Robot attitude intelligent control method and system based on ad hoc network communication architecture

Abstract

The invention relates to the technical field of robot attitude control, and discloses a robot attitude intelligent control method and system based on an ad hoc network communication architecture, and the method comprises the steps: obtaining multi-dimensional environment initial data and current motion state data of a robot; performing multi-source data integration and key feature extraction operation according to the multi-dimensional environment initial data to obtain a fused environment feature set; according to the current motion state data of the robot and the fused environment feature set, risk identification and posture adjustment strategy generation operation are carried out, and a preliminary posture adjustment strategy signal is obtained; and performing parameter optimization and secondary calibration operation on the preliminary attitude adjustment strategy signal to obtain a final accurate attitude control instruction. According to the method, precise attitude control of the robot in a complex dynamic environment can be realized, and the operation safety and efficiency are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of robot posture control, and in particular to a robot posture intelligent control method and system based on an ad hoc network communication architecture. BACKGROUND

[0002] At present, in the field of modern intelligent manufacturing and automation, precise perception and intelligent control of complex dynamic environments are the core requirements for improving the efficiency and safety of robot operation, especially in the aspects of multi-dimensional information processing and real-time response.

[0003] In one prior art, environmental data is collected by a single or a few sensors, and then these data are input into a control algorithm for real-time analysis to determine the deviation between the current posture of the robot and the desired posture. Then, according to the analysis result, corresponding control instructions are generated to drive the actuator such as a motor or a hydraulic device to dynamically adjust the posture of the robot. However, the prior art has significant shortcomings in dealing with highly dynamic and complex actual scenarios. The environmental perception means cannot effectively integrate visual data from multiple angles, and cannot realize real-time acquisition and deep analysis of all-around information, which directly leads to limited accurate judgment of the posture of the robot itself. The lack of perception ability further restricts the real-time performance of posture control, making it difficult for the robot to quickly generate and execute effective posture adjustment instructions when facing sudden obstacles or terrain changes.

[0004] In summary, the prior art has the problem of insufficient accuracy of robot posture adjustment instructions in complex environments. SUMMARY

[0005] The present application relates to the field of robot posture control, and in particular to a robot posture intelligent control method and system based on an ad hoc network communication architecture, to solve the problem of insufficient accuracy of robot posture adjustment instructions in complex dynamic environments in the prior art.

[0006] In a first aspect, to solve the above technical problems, the present application provides a robot posture intelligent control method based on an ad hoc network communication architecture, comprising: acquiring multi-dimensional environmental initial data; performing multi-source data integration and key feature extraction operations according to the multi-dimensional environmental initial data to obtain a fused environmental feature set; performing multi-angle visual data space distribution and change trend analysis operations according to the environmental feature set to obtain a panoramic information view of the dynamic environment; dividing the panoramic information view of the dynamic environment into a set of divided view units, and identifying operation risk points according to the set of view units; If there is a potential operating risk point, a risk response trigger and a preliminary attitude adjustment strategy signal generation operation are performed to obtain a preliminary attitude adjustment strategy signal; According to the preliminary attitude adjustment strategy signal, an adjustment instruction parameter optimization operation is performed to obtain a final precise attitude control instruction.

[0007] In an optional implementation, the obtaining of the multi-dimensional environment initial data includes: Multi-angle data are acquired from the dynamic environment by a sensor array, and preprocessed to obtain an initial multi-dimensional data set; Visual information in the initial multi-dimensional data set is subjected to denoising and edge detection processing to extract clear visual feature data; According to distance information and terrain feature information in the initial multi-dimensional data set, spatial distribution in the environment is calculated to determine spatial structure data of the environment; If a matching degree of the spatial structure data and the visual feature data is lower than a preset matching degree threshold, calibration processing is performed on the two to obtain calibrated environment information matrix; The calibrated environment information matrix is subjected to interpolation filling and standardization processing to obtain a standardized environment perception data set; According to the environment perception data set, grouping processing is performed, and distribution characteristics of each group of data are judged to obtain multi-dimensional environment initial data.

[0008] In an optional implementation, according to the multi-dimensional environment initial data, multi-source data integration and key feature extraction operations are performed to obtain a fused environment feature set, including: The multi-dimensional environment initial data are subjected to data cleaning, and the cleaned data are subjected to timestamp synchronization processing and missing value filling to obtain a filled complete data set; According to the filled complete data set, each dimension of data is subjected to screening processing to acquire significant feature data related to environment perception; The significant feature data are subjected to classification processing to obtain a grouped feature data set; Through the grouped feature data set, different groups of data are subjected to integration processing, and if a matching degree of the integrated data is lower than a preset threshold, consistency adjustment is performed to obtain the fused environment feature set.

[0009] In an optional implementation, according to the environment feature set, multi-angle visual data spatial distribution and variation trend analysis operations are performed to obtain a panoramic information view of the dynamic environment, including: The spatial coordinate conversion processing is performed on the environment feature set to obtain point data corresponding to the spatial distribution, and region feature data before classification is obtained. The data is converted from the sensor local coordinate system to the global coordinate system to generate point data containing three-dimensional coordinates.

[0010] According to the region feature data before classification, classification processing is performed according to different regions to obtain a region feature set after classification. The region feature data after classification is subjected to change trend tracking processing to obtain dynamic environmental change data over time, and the integrity of the dynamic environmental change data is judged. If the integrity of the dynamic environmental change data is lower than a preset integrity threshold, the dynamic environmental change data is subjected to difference filling to obtain an adjusted dynamic environmental data set. Based on the adjusted dynamic environmental data set, a comprehensive visual network model is constructed to analyze the spatial distribution and change trend of multi-angle visual data, and a panoramic information view of the dynamic environment is obtained.

[0011] In an optional implementation, the panoramic information view of the dynamic environment is divided into a set of divided view units, and operation risk point identification is performed according to the set of view units, including: The terrain changes and obstacle regions in the panoramic information view are preliminarily divided to obtain at least one independent terrain unit and obstacle unit, and a set of divided view units is obtained. According to the set of view units, the terrain unit and the obstacle unit are subjected to detail feature grabbing, and the grabbed detail features are subjected to real-time classification to determine the specific category and attribute state of each unit. The specific category and the attribute state are matched with a pre-established threat database, and if the attribute state matches the potential threat features in the database, the unit is marked as a high-risk unit to obtain a set of marked risk units. According to the set of marked risk units, the operation safety region is dynamically adjusted, the panoramic information view of the dynamic environment is refreshed in real time in combination with a preset view update mechanism, and the distribution of potential operation risk points is judged.

[0012] In an optional implementation, if there is a potential operation risk point, a risk response trigger and a preliminary posture adjustment strategy signal generation operation are performed to obtain a preliminary posture adjustment strategy signal, including: If there is a potential operation risk point, the panoramic information view of the dynamic environment is divided into at least one risk region unit to obtain a set of divided risk regions. detailed capture is performed on the spatial position and risk category of a risk region unit in the risk region set, to determine the specific position coordinates and category label of the unit; If the category label matches a high-risk feature in a pre-established threat assessment database, a corresponding real-time response instruction is generated, and a response signal triggered by the real-time response instruction is obtained; According to the response signal, a target operation region is dynamically planned, and a preliminary adjustment signal is generated in combination with a preset posture adjustment logic; If the preliminary adjustment signal meets the preset safety range, the preliminary adjustment signal is determined as the preliminary posture adjustment strategy signal.

