Slam real three-dimensional space scanning imaging system based on quadruped robot

Abstract

The application discloses the technical field of three-dimensional space scanning imaging, and particularly discloses a slam real three-dimensional space scanning imaging system based on a four-legged robot, which comprises a self-adaptive posture compensation module, is used for collecting the posture information of the robot in real time, calculating the angle and position of the scanning device that need to be adjusted according to the posture information, and then adjusting the scanning device in real time, a multi-sensor fusion sensing module, is used for collecting the environment and robot data through multiple sensors, and then calibrating, fusing and complementing the data collected by different sensors through a multi-sensor data fusion algorithm. The self-adaptive posture compensation module can offset the posture deviation caused by the robot movement in real time, so that the scanning device always maintains a stable position and posture, and the accuracy of three-dimensional space scanning is significantly improved.

Description

Technical Field

[0001] This invention relates to the field of three-dimensional spatial scanning imaging technology, specifically to a SLAM-based real-time three-dimensional spatial scanning imaging system based on a quadruped robot. Background Technology

[0002] In today's era of rapid technological advancement, the integration of robotics and 3D spatial perception technologies has become an important research direction. Among these, Simultaneous Localization and Mapping (SLAM) technology, as the core technology for robots to achieve autonomous localization and map creation in unknown environments, has extremely high research value and broad application prospects.

[0003] Quadruped robots, with their unique mechanical structure and locomotion characteristics, possess the ability to move flexibly in complex terrains (such as mountains, ruins, and jungles), reaching areas that are difficult for wheeled and tracked robots to access. Combining SLAM technology with quadruped robots to build a quadruped robot-based SLAM real-time 3D spatial scanning and imaging system can achieve real-time 3D spatial scanning and imaging of unknown environments, providing crucial technical support for numerous fields such as environmental detection, disaster relief, industrial inspection, and archaeological research.

[0004] For example, in earthquake disaster relief, the system can be mounted on a quadruped robot to go deep into the interior of collapsed buildings, scan and build a three-dimensional spatial map in real time, and provide rescuers with accurate information about the internal environment to help them develop more effective rescue plans. In the industrial field, it can be used to perform three-dimensional scanning of the interior of large industrial equipment or complex factory areas to realize equipment status monitoring and digital management of the factory area.

[0005] Although SLAM-based real-world 3D spatial scanning imaging systems using quadruped robots have broad application prospects, there are still many problems in practical applications, mainly in the following aspects:

[0006] 1. Insufficient Scanning Accuracy: During movement, quadruped robots experience bumps and posture changes due to uneven terrain and their own movement characteristics. This causes the position and orientation of the onboard scanning equipment (such as LiDAR and cameras) to shift, affecting the accuracy of 3D spatial scanning. For example, when a quadruped robot walks on a rugged mountain surface, its body will undulate up and down and sway left and right, causing deviations in the angle and position of the laser beam emitted by the LiDAR. Ultimately, the constructed 3D model will appear distorted and inaccurate.

[0007] 2. Poor Environmental Adaptability: In complex environments, such as dimly lit underground caves or dense jungles with numerous obstructions, existing SLAM systems often struggle to accurately extract environmental features, leading to errors in localization and map building. For example, in underground caves, the lack of effective lighting prevents cameras from acquiring clear image information, causing vision-based SLAM algorithms to fail. In dense jungles, the shading of tree branches and leaves results in point cloud data from LiDAR scans containing a large amount of noise and invalid information, affecting the quality of map construction.

[0008] 3. Low Real-Time Data Processing Capacity: 3D spatial scanning generates massive amounts of point cloud and image data. Real-time processing and analysis of this data requires powerful computing capabilities. Currently, quadruped robots, limited by their size and weight, cannot be equipped with high-performance computing devices, resulting in slow data processing speeds and difficulty in meeting the demands of real-time 3D spatial imaging. For example, when scanning a large factory, millions of point cloud data points may be generated per second. Existing quadruped robot processors require several seconds or even tens of seconds to process this data, making it impossible to generate 3D spatial images in real time.

[0009] 4. High system energy consumption: The SLAM scanning and imaging process requires the simultaneous operation of scanning equipment, computing equipment, and robot motion drive equipment, consuming a significant amount of energy. Quadruped robots are typically battery-powered, and the limited battery capacity restricts the system's operating time, limiting its application in scenarios requiring extended operation. For example, when performing 3D scanning of a large mine, existing systems may only be able to operate for 2-3 hours before needing battery replacement, severely impacting work efficiency.

[0010] 5. Dynamic environmental interference leads to decreased map accuracy: In dynamic environments with moving objects (such as robotic arms in factory workshops or personnel activities at disaster relief sites), existing SLAM systems are prone to misclassifying the features of dynamic objects as static environmental features, resulting in ghosting or drifting in the map. For example, when scanning a busy production workshop, moving conveyor belts and robotic arms may be included in the 3D model, causing the subsequently generated map to deviate from the actual static environment.

