used as the mixed weight for each anchor point.

5. The method for determining control parameters of the human-machine collaborative embody system as described in claim 1, characterized in that, The method for determining the control parameters of the human-machine collaborative embody system further includes the following steps: Determine the geometric distance between the historical terrain vector and the terrain vector corresponding to each anchor point; If the geometric distance between the historical terrain vector and the terrain vectors corresponding to all anchor points is greater than a preset distance threshold, then the ground point cloud corresponding to the historical terrain vector is taken as a blind zone terrain point. Clustering the terrain points in the blind area yields the blind area terrain vector; Construct new anchor points in the terrain base library, and use the blind zone terrain vector as the terrain vector corresponding to the new anchor points; The set of control parameters corresponding to the new anchor point is determined based on the terrain vector of the blind zone.

6. The method for determining control parameters of the human-machine collaborative embody system as described in claim 5, characterized in that, The step of determining the control parameter set corresponding to the new anchor point based on the blind zone terrain vector includes: The first neighboring anchor point of the new anchor point is determined based on the geometric distance between the terrain vector of the blind zone and the terrain vector corresponding to each anchor point; The parameter weights of the first neighbor anchor points are determined based on the geometric distance corresponding to the first neighbor anchor point. The control parameter set corresponding to the first neighbor anchor point is fused according to the parameter weight to obtain the control parameter set corresponding to the new anchor point.

7. The method for determining control parameters of the human-machine collaborative embody system as described in claim 5, characterized in that, Before the step of clustering the blind area terrain points to obtain the blind area terrain vector, the following steps are also included: Traverse all anchor points in the terrain base library and determine the second neighbor anchor point of the current anchor point. Determine a first difference between the terrain vectors of the current anchor point and the second neighboring anchor point, and a second difference between the control parameter sets of the current anchor point and the second neighboring anchor point; The sensitivity of the current parameter region is determined based on the ratio of the first difference to the second difference, wherein the current parameter region is the parameter range of the control parameter set corresponding to the current anchor point and the second neighboring anchor point; Based on the sensitivity of the current parameter region, determine the corresponding anchor point density within the current parameter region; The number of clusters of the blind area terrain points is determined based on the anchor point density, and the step of clustering the blind area terrain points to obtain blind area terrain vectors is performed based on the number of clusters, wherein the number of blind area terrain vectors and the number of new anchor points are consistent with the number of clusters.

8. The method for determining control parameters of the human-machine collaborative embody system as described in claim 1, characterized in that, After the steps of determining the target filtering coefficients based on the change magnitude of the terrain semantic vector, and smoothing the candidate parameter set based on the target filtering coefficients to obtain the target control parameters, the method further includes: When the mixed weight of any target anchor point is greater than the preset weight threshold, the control deviation index corresponding to the target control parameter is determined. When the control deviation index is greater than the control deviation index of other anchor points, the recalibration of the control parameter set corresponding to the target anchor point is triggered or the target anchor point is deleted.

9. A device for determining control parameters of a human-machine collaborative embody system, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the method for determining control parameters of the human-machine collaborative embody system as described in any one of claims 1 to 8.

10. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the method for determining control parameters of the human-machine collaborative embody system as described in any one of claims 1 to 8.

Images