[0013] In an optional embodiment, the adjustment instruction parameter optimization operation is performed according to the preliminary posture adjustment strategy signal to obtain a final precise posture control instruction, including: According to the preliminary posture adjustment strategy signal, current motion state data of the robot is obtained; The current motion state data is compared with a pre-established task matching rule, and if the deviation of the current motion state data from a target task exceeds a preset deviation threshold, a preliminary adjustment instruction is generated to obtain a basic parameter set of instruction adjustment; According to the basic parameter set, the robot is dynamically adjusted to obtain dynamic adjustment data of the robot, and parameter optimization is performed in combination with a preset constraint condition to obtain a parameter optimization result; If the parameterization result meets the preset precise control standard, an intermediate parameter set suitable for the current scene is obtained; According to the intermediate parameter set, the robot is dynamically adjusted, and the robot state is monitored in real time to obtain the latest motion state information; If the motion state information does not meet the preset target task matching degree, secondary calibration is performed to obtain a final control parameter set; According to the final control parameter set, the posture is adjusted, and it is judged whether the adjusted state meets the preset target task requirement, and if so, a final precise posture control instruction is obtained.

[0014] In an optional embodiment, the sensor array includes a visual sensor, a distance sensor, and a terrain sensor, the visual sensor is used to collect the environmental visual information, the distance sensor is used to collect the environmental distance information, and the terrain sensor is used to collect the environmental terrain feature information.

[0015] In a second aspect, the application provides a robot posture intelligent control system based on a self-organizing network communication architecture, including: A data acquisition module is configured to acquire multi-dimensional environment initial data. A data fusion module is configured to perform multi-source data integration and key feature extraction based on the multi-dimensional environment initial data to obtain a fused environment feature set. A panoramic view construction module is configured to perform multi-angle visual data spatial distribution and variation trend analysis based on the environment feature set to obtain a panoramic information view of a dynamic environment. A risk identification module is configured to divide the panoramic information view of the dynamic environment into view unit sets, and identify operation risk points based on the view unit sets. A preliminary strategy generation module is configured to generate a risk response trigger and a preliminary attitude adjustment strategy signal if there is a potential operation risk point, and obtain a preliminary attitude adjustment strategy signal. A precise instruction determination module is configured to perform adjustment instruction parameter optimization based on the preliminary attitude adjustment strategy signal to obtain a final precise attitude control instruction.

[0016] In a third aspect, the present application further provides an electronic device comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the robot attitude intelligent control method based on the self-organizing network communication architecture according to any one of the above.

[0017] In a fourth aspect, the present application further provides a computer readable storage medium comprising a stored computer program, wherein the computer program controls a device where the computer readable storage medium is located to execute the power control method for subway emergency lighting according to any one of the above when the computer program is running.

[0018] Compared with the prior art, the present application has the following beneficial effects: (1) The multi-source heterogeneous data collected by the present application from a dynamic environment is filtered for noise, time stamps are aligned and missing values are completed, and then the environment data dynamic trend is integrated and tracked through spatial coordinate conversion and region classification modeling. Finally, high-precision perception of dynamic environment panoramic information is achieved, the robot can more accurately grasp the operation scene layout and element motion state, and the problems of one-sidedness of traditional single sensor data and lagging environment response are solved, thereby laying a solid data foundation for risk identification and attitude adjustment.

[0019] (2) The application divides the dynamic environment panoramic view by region growing algorithm, extracts and classifies features to determine the unit attribute, matches and marks high-risk units in the threat database, synchronously adjusts the safe area and refreshes the view in real time. Finally, the dynamic and accurate identification of potential operational risk points is realized, the robot's risk prediction and avoidance ability is improved, the problem of relying on experience in traditional risk identification and response lag is solved, and the safety barrier of complex scene operation is built.

[0020] (3) The application collects robot motion state data, compares with task rules, generates a basic parameter set when the deviation exceeds the limit, optimizes the basic parameter set by PID algorithm, multi-round calibration and safety verification. Finally, the high-precision iteration of the posture control instruction is realized, the robot operation posture is adapted to the task demand, the problem of traditional instruction roughness and difficulty in adapting to dynamic scene is solved, and the operation efficiency and quality are improved. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 is a flowchart of the robot posture intelligent control method based on the self-organizing network communication architecture provided by the first embodiment of the application; Figure 2 is a schematic diagram of the structure of the robot posture intelligent control system based on the self-organizing network communication architecture provided by the second embodiment of the application. DETAILED DESCRIPTION

[0022] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the application.

[0023] Referring to Figure 1 The first embodiment of the application provides a robot posture intelligent control method based on a self-organizing network communication architecture, including the following steps: S101, obtaining multi-dimensional environment initial data; S102, performing multi-source data integration and key feature extraction operation according to the multi-dimensional environment initial data, to obtain a fused environment feature set; S103, performing multi-angle visual data space distribution and change trend analysis operation according to the environment feature set, to obtain a panoramic information view of the dynamic environment; S104, dividing the panoramic information view of the dynamic environment into regions to obtain a set of divided view units, and identifying operational risk points according to the set of view units; S105, if there is a potential operating risk point, a risk response trigger and a preliminary attitude adjustment strategy signal generation operation are performed to obtain a preliminary attitude adjustment strategy signal; S106, according to the preliminary attitude adjustment strategy signal, an adjustment instruction parameter optimization operation is performed to obtain a final accurate attitude control instruction.

[0024] In step S101, multi-dimensional environment initial data is acquired.

[0025] In an implementable manner, the acquisition of multi-dimensional environment initial data includes: Multi-angle data is acquired from a dynamic environment by a sensor array and preprocessed to obtain an initial multi-dimensional data set; Visual information in the initial multi-dimensional data set is denoised and edge detected to extract clear visual feature data; According to distance information and terrain feature information in the initial multi-dimensional data set, the spatial distribution in the environment is calculated to determine the spatial structure data of the environment; If the matching degree of the spatial structure data and the visual feature data is lower than a preset matching degree threshold, calibration processing is performed on both to obtain a calibrated environment information matrix; The calibrated environment information matrix is interpolated and standardized to obtain a standardized environment perception data set; According to the environment perception data set, grouping processing is performed, and the distribution characteristics of each group of data are judged to obtain multi-dimensional environment initial data.

[0026] It should be noted that multi-angle data in a dynamic environment is synchronously collected by a sensor array, which includes a visual sensor, a distance sensor, and a terrain feature sensor. After collection, the visual information is classified by resolution, the distance information is classified by measurement range, and the terrain feature information is classified by type to form an initial multi-dimensional data set. For example, in an autonomous driving scenario, a camera captures a road image, a laser radar measures the distance of a front obstacle (e.g., 10.5 meters), and an inertial measurement unit combines a digital map to obtain a slope and a road roughness to form a structured data set after classification.

[0027] Then, the visual information in the initial data set is denoised to filter out noise interference caused by light fluctuations, and then contour features are extracted through an edge detection algorithm to obtain clear visual feature data. For example, after road images are denoised by Gaussian filtering, continuous lane line edge contours are extracted by a Canny operator.

[0028] Meanwhile, based on distance information and terrain feature information, the spatial distribution of the environment is calculated through a triangulation algorithm. Specifically, by combining laser radar ranging data and terrain slope information, the three-dimensional spatial coordinates of the obstacles are calculated, and the spatial structure data of the environment is generated. For example, the laser radar measures the distance of an obstacle to be 10.5 meters, and combined with the 5-degree slope data, the precise position of the obstacle in the vehicle coordinate system is calculated (X=10.5m, Y=2.3m, Z=0.2m).