[0011] Based on the above, a SLAM real-time three-dimensional spatial scanning imaging system based on a quadruped robot is invented. Summary of the Invention

[0012] To address the aforementioned technical problems, according to one aspect of the present invention, the present invention provides the following technical solution:

[0013] A SLAM-based real-world 3D spatial scanning imaging method using a quadruped robot, comprising:

[0014] S1: First, collect the robot's posture information in real time, and calculate the angle and position that the scanning device needs to be adjusted based on the posture information, and then adjust the scanning device in real time.

[0015] S2: First, environmental and robot data are collected through multiple sensors. Then, a multi-sensor data fusion algorithm is used to calibrate, fuse, and complement the data collected by different sensors.

[0016] S3: Based on the dynamic feature recognition model of deep learning, the fused point cloud data and image sequence are analyzed to identify moving dynamic targets and remove their feature points in the point cloud; at the same time, the static environment features are optimized through the temporal consistency verification algorithm to fill the map blank areas caused by dynamic object occlusion and generate a pure static map without dynamic interference.

[0017] S4: Construct a real-time environmental map based on the optimized environmental data, and use the fast exploration random tree algorithm combined with the target area of ​​the scanning task to plan the optimal movement path. At the same time, adjust the path in real time by continuously monitoring dynamic or static obstacles in the environment to avoid collisions between the robot and obstacles.

[0018] S5: Performs real-time preprocessing on scanned data to reduce the amount of data transmitted and the complexity of subsequent processing. At the same time, it transmits the preprocessed data to the cloud server via wireless network, enabling the cloud server to utilize its powerful computing resources for high-precision 3D map construction, data analysis, and model optimization.

[0019] S6: Real-time monitoring of battery information enables dynamic adjustment of power consumption of each device based on system operating status and task requirements. Simultaneously, an energy recovery device is installed to convert some kinetic energy into electrical energy and store it in the battery when the robot is going downhill or decelerating, thereby extending the system's operating time.

[0020] As a preferred embodiment of the SLAM real-world three-dimensional spatial scanning imaging method based on a quadruped robot described in this invention, step S3 includes the following steps:

[0021] S31: Receive the fused point cloud data and image sequence data output by S2, and simultaneously acquire the robot's motion state information;

[0022] S32: The improved YOLOv8 algorithm is used to perform target detection on the image sequence, identify possible dynamic target candidate regions, and combine optical flow method to calculate the motion vector of pixels in the image, mark the regions with obvious motion trajectories, fuse the two results, and preliminarily determine the dynamic feature regions.

[0023] S33: Based on the calibration relationship between the image and the point cloud, the initially identified dynamic feature regions are mapped to the point cloud data, the corresponding point cloud feature points are determined, and the motion consistency analysis of the point cloud feature points is performed. If the motion pattern conforms to the characteristics of the dynamic target, it is removed from the point cloud data.

[0024] S34: Analyze the point cloud data after removing dynamic features according to the time series, calculate the consistency of static feature points at the same spatial location at different times, and fill in the missing static feature regions at a certain time due to occlusion by dynamic objects by interpolation based on the distribution of static features before and after the time series.

[0025] S35: Integrate the verified static feature point cloud data, construct a clean static 3D map without dynamic interference, and output it to S4.

[0026] As a preferred embodiment of the SLAM real-world three-dimensional spatial scanning imaging method based on a quadruped robot described in this invention, step S4 includes the following steps:

[0027] S41: Receive the clean static map output by S3 and combine it with the real-time sensor data to construct a real-time environmental map containing the current environmental state.

[0028] S42: Based on the target area of ​​the scanning task and the robot's current position, use the fast exploration random tree algorithm to search for paths in the real-time environment map and plan an initial optimal path from the current position to the target area.

[0029] S43: By continuously receiving sensor data, monitor whether new dynamic or static obstacles appear in the environment. For dynamic obstacles, track their movement trajectory and predict their position changes in the future.

[0030] S44: When an obstacle is detected that may conflict with the current planned path, a new path to avoid the obstacle is searched again in the real-time environment map based on the dynamic replanning mechanism of the fast exploration random tree algorithm.

[0031] S45: Send the adjusted path instructions to the quadruped robot's motion control system to control the robot to move along the planned path. At the same time, collect the robot's actual motion trajectory in real time and compare it with the planned path. If there is a deviation, correct it in time.