[0029] If the matching degree of the spatial structure data and the visual feature data is lower than the preset threshold (such as the position deviation exceeds 5 meters), calibration processing is performed: first, the time stamps of the two types of data are aligned to ensure time synchronization; then the position deviation is corrected through spatial coordinate transformation method. For example, when the position deviation of the rectangular profile of the obstacle detected by vision and the laser radar positioning reaches 5 meters, after aligning the data time stamps, the visual coordinates are mapped to the spatial coordinate system through affine transformation to realize position calibration.

[0030] For the missing areas in the calibrated environment information matrix, linear interpolation is used to fill in the missing values based on the surrounding effective data. Then standardization processing is performed: the visual data is converted to a gray matrix in the range of 0-255, the distance data is uniformly converted to meters, and the terrain data is normalized to a slope coefficient in the range of 0-1. For example, the distance data of a certain area is missing, and the missing value is estimated to be 10.5 meters according to the data of the previous frame of 12 meters and the next frame of 9 meters; the terrain slope of 15% is normalized to 0.15.

[0031] Finally, the standardized data set is grouped according to the feature dimensions, such as dividing the distance data into near field (0-5 meters), middle field (5-20 meters), etc. groups, analyzing the distribution characteristics of each group (such as detecting 3 obstacles in the near field group and distributing densely), and generating the final multi-dimensional environment initial data.

[0032] In summary, through multi-source data classification collection, visual feature optimization extraction, spatial structure accurate calculation, cross-modal data calibration, missing value interpolation and standardization grouping processing, the integrity and reliability of the environment perception data are significantly improved. For example, in automatic driving test, the obstacle on the 15% slope 10.5 meters in front can be accurately identified, providing high-precision input for subsequent real-time attitude control, effectively solving the problem of fragmented perception data in dynamic environment. The standardization and grouping mechanism further accelerates the analysis efficiency of the system on environmental features, supporting rapid decision-making in complex scenarios.

[0033] In step S102, multi-source data integration and key feature extraction operations are performed according to the multi-dimensional environment initial data, and a fused environment feature set is obtained.

[0034] In an implementable manner, according to the multi-dimensional environment initial data, multi-source data integration and key feature extraction operations are performed to obtain a fused environment feature set, including: The multi-dimensional environment initial data is subjected to data cleaning, and the cleaned data is subjected to timestamp synchronization processing and missing value filling to obtain a filled complete data set; According to the filled complete data set, each dimension data is subjected to screening processing to obtain significant feature data related to environment perception; The significant feature data is subjected to classification processing to obtain a grouped feature data set; Through the grouped feature data set, the data of different groups is subjected to integration processing, and if the matching degree of the integrated data is lower than a preset threshold, consistency adjustment is performed to obtain the fused environment feature set.

[0035] It should be noted that the multi-dimensional environment initial data is subjected to data cleaning to filter out redundant noise caused by sensor abnormalities or environmental interference. Specifically, for isolated points in distance data (such as invalid values exceeding the effective range) and fuzzy areas in visual data, the cleaned data set is formed. For example, in the automatic driving scene, if the point cloud data collected by the laser radar has invalid ranging points exceeding 100 meters, it is directly excluded; the patch noise caused by strong light reflection in the image is filtered and removed through the pixel threshold.

[0036] The cleaned data set is subjected to timestamp synchronization processing to align the sensor data of different collection frequencies to a unified time reference. If there is a missing time node, the missing value is filled by linear interpolation based on the valid data of adjacent time periods to generate a filled complete data set. For example, when the camera collects at 30 frames / second and the laser radar collects at 10 times / second, the radar data is matched to the nearest frame image timestamp; if the radar data is missing at a certain time, the 10-meter distance data at that time is generated by interpolation based on the 9-meter and 11-meter ranging values of the previous and next frames.

[0037] According to the filled complete data set, significant feature data directly related to environment perception is screened: obstacle edge contour and lane line features are extracted from visual data, key obstacle information within a preset range (such as targets within 20 meters) is retained from distance data, and regions with slope changes exceeding a set threshold (such as a slope difference ≥ 5%) are identified from terrain data. For example, in a road environment, the pedestrian contour features in the image, the 15-meter vehicle distance data detected by the laser radar, and the road segment information with a slope change exceeding 8% are extracted.

[0038] The screened significant feature data is classified and processed according to spatial distribution or functional attributes: distance data is divided into a near distance group (0-5 meters) and a middle distance group (5-20 meters), and visual features are divided into a static object group (such as traffic signs) and a dynamic object group (such as pedestrians), to form a feature data set after grouping. For example, a pedestrian 3 meters in front is classified into a “near distance dynamic object group”, and a stationary vehicle 20 meters away is classified into a “middle distance static object group”.

[0039] The feature data of different groups is integrated and processed, and if the matching degree of the integrated data is lower than a preset threshold (such as the deviation of radar positioning from vision is more than 2 meters), dynamic consistency adjustment is performed: the data sources are aligned through spatial coordinate transformation, and finally a fused environmental feature set is generated. For example, when the position of a pedestrian identified by image recognition deviates from radar positioning by 2.5 meters, the visual coordinates are mapped to a unified space system through affine transformation based on radar coordinates, to form a “3-meter dynamic pedestrian in front” feature item with consistent position.

[0040] In summary, the data quality is improved by noise filtering, the time sequence consistency is ensured by time synchronization, the key environmental elements are locked by feature screening, the data structure is optimized by classification processing, and the cross-sensor deviation is solved by dynamic calibration, which significantly enhances the accuracy of environmental perception. For example, in autonomous driving, the integrated feature set can accurately describe the complex scene of “a stationary vehicle on a road section with a 8% slope 15 meters in front”, providing high reliability input for subsequent risk judgment. The efficient fusion mechanism of multi-source data is especially suitable for rapid identification of sudden obstacles in dynamic environments, solving the problem of decision lag caused by data fragmentation in traditional methods.

[0041] In step S103, according to the environmental feature set, a multi-angle visual data spatial distribution and change trend analysis operation is performed to obtain a panoramic information view of the dynamic environment.

[0042] In an optional implementation, the multi-angle visual data spatial distribution and change trend analysis operation according to the environmental feature set to obtain a panoramic information view of the dynamic environment includes: According to the environmental feature set, a multi-angle visual data spatial distribution and change trend analysis operation is performed to obtain a panoramic information view of the dynamic environment.

[0043] The environmental feature set is subjected to spatial coordinate conversion processing to obtain point data corresponding to spatial distribution, to obtain region feature data before classification; According to the region feature data before classification, classification processing is performed according to different regions to obtain a region feature set after classification; The region feature data after classification is subjected to change trend tracking processing to obtain dynamic environmental change data over time, and the integrity of the dynamic environmental change data is judged. If the integrity of the dynamic environment change data is lower than a preset integrity threshold, the dynamic environment change data is interpolated to obtain an adjusted dynamic environment data set. It should be noted that when the spatial coordinate conversion processing is performed on the environment feature set, the local coordinate system data of different sensors needs to be uniformly converted to the global coordinate system to generate point data containing three-dimensional coordinates (such as X, Y, and Z axis coordinates). The coordinate system is aligned through a homogeneous coordinate transformation matrix, and the two-dimensional pixel coordinates (u, v) collected by the camera are combined with the depth information to convert them into three-dimensional space coordinates (x, y, z) centered on the robot body. For example, in an autonomous driving scenario, the camera detects an obstacle pixel point with image coordinates (320, 240), and combines the 10.5-meter distance data measured by the laser radar to calculate the actual spatial position of the obstacle in the vehicle coordinate system as (10.5, -2.3, 0.2) meters through the calibrated camera intrinsic matrix and extrinsic matrix, forming the regional feature data before classification.

[0044] When the regional feature data before classification is classified according to different regions, it is divided into forward, lateral, and rearward regions according to the spatial orientation, wherein the forward region can be defined as the range of 0-30 meters in front of the robot, and the lateral region is the range of 5 meters on the left and right, each region containing a sub-set of vision, distance, and terrain features. For example, the forward region sub-set can include the visual profile of the forward obstacle, the ranging data at a distance of 10.5 meters, and the terrain information with a slope of 5 degrees, thereby forming the regional feature set after classification.

[0045] When tracking the change trend of the classified regional feature data, a time series model is established for the classified regional feature data, and the Kalman filter or recursive least squares method is used to track the motion trajectory of the feature points. The position and velocity vector of the obstacle in each region are collected at a fixed time interval (such as 100 ms) to generate a sequence of dynamic environment change data. For example, if the coordinates of a certain obstacle in the forward region are detected as (20, 0)→(18, 0)→(16, 0) meters in 5 consecutive frames, it is determined that it is approaching at a speed of 2 meters / second, and the trend data is recorded. Set the integrity threshold (such as 85%), and calculate the effective sampling rate of the dynamic environment change data within a unit time window. If more than 2 out of 10 consecutive sampling points are missing (integrity 80%<85%), the missing values are reconstructed based on the previous and subsequent valid data using the cubic spline interpolation method. For example, the obstacle positions at t1-t3 are P1(10, 2), P2 (missing), and P3(8, 2), and according to the coordinates and velocity direction of P1 and P3, the coordinates at P2 are interpolated to be (9, 2) meters, forming a complete dynamic environment data set.

[0046] Based on the adjusted dynamic environment data set, the regional feature data is fused to construct a full-view visual network model in the following specific ways: First, for the regional feature data (including images, distances, speeds, etc.) collected by each node in the ad hoc communication architecture, an adaptive weighted fusion algorithm is used to assign dynamic weights to different regional data according to the credibility of the data source (such as node device accuracy, communication signal strength), for example, higher weights are given to obstacle distance data collected by high-precision laser radar nodes, and appropriate weights are reduced for visual sensor data affected by shielding; Second, the attention mechanism module is introduced to focus on the feature data of the key areas around the robot (such as the 10-meter range in front), and to strengthen the feature extraction of dynamic target dense areas; Finally, integrate different regional data, combined with multi-layer convolutional neural network (CNN) to extract spatial features of each region (such as obstacle shape, size), which includes input layer (receiving fused feature map), 3 convolutional layers (32, 64, 128 3x3 convolutional kernels are used to extract features respectively), and 2 pooling layers (2x2 max pooling), while the 2-layer LSTM of recurrent neural network (RNN) processes the time correlation information (such as the position change of continuous motion of the target), without additional training, directly using the preset network structure parameters for feature fusion, to construct a full-view visual network model with spatial perception and time correlation ability.

[0047] When combining time series analysis to label the motion trajectory of dynamic targets, Kalman filter algorithm is used to smooth the historical position data of dynamic targets: taking position and speed as state variables, state transition matrix is constructed according to the sampling period of the sensor (such as 100ms). Set the noise covariance matrix corresponding to the measurement accuracy of the sensor, in the prediction stage, according to the state transition matrix, the predicted state and covariance at the next time are calculated; in the update stage, combined with the position data collected by the current sensor, the optimal estimate value is calculated through Kalman gain, and iteration optimization is performed to eliminate measurement noise. Based on the historical trajectory, the motion trend within 3 seconds in the future is predicted, and then a dynamic environment panoramic information view containing spatial distribution and time trend is generated. Taking the front area as an example, this view integrates the calibrated obstacle spatial distribution (such as coordinates (9, 2) meters), motion trend (-1 meter / second along the X axis), and regional classification label (high-risk area), and visualizes the 360-degree environment state around the robot in the bird's eye projection mode, providing a comprehensive environment basis for subsequent risk identification.

[0048] In summary, this process realizes the unified calibration of environmental feature data in the spatial dimension and the continuous tracking in the time dimension, effectively integrates the spatial distribution and change trend of multi-angle visual data, provides comprehensive and accurate environmental basis for subsequent risk identification and posture adjustment, and significantly improves the perception ability and adaptation efficiency of robots in complex dynamic environments.

[0049] In step S104, the panoramic information view of the dynamic environment is regionally divided to obtain a set of divided view units, and operation risk point identification is performed according to the set of view units.

[0050] In an implementable manner, the panoramic information view of the dynamic environment is regionally divided to obtain a set of divided view units, and operation risk point identification is performed according to the set of view units, including: The terrain changes and obstacle regions in the panoramic information view are preliminarily divided to obtain at least one independent terrain unit and obstacle unit, thereby obtaining the set of divided view units; According to the set of view units, detailed features of the terrain units and the obstacle units are captured, and the captured detailed features are classified in real time to determine the specific category and attribute state of each unit; The specific category and the attribute state are matched with a pre-established threat database, and if the attribute state matches the potential threat features in the database, the unit is marked as a high-risk unit to obtain a set of marked risk units; According to the set of marked risk units, the operation safety region is dynamically adjusted, and the panoramic information view of the dynamic environment is refreshed in real time in combination with a preset view updating mechanism to determine the distribution of potential operation risk points.

[0051] It should be noted that based on the pixel distribution characteristics of the panoramic information view, the terrain change region and the obstacle region are automatically divided by a pre-trained semantic segmentation convolutional neural network model. According to the pixel depth difference and texture continuity, the model aggregates spatially adjacent and feature-similar pixels into independent units: terrain units (such as continuous slope regions) and obstacle units (such as closed contour regions). For example, in the panoramic view of autonomous driving, the model divides the region with a depth value mutation 10 meters in front into a "gully terrain unit", and identifies the moving elliptical contour on the left as a "pedestrian obstacle unit", thereby generating a set of view units containing multiple independent units.

[0052] It should be noted that the core of the "pre-trained semantic segmentation convolutional neural network model" is to use a deep convolutional neural network to perform semantic classification and region division on the pixel features of the panoramic information view to accurately divide the terrain change region and the obstacle region.

[0053] Model structure: The semantic segmentation convolutional neural network model adopted by the present application is an improved U-Net architecture. The input layer of the network receives pixel data of the panoramic information view (including RGB color channels, depth channels and texture feature channels), each channel corresponds to a normalized pixel feature parameter (such as pixel brightness value, depth coordinate value, texture gradient value). The encoder part of the network contains 4 layers of convolution modules, each layer is composed of two 3x3 convolution kernels, a BatchNormalization layer and a ReLU activation function, which gradually extracts the local features and global context information of the pixels, and realizes feature dimension reduction through the maximum pooling layer; the decoder part is symmetrical to the encoder, restores the feature map size through upsampling operation, and performs jump connection with the feature map of the corresponding layer of the encoder, fuses the feature information of different scales; finally, the output layer with 1x1 convolution kernel generates a semantic segmentation mask consistent with the size of the input view, wherein each pixel point corresponds to the class label of the terrain or obstacle (such as "flat ground", "slope", "pedestrian", "vehicle", etc.).