[0032] As a preferred embodiment of the SLAM real-world three-dimensional spatial scanning imaging method based on a quadruped robot described in this invention, step S5 includes the following steps:

[0033] S51: Receives raw data collected by each sensor;

[0034] S52: A point cloud feature-based compression algorithm is used to compress redundant point cloud data, and an adaptive resolution compression method is used for image data to reduce the amount of data.

[0035] S53: Use algorithms to remove noise points from point cloud data, smooth image data, and eliminate image noise;

[0036] S54: Extract key feature points from preprocessed point cloud data and extract feature descriptors from image data;

[0037] S55: Stores the preprocessed data locally temporarily and transmits it to the cloud server via wireless network;

[0038] S56: After receiving preprocessed data transmitted from the edge terminal, the cloud server performs time synchronization and spatial registration on data collected at different times and locations, and integrates them into a unified dataset.

[0039] S57: Utilizing the powerful computing resources of the cloud, it uses a globally optimized SLAM algorithm to process the integrated dataset and construct a high-precision global 3D map.

[0040] S58: Perform accuracy analysis and error assessment on the constructed 3D map to optimize SLAM algorithm parameters and sensor calibration parameters based on the analysis results. At the same time, train dynamic feature recognition model and path planning model based on historical data to improve model performance.

[0041] S59: Provide feedback on the optimized algorithm parameters, model, and high-precision 3D map to guide subsequent scanning imaging and path planning.

[0042] As a preferred embodiment of the SLAM real-world three-dimensional spatial scanning imaging method based on a quadruped robot described in this invention, step S6 includes the following steps:

[0043] S61: Real-time acquisition of battery power, voltage, current, and power consumption data of each device;

[0044] S62: Analyze the power consumption data of each device and establish a correlation model between device power consumption and working status to predict the energy consumption required to complete the remaining tasks based on the current scanning task progress and environmental conditions.

[0045] S63: Develop a power consumption adjustment strategy based on the energy consumption prediction results and the current battery level;

[0046] S64: When the robot is going downhill or decelerating, the energy recovery device can be activated, the drive motor can be switched to generator mode, the kinetic energy can be converted into electrical energy, and the recovered electrical energy can be stored in the battery through the charging management circuit.

[0047] S65: Real-time feedback of battery status, energy consumption, and adjustment strategies to the robot control system. At the same time, when the battery level is lower than the preset threshold, a low battery alarm is issued to prompt the operator to replace the battery or pause the task in time.

[0048] A quadruped robot-based SLAM real-world 3D spatial scanning imaging system includes:

[0049] The adaptive posture compensation module is used to first collect the robot's posture information in real time, calculate the angle and position that the scanning device needs to be adjusted based on the posture information, and then adjust the scanning device in real time.

[0050] The multi-sensor fusion perception module is used to first collect environmental and robot data through multiple sensors, and then to calibrate, fuse and complement the data collected by different sensors through a multi-sensor data fusion algorithm.

[0051] The dynamic feature removal and static map optimization module is used to analyze the fused point cloud data and image sequence based on the deep learning dynamic feature recognition model, identify moving dynamic targets, and remove their feature points in the point cloud; at the same time, through the temporal consistency verification algorithm, the static environment features are optimized to fill the map blank areas caused by dynamic object occlusion and generate a clean static map without dynamic interference.

[0052] The dynamic path planning and obstacle avoidance module is used to build a real-time environment map based on the optimized environment data, and to plan the optimal movement path by using a fast exploration random tree algorithm combined with the target area of ​​the scanning task. At the same time, by continuously monitoring dynamic or static obstacles in the environment, the path is adjusted in real time to avoid collisions between the robot and obstacles.

[0053] The edge computing and cloud collaborative processing module is used to preprocess the scanned data in real time, reducing the amount of data transmission and the complexity of subsequent processing. At the same time, the preprocessed data is transmitted to the cloud server via wireless network, so that the cloud server can use its powerful computing resources to perform high-precision 3D map construction, data analysis and model optimization.

[0054] The intelligent energy management module is used to monitor battery information in real time, so that it can dynamically adjust the power consumption of each device according to the system's working status and task requirements. At the same time, an energy recovery device is set up to convert some of the kinetic energy into electrical energy and store it in the battery when the robot goes downhill or decelerates, thus extending the system's working time.

[0055] As a preferred embodiment of the SLAM real-world 3D spatial scanning imaging system based on a quadruped robot described in this invention, the dynamic feature removal and static map optimization module includes:

[0056] The data input module is used to receive fused point cloud data and image sequence data output by the multi-sensor fusion perception module, and at the same time acquire the robot's motion state information;

[0057] The dynamic feature preliminary identification module is used to perform target detection on image sequences using the improved YOLOv8 algorithm, identify possible dynamic target candidate regions, and combine optical flow method to calculate the motion vector of pixels in the image, mark the regions with obvious motion trajectories, and fuse the two results to preliminarily determine the dynamic feature regions.