[0054] Model training: The training of the model is supervised learning. The training data comes from the panoramic environment views collected by the robot in different scenes and the manually labeled semantic segmentation labels (including terrain types and obstacle categories). The pixel data of the panoramic environment view is used as the input, and the manually labeled semantic segmentation label is used as the target output. The cross-entropy loss function is used to calculate the difference between the predicted label and the true label, and the SGD optimizer (the learning rate is initially set to 0.001, and is reduced to 0.1 of the original every 10 epochs) is used to perform iterative training on a large-scale dataset. The generalization ability of the model is improved through data enhancement techniques such as random cropping, rotation and brightness adjustment, until the loss function converges and the intersection over union (IoU) on the validation set reaches a preset threshold (such as 0.85 or higher). The trained model can automatically identify and segment the terrain change area and obstacle area from new panoramic information views.

[0055] When capturing detailed features from the view unit set, the terrain unit extracts the slope angle (by calculating the arctangent value of the unit boundary point depth coordinate), the surface curvature (based on the change rate of the normal vector of the adjacent triangular surface); the obstacle unit extracts the motion speed (the displacement of the unit centroid in two consecutive frames divided by the time difference), the size parameter (the volume of the smallest circumscribed cube). The feature vector is input into the cascaded classifier, the first level is based on support vector machine to distinguish the unit class (terrain / obstacle), and the second level is through decision tree to subdivide the small class (such as slope / steps, pedestrian / vehicle). For example, if the unit slope angle is 15° and the curvature is greater than 0.25, it is classified as "steep terrain unit"; if the moving unit size is 0.5m and the speed is 1.5m / s, it is classified as "riding personnel obstacle unit".

[0056] When the specific category and attribute state are matched with the threat database, the quantified threat feature threshold is stored in the database, such as "a slope > 12° ground unit is at risk of overturning" and "a moving obstacle with a distance < 3 m and a speed > 1.5 m / s is at risk of collision". If a certain obstacle unit is 2.5 m away from the robot and has a speed of 2 m / s, its attribute state matches the "high-risk collision risk" feature in the database, and it is marked as a high-risk unit, forming a risk unit set.

[0057] Based on the spatial distribution of the risk unit set, the safety area is dynamically divided: taking the robot position as the base point and the risk unit coordinates as the core, a dangerous influence area (such as a circular exclusion zone with a radius of unit speed x response time) is generated, and a polygon safety area for generating an avoidance path is generated in the reachable space. At the same time, incremental view updating is started: every 100 milliseconds, new laser radar point cloud data is fused, and the panoramic view terrain model is updated through triangular mesh reconstruction.

[0058] The panoramic view terrain model is updated through the following specific process: first, for the newly collected laser radar point cloud data, a voxel grid-based downsampling algorithm is used for data reduction, and the point cloud resolution is set to 0.05 meters to retain key feature points, while redundant points are removed to reduce data volume; second, the ground and non-ground points in the point cloud are segmented by the random sample consensus (RANSAC) algorithm to exclude dynamic target interference and focus on terrain structure data; then, based on the incremental Poisson surface reconstruction algorithm, the new point cloud data and the panoramic view terrain model are fused: first, a directed point set containing new point clouds and historical model vertices is constructed, the normal vector of each point is calculated by fitting a plane through the neighborhood points and determining the direction, which is used as the gradient constraint condition of the Poisson equation; then, a three-dimensional voxel grid is established, and the Poisson equation system is constructed with the grid nodes as unknowns, and the source term of the equation is determined by the projection of the normal vector of the directed point on the grid; the implicit signed distance function (the function value is negative indicating the inside of the model, and positive indicating the outside) is obtained by solving the equation; then, the initial triangular mesh that fuses the new point cloud features is generated by extracting the isosurface of the signed distance function combined with the triangular mesh boundary of the panoramic view terrain model; for the overlapping area of the initial mesh and the historical model, matching and alignment are achieved by calculating the vertex distance, the higher precision part (such as recent updated terrain details) in the historical model is retained, only the mesh in the newly added point cloud coverage area is reconstructed, and seamless fusion of new and old data is achieved, and the terrain surface details are updated through triangular mesh topology optimization (such as edge shrinkage and vertex smoothing) to ensure model continuity; finally, a time decay factor is introduced to reduce the weight of point cloud data in areas that have not been updated for more than 3 seconds, and the latest observed terrain features are preferentially retained, thereby realizing dynamic updating of the panoramic view terrain model.

[0059] For example, when a new rolling stone unit is detected in the right front (coordinates (5, 7), speed 3 m / s), a circular exclusion zone with a diameter of 6 meters is generated in the safety area, and the rolling stone trajectory prediction area is displayed in the updated panoramic view as a heat map (orange gradient area).

[0060] Through the above steps, the whole-process quantitative processing from region division to risk positioning is realized, which not only solves the one-sidedness of single feature recognition, but also ensures the timeliness of risk identification through real-time updating mechanism, and provides high-reliability risk coordinates and level basis for subsequent triggering of accurate attitude adjustment strategy.

[0061] In step S105, if there is a potential operating risk point, a risk response trigger and a preliminary attitude adjustment strategy signal generation operation are performed to obtain a preliminary attitude adjustment strategy signal.

[0062] In an implementable manner, if there is a potential operating risk point, a risk response trigger and a preliminary attitude adjustment strategy signal generation operation are performed to obtain a preliminary attitude adjustment strategy signal, including: If there is a potential operating risk point, the panoramic information view of the dynamic environment is regionally divided to obtain at least one risk region unit, and a divided risk region set is obtained; The spatial position and risk category of the risk region unit in the risk region set are captured in detail to determine the specific position coordinates and category label of the unit; If the category label matches the high-risk features in the pre-established threat evaluation database, a corresponding real-time response instruction is generated, and a response signal triggered by the real-time response instruction is obtained; According to the response signal, the target operating area is dynamically planned, and a preliminary adjustment signal is generated in combination with a preset attitude adjustment logic; It is judged whether the preliminary adjustment signal meets the preset safety range, and if so, the preliminary adjustment signal is determined as the preliminary attitude adjustment strategy signal.

[0063] It should be noted that based on the panoramic information view of the dynamic environment, the region is divided. The specific execution process is: first, the panoramic view is scanned at the pixel level, and the pixel cluster of the terrain mutation or obstacle profile is identified by the edge detection operator. For the detected pixel cluster, calculate its minimum enclosing rectangle area, if the area exceeds the preset threshold (for example 0.5 square meters), it is determined as an effective risk area unit. Each risk area unit records its geometric center point coordinates, boundary vertex coordinates and region type identifier. For example, in the automatic driving scene, when an irregular pit is detected in front of the road, the system extracts the pit edge pixels through the Sobel operator, calculates the minimum enclosing rectangle area of 1.2 square meters (1.8 meters long x 0.7 meters wide), generates a risk unit numbered #R01 and records the center point coordinates as the vehicle coordinate system (X=15.3m, Y=0.2m), and marks it as "road defect" type.

[0064] Feature extraction is performed on the divided risk region set: for the spatial position of each risk unit, the absolute coordinates are converted to polar coordinate system data (distance d, azimuth angle θ) relative to the robot body through the coordinate transformation matrix. At the same time, the unit type identifier is analyzed, and the dynamic risk parameters are calculated combined with real-time motion state data (such as the current speed v of the robot, the attitude angle φ). Taking the moving obstacle as an example, if the coordinates of the lateral region unit #R02 (type "moving object") are (d=3.5m, θ=45°), combined with the current speed v=1.5m / s of the robot and the object motion trajectory, the minimum relative distance d_min=1.8m within 2 seconds in the future is predicted. The parameter is matched with the threat assessment database: the database pre-stores the corresponding relationship between various high-risk features and response instructions, for example, the high-risk feature item "d_min<2m and θ∈[30°, 150°]", the current parameter completely matches the item, and triggers the obstacle avoidance response instruction of code #Alert02. The database adopts a tree index structure, and the matching response time is controlled within 10ms.

[0065] According to the real-time response instruction, dynamic path planning is performed: taking the current position of the robot as the starting point of the path and the target operation area as the end point, a two-dimensional grid environment model (grid resolution 0.1m x 0.1m) is established.

[0066] The two-dimensional grid environment model is established through the following specific process: first, based on the segmented ground cell and obstacle cell data in the panoramic information view, the 360-degree environment space around the robot is divided into uniform grid cells with a size of 0.1m*0.1m, each grid corresponds to a small area in the actual physical space; second, each grid is assigned a multi-dimensional attribute label, including terrain attributes (such as "flat", "slope", "gully", determined based on the slope and flatness data of the ground cell), obstacle attributes (such as "no obstacle", "static obstacle", "dynamic obstacle", labeled according to the type and motion state of the obstacle cell), risk level (such as "low risk", "medium risk", "high risk", calculated in combination with the influence range and motion trend of the risk cell); finally, through the self-organizing network communication architecture, the updated environment data of each node is synchronized in real time, and the grid attributes are dynamically corrected, for example, when a new dynamic obstacle is detected in a certain grid, its obstacle attribute is immediately updated to "dynamic obstacle" and the risk level is increased, thereby constructing a two-dimensional grid environment model that can reflect the changes in the environment in real time.

[0067] An improved A* algorithm is used for path search, in which the heuristic function increases the risk cell penalty term: is the spatial coordinates of the current path node, is the spatial coordinates of the target endpoint, is the distance from node n to the kth risk cell, is the cell risk coefficient, and are the path length weight coefficient and risk avoidance intensity coefficient, respectively set based on the motion efficiency requirement of the robot and the safety level requirement, for example, 0.8 and 0.7. For example, for the #R01 pit cell ( =0.8), a safety buffer of 0.5m from its boundary is automatically generated when planning the path. After planning, the attitude parameters are calculated: for wheeled robots, based on the differential drive model, the left and right wheel speed difference Δω=K·Δθ (K is the steering gain coefficient, Δθ is the path yaw angle); for legged robots, the joint angle sequence is calculated through inverse kinematics. Finally, the preliminary adjustment signal containing the target pose (x, y, φ) and motion parameters (v, ω) is generated.

[0068] ​Safety check on the generated preliminary adjustment signal: First, establish a robot dynamics constraint model, for example, the maximum centripetal acceleration constraint of a wheeled robot is a_max≤μ·g (μ is the friction coefficient, g is the acceleration of gravity). For steering command ω, the check condition is |v·ω|≤a_max; for speed command v, the acceleration change rate |Δv / Δt|≤j_max (j_max is the maximum jerk). For example, when the planned path requires the vehicle to complete an 8° turn (ω=0.14 rad / s) in 1 second and the speed v=0.8 m / s, the centripetal acceleration a=v·ω=0.112 m / s Under dry road conditions (μ=0.8, a_max=7.84 m / s ), the constraints a≤a_max and acceleration change rate |Δv / Δt|≤0.4 m / s ≤3 m / s are satisfied, and the signal is determined to be valid. If the check fails, start the command iterative optimization: adjust ω and v by gradient descent method until all safety constraints are satisfied.

[0069] In summary, this complete process allows the robot to quickly and reasonably generate an initial risk response posture command in a complex dynamic scenario through hierarchical identification, precise matching, dynamic planning, and safety checking, laying a solid foundation for subsequent command optimization and precise control, effectively improving the timeliness and scientificity of the robot's response to potential risks, and ensuring safe operation.

[0070] In step S106, according to the preliminary posture adjustment strategy signal, an adjustment instruction parameter optimization operation is performed to obtain a final precise posture control instruction.

[0071] In an implementable manner, according to the preliminary posture adjustment strategy signal, an adjustment instruction parameter optimization operation is performed to obtain a final precise posture control instruction, including: According to the preliminary posture adjustment strategy signal, the current motion state data of the robot is obtained; The current motion state data is compared with the pre-established task matching rule, and if the deviation of the current motion state data from the target task exceeds the pre-set deviation threshold, a preliminary adjustment instruction is generated to obtain a basic parameter set of instruction adjustment; According to the basic parameter set, the robot is dynamically adjusted to obtain dynamic adjustment data of the robot, and parameter optimization is performed in combination with the pre-set constraint condition to obtain a parameter optimization result; It is judged whether the parameterized result meets the pre-set precise control standard, and if it meets, an intermediate parameter set suitable for the current scenario is obtained; According to the intermediate parameter set, the robot is dynamically adjusted, and the latest motion state information is obtained by monitoring the robot state in real time. If the motion state information does not match the preset target task, secondary calibration is performed to obtain a final control parameter set; According to the final control parameter set, posture adjustment is performed, and it is determined whether the adjusted state meets the preset target task requirement. If yes, a final precise posture control instruction is obtained.

[0072] It should be noted that the current motion state data is obtained in real time by the state monitoring module of the robot. The data includes physical quantities such as joint angle, linear velocity, angular velocity, and acceleration. For example, in the industrial robot arm scene, the real-time angle of each joint is collected by a high-precision encoder (for example, the angle of joint 1 is 35.2°±0.1°), and the three-dimensional space velocity (for example, the X-axis direction velocity is 0.48 m / s) and the attitude angle (for example, the pitch angle is 2.3°) of the end effector are obtained by an inertial measurement unit (IMU). These data are updated at a sampling frequency of 100 Hz to form a dynamic state data set.

[0073] The current motion state data is compared with the preset task matching rule. The task matching rule defines the key parameter threshold of the target task, for example, the position error threshold of the target trajectory tracking is ±0.05 m, and the velocity error threshold is ±0.1 m / s. If it is detected that the deviation exceeds the threshold (for example, the measured velocity 0.48 m / s is lower than the target velocity 0.6 m / s, and the deviation 0.12 m / s exceeds the allowed 0.1 m / s), the built-in control algorithm is started to generate a preliminary adjustment instruction. Specifically, by establishing a robot dynamics model, a proportional-integral-derivative (PID) control algorithm is used to calculate the compensation amount: based on the velocity deviation value 0.12 m / s, the motor torque increment is calculated in combination with the robot mass and the friction coefficient, and the basic parameter set {torque increment: +6.4 N·m, duration: 0.3 s} is generated.

[0074] After executing the basic parameter set instruction, dynamic adjustment data is collected in real time (for example, the actual velocity after adjustment rises to 0.58 m / s). In combination with the preset constraint condition (for example, the maximum joint torque limit is 20 N·m, and the maximum steering angular velocity is 10° / s), a quadratic programming (QP) optimization algorithm is used for parameter optimization: a target function min||v_actual-v_target|| is constructed , and the constraint condition is |joint torque|≤18 N·m (a safety margin is reserved). For example, when it is detected that the road adhesion coefficient is low, the theoretically calculated steering angular velocity 15° / s is optimized to 8° / s, and the optimized parameter set {steering angular velocity: 8° / s, acceleration: 0.3 m / s } is output.