[0058] The point cloud dynamic feature matching and removal module is used to map the initially identified dynamic feature regions to the point cloud data according to the calibration relationship between the image and the point cloud, determine the corresponding point cloud feature points, and perform motion consistency analysis on the point cloud feature points. If their motion pattern conforms to the characteristics of the dynamic target, they are removed from the point cloud data.

[0059] The static feature temporal consistency verification module is used to analyze the point cloud data after removing dynamic features according to the time series, calculate the consistency of static feature points at the same spatial location at different times, and fill in the missing static feature regions at a certain time due to dynamic object occlusion by interpolation based on the static feature distribution before and after the time.

[0060] The clean static map generation module is used to integrate the verified static feature point cloud data, construct a clean static 3D map without dynamic interference, and output it to the dynamic path planning and obstacle avoidance module.

[0061] As a preferred embodiment of the SLAM real-world three-dimensional spatial scanning imaging system based on a quadruped robot described in this invention, the dynamic path planning and obstacle avoidance module includes:

[0062] The environment map loading module is used to receive the clean static map output by the dynamic feature removal and static map optimization module, and combine it with real-time collected sensor data to construct a real-time environment map containing the current environmental status.

[0063] The initial path planning module is used to search for a path in the real-time environment map based on the target area of ​​the scanning task and the robot's current position, and to plan an initial optimal path from the current position to the target area.

[0064] The obstacle real-time monitoring module is used to continuously receive sensor data to monitor whether new dynamic or static obstacles appear in the environment. For dynamic obstacles, it tracks their movement trajectory and predicts their position changes in the future.

[0065] The path dynamic adjustment module is used to search for a new path to avoid the obstacle in the real-time environment map when an obstacle is detected that may conflict with the current planned path. This is based on the dynamic replanning mechanism of the fast exploration random tree algorithm.

[0066] The path execution and feedback module is used to send the adjusted path instructions to the quadruped robot's motion control system to control the robot to move along the planned path. At the same time, it collects the robot's actual motion trajectory in real time, compares it with the planned path, and corrects it in time if there is a deviation.

[0067] As a preferred embodiment of the SLAM real-world 3D spatial scanning imaging system based on a quadruped robot described in this invention, the edge computing and cloud collaborative processing module includes:

[0068] The data acquisition module is used to receive raw data collected by each sensor;

[0069] The data compression module is used to compress redundant point cloud data using a compression algorithm based on point cloud features, and to reduce the amount of data by using an adaptive resolution compression method for image data.

[0070] The noise filtering module is used to remove noise points from point cloud data using algorithms, smooth image data, and eliminate image noise.

[0071] The feature extraction module is used to extract key feature points from preprocessed point cloud data and feature descriptors from image data.

[0072] The local storage and transmission module is used to temporarily store the preprocessed data locally and transmit it to the cloud server via a wireless network;

[0073] The data receiving and integration module is used to receive preprocessed data transmitted from the edge terminal on the cloud server, and then perform time synchronization and spatial registration on data collected at different times and locations, integrating them into a unified dataset.

[0074] The high-precision 3D map building module is used to utilize the powerful computing resources in the cloud and employ a globally optimized SLAM algorithm to process the integrated dataset and build a high-precision global 3D map.

[0075] The data analysis and model optimization module is used to perform accuracy analysis and error assessment on the constructed 3D map, so as to optimize the SLAM algorithm parameters and sensor calibration parameters based on the analysis results. At the same time, it trains dynamic feature recognition model and path planning model based on historical data to improve model performance.

[0076] The results feedback module is used to provide feedback on the optimized algorithm parameters, model, and high-precision 3D map to guide subsequent scanning imaging and path planning.

[0077] As a preferred embodiment of the SLAM real-world three-dimensional spatial scanning imaging system based on a quadruped robot described in this invention, the intelligent energy management module includes:

[0078] The energy status monitoring module is used to collect real-time data on battery charge, voltage, current, and power consumption of various devices.

[0079] The power consumption analysis and prediction module is used to analyze the power consumption data of each device, establish a correlation model between device power consumption and working status, and predict the energy consumption required to complete the remaining tasks based on the current scanning task progress and environmental conditions.

[0080] The dynamic power consumption adjustment module is used to formulate power consumption adjustment strategies based on energy consumption prediction results and the current battery level.

[0081] The energy recovery control module is used to activate the energy recovery device when the robot is going downhill or decelerating, switch the drive motor to generator mode, convert kinetic energy into electrical energy, and store the recovered electrical energy in the battery through the charging management circuit.

[0082] The status feedback and alarm module is used to provide real-time feedback on battery status, energy consumption, and adjustment strategies to the robot control system. At the same time, when the battery level is lower than