[0075] Determine whether the parameter optimization result meets the precision control standard (e.g., the speed is stable within ±0.05 m / s of the target value). If it meets the standard, the intermediate parameter set is determined. For example, the optimized speed is stable at 0.58 m / s, which meets the standard and is used as the intermediate parameter set for subsequent adjustment.

[0076] Control the robot based on the intermediate parameter set, while monitoring the latest motion state (e.g., actual steering angle 4.6° vs. target 5.0°) in real time through a laser tracker or visual sensor. If the matching degree is insufficient (deviation > 0.5°), initiate secondary calibration: use the iterative learning control (ILC) algorithm to predict the compensation amount based on the historical error sequence [ ]=[0.4°, 0.3°, 0.2°], and generate the final control parameter set {steering angle compensation: +0.53°, torque compensation: +7.2 N·m}. The calibration process introduces Kalman filtering to eliminate sensor noise and ensure parameter reliability.

[0077] The final control parameter set is converted into a PWM signal (e.g., duty cycle 72%) by a motion control card and sent to the servo motor. After execution, continuously monitor the state data (500 Hz sampling rate). When the position error is <0.01 meters and the attitude stability time is ≥1 second for 5 consecutive sampling periods, it is determined that the target task requirements are met, and the locked precision attitude control command is output. For example, the end-of-arm trajectory tracking error is stable within ±0.008 meters, and the system records the final command parameters and solidifies them to the control register.

[0078] In summary, this process effectively improves the precision of attitude control commands through multi-stage parameter optimization and calibration, fully combines the robot motion state and task constraints, and ensures that the robot can stably and efficiently complete the target task in a complex dynamic environment, providing a strong guarantee for the safety and reliability of robot operation.

[0079] Referring to Figure 2 , the second embodiment of the present application provides a robot attitude intelligent control system based on a self-organizing network communication architecture, comprising: A data acquisition module for acquiring multi-dimensional environment initial data; A data fusion module for performing multi-source data integration and key feature extraction operations based on the multi-dimensional environment initial data to obtain a fused environment feature set; A panoramic view construction module for performing multi-angle visual data spatial distribution and change trend analysis operations based on the environment feature set to obtain a panoramic information view of the dynamic environment; A risk identification module for dividing the panoramic information view of the dynamic environment into regions to obtain a set of divided view units, and identifying operation risk points based on the set of view units; The preliminary strategy generation module is configured to generate a preliminary attitude adjustment strategy signal if there is a potential operation risk point. The precise instruction determination module is configured to perform an adjustment instruction parameter optimization operation according to the preliminary attitude adjustment strategy signal, and obtain a final precise attitude control instruction.

[0080] It should be noted that the robot attitude intelligent control device based on the ad hoc network communication architecture provided in the embodiments of the present application is used to execute all process steps of the robot attitude intelligent control method based on the ad hoc network communication architecture in the above embodiments, and the working principles and beneficial effects of the two are one-to-one corresponding, thus no longer being repeated.

[0081] To sum up, the present application provides a robot attitude intelligent control method based on an ad hoc network communication architecture, comprising: acquiring multi-dimensional environment initial data; performing multi-source data integration and key feature extraction operations according to the multi-dimensional environment initial data to obtain a fused environment feature set; performing multi-angle visual data space distribution and change trend analysis operations according to the environment feature set to obtain a panoramic information view of a dynamic environment; dividing the panoramic information view of the dynamic environment into regions to obtain a set of divided view units, and performing operation risk point identification according to the set of view units; if there is a potential operation risk point, performing risk response triggering and preliminary attitude adjustment strategy signal generation operations to obtain a preliminary attitude adjustment strategy signal; and performing adjustment instruction parameter optimization operations according to the preliminary attitude adjustment strategy signal to obtain a final precise attitude control instruction. The method realizes intelligent and precise control of the robot attitude in a complex dynamic environment through sensor array collection of multi-dimensional environment data, multi-source integration, panoramic view construction, risk identification and instruction optimization, effectively solves the problem of low precision of attitude adjustment instructions caused by incomplete single sensor information and insufficient environment perception, improves the adaptability of the robot to dynamic obstacles and terrain changes, reduces operation errors, and improves work efficiency and operation safety.

[0082] The embodiments of the present application also provide an electronic device. The electronic device comprises a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a robot attitude intelligent control program based on an ad hoc network communication architecture. The processor implements the steps in each of the above robot attitude intelligent control method embodiments based on an ad hoc network communication architecture when executing the computer program, such as Figure 1 The steps S101 shown. Alternatively, the processor implements the functions of each module / unit in each of the above device embodiments when executing the computer program, such as the data acquisition module.

[0083] For example, the computer program can be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present application. The one or more modules / units can be a series of computer program instruction segments capable of completing a specific function, which are used to describe the execution process of the computer program in the electronic device.

[0084] The electronic device can be a computing device such as a desktop computer, a notebook computer, a palm computer, and a smart tablet. The electronic device can include, but is not limited to, a processor, a memory. Those skilled in the art can understand that the above components are only examples of the electronic device and do not constitute a limitation on the electronic device, and can include more or fewer components than the above, or combine certain components, or different components, for example, the electronic device can also include an input / output device, a network access device, a bus, etc.

[0085] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic components, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The processor is the control center of the electronic device, which connects all parts of the electronic device through various interfaces and lines.

[0086] The memory can be used to store the computer program and / or modules, and the processor realizes various functions of the electronic device by running or executing the computer program and / or modules stored in the memory, and calling data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required by a function (such as a sound playing function, an image playing function, etc.), and the like; and the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), and the like. In addition, the memory can include a high-speed random access memory, and can also include a nonvolatile memory, for example, a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state memory devices.

[0087] The modules / units integrated in the electronic device, if realized in the form of software function units and sold or used as independent products, can be stored in a computer readable storage medium. Based on this understanding, all or part of the processes in the above-mentioned embodiment methods can also be completed by a computer program instructing related hardware, and the computer program can be stored in a computer readable storage medium. The computer program can realize the steps of the above-mentioned various method embodiments when executed by a processor. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or some intermediate forms, etc. The computer readable medium can include any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the contents included in the computer readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction, for example, in some jurisdictions, according to legislation and patent practice, the computer readable medium does not include electrical carrier signals and telecommunication signals.

[0088] It should be noted that the apparatus embodiments described above are merely illustrative, and the units described as separate units can or can not be physically separate, and the units displayed as units can or can not be physical units, i.e. can be located in one place, or can be distributed to multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiment scheme according to actual needs. In addition, the connection relationship between the modules in the apparatus embodiment provided by the present application indicates that there is a communication connection between them, which can be implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement it without creative labor.

[0089] The above specific embodiments further illustrate the purpose, technical scheme and beneficial effects of the present application. It should be understood that the above description is only a specific embodiment of the present application and is not intended to limit the protection scope of the present application. It is particularly pointed out that any modification, equivalent replacement, improvement, etc. made by those skilled in the art within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A robot posture intelligent control method based on an ad hoc network communication architecture, characterized by, The method comprises the following steps: acquiring multi-dimensional environment initial data; performing multi-source data integration and key feature extraction operations according to the multi-dimensional environment initial data to obtain a fused environment feature set; performing multi-angle visual data space distribution and trend analysis operations according to the environment feature set to obtain a panoramic information view of a dynamic environment; dividing the panoramic information view of the dynamic environment into regions to obtain a set of divided view units, and performing operation risk point identification according to the set of view units; if there is a potential operation risk point, performing risk response triggering and preliminary attitude adjustment strategy signal generation operations to obtain a preliminary attitude adjustment strategy signal; performing adjustment instruction parameter optimization operations according to the preliminary attitude adjustment strategy signal to obtain a final precise attitude control instruction. 2.The robot gesture intelligent control method based on ad hoc communication architecture according to claim 1, wherein, The method of acquiring multi-dimensional environment initial data comprises the following steps: acquiring multi-angle data from a dynamic environment through a sensor array and performing preprocessing to obtain an initial multi-dimensional data set; performing denoising and edge detection processing on visual information in the initial multi-dimensional data set to extract clear visual feature data; calculating the spatial distribution in the environment according to distance information and terrain feature information in the initial multi-dimensional data set to determine spatial structure data of the environment; if the matching degree of the spatial structure data and the visual feature data is lower than a preset matching degree threshold, performing calibration processing on the two to obtain a calibrated environment information matrix; performing interpolation filling and standardization processing on the calibrated environment information matrix to obtain a standardized environment perception data set; performing grouping processing according to the environment perception data set and judging the distribution characteristics of each group of data to obtain multi-dimensional environment initial data. 3.The robot pose intelligent control method based on ad hoc communication architecture according to claim 1, wherein, The method of performing multi-source data integration and key feature extraction operations according to the multi-dimensional environment initial data to obtain a fused environment feature set comprises the following steps: performing data cleaning on the multi-dimensional environment initial data, and performing timestamp synchronization processing and missing value filling on the cleaned data to obtain a filled complete data set; performing screening processing on each dimension of data according to the filled complete data set to obtain significant feature data related to environment perception; performing classification processing on the significant feature data to obtain a set of grouped feature data; performing integration processing on data of different groups through the set of grouped feature data, and if the matching degree of the integrated data is lower than a preset threshold, performing consistency adjustment to obtain the fused environment feature set. 4.The robot gesture intelligent control method based on ad hoc communication architecture according to claim 1, wherein, The method of performing multi-angle visual data space distribution and trend analysis operations according to the environment feature set to obtain a panoramic information view of a dynamic environment comprises the following steps: performing spatial coordinate conversion processing on the environment feature set to obtain point data corresponding to spatial distribution, thereby obtaining region feature data before classification; performing classification processing according to different regions according to the region feature data before classification to obtain a set of region features after classification; The classified region feature data is subjected to trend tracking processing to obtain dynamic environment change data over time, and the integrity of the dynamic environment change data is judged; If the integrity of the dynamic environment change data is lower than a preset integrity threshold, the dynamic environment change data is subjected to difference filling to obtain an adjusted dynamic environment data set; Based on the adjusted dynamic environment data set, a comprehensive visual network model is constructed to analyze the spatial distribution and trend of multi-angle visual data, and a panoramic information view of the dynamic environment is obtained.

5. The method of claim 1, wherein, The panoramic information view of the dynamic environment is divided into a set of divided view units, and operation risk point identification is performed based on the set of view units, including: The terrain changes and obstacle regions in the panoramic information view are preliminarily divided to obtain at least one independent terrain unit and obstacle unit, and a set of divided view units is obtained; Based on the set of view units, detailed features of the terrain unit and the obstacle unit are captured, and the captured detailed features are classified in real time to determine the specific category and attribute state of each unit; The specific category and attribute state are matched with a pre-established threat database, and if the attribute state matches the potential threat features in the database, it is marked as a high-risk unit to obtain a set of marked risk units; Based on the set of marked risk units, the operation safety region is dynamically adjusted, and the panoramic information view of the dynamic environment is refreshed in real time in combination with a preset view update mechanism to determine the distribution of potential operation risk points. 6.The robot pose intelligent control method based on ad hoc communication architecture according to claim 1, wherein, If there is a potential operation risk point, a risk response trigger and a preliminary posture adjustment strategy signal generation operation are performed to obtain a preliminary posture adjustment strategy signal, including: If there is a potential operation risk point, the panoramic information view of the dynamic environment is divided into at least one risk region unit to obtain a set of divided risk regions; The spatial position and risk category of the risk region unit in the set of risk regions are captured in detail to determine the specific position coordinates and category label of the unit; If the category label matches the high-risk features in the pre-established threat evaluation database, a corresponding real-time response instruction is generated, and a response signal triggered by the real-time response instruction is obtained; Based on the response signal, the target operation region is dynamically planned, and a preliminary adjustment signal is generated in combination with a preset posture adjustment logic; It is judged whether the preliminary adjustment signal meets the preset safety range, and if so, the preliminary adjustment signal is determined as the preliminary posture adjustment strategy signal.

7. The method of claim 1, wherein the method further comprises: Based on the preliminary posture adjustment strategy signal, an adjustment instruction parameter optimization operation is performed to obtain a final precise posture control instruction, including: Based on the preliminary posture adjustment strategy signal, the current motion state data of the robot is obtained; The current motion state data is compared with a pre-established task matching rule, and if the deviation of the current motion state data from a target task exceeds a preset deviation threshold, a preliminary adjustment instruction is generated, and a basic parameter set of the instruction adjustment is obtained; The robot is dynamically adjusted according to the basic parameter set, dynamic adjustment data of the robot is obtained, and parameter optimization is performed in combination with a preset constraint condition, and a parameter optimization result is obtained; It is judged whether the parameterization result meets a preset precision control standard, and if so, an intermediate parameter set applicable to the current scene is obtained; The robot is dynamically adjusted according to the intermediate parameter set, and the latest motion state information is obtained by monitoring the state of the robot in real time; If the motion state information does not meet a preset target task matching degree, secondary calibration is performed, and a final control parameter set is obtained; According to the final control parameter set, the posture is adjusted, and it is judged whether the adjusted state meets a preset target task requirement, and if so, a final precision posture control instruction is obtained.

8. The method of claim 1, wherein, The sensor array includes a vision sensor, a distance sensor, and a terrain sensor, the vision sensor is used to collect the environmental visual information, the distance sensor is used to collect the environmental distance information, and the terrain sensor is used to collect the environmental terrain feature information.

9. A robot posture intelligent control system based on an ad hoc network communication architecture, characterized by, It includes: A data acquisition module is configured to acquire multi-dimensional environment initial data, a current motion state of a robot, and target task constraints; A data fusion module is configured to perform multi-source data integration and key feature extraction operations based on the multi-dimensional environment initial data to obtain a fused environment feature set; A panoramic view construction module is configured to perform multi-angle visual data spatial distribution and change trend analysis operations based on the environment feature set to obtain a panoramic information view of a dynamic environment; A risk identification module is configured to divide the panoramic information view of the dynamic environment into a set of divided view units, and identify operation risk points based on the set of view units; A preliminary strategy generation module is configured to generate a preliminary posture adjustment strategy signal if there is a potential operation risk point, and perform risk response triggering and preliminary posture adjustment strategy signal generation operations to obtain the preliminary posture adjustment strategy signal; A precision instruction determination module is configured to perform adjustment instruction parameter optimization operations based on the preliminary posture adjustment strategy signal, the current motion state, and the target task constraints to obtain a final precision posture control instruction.

10. An electronic device, comprising: It includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and the processor implements a robot posture intelligent control method based on a self-organizing network communication architecture as claimed in any one of claims 1 to 8 when executing the computer program.

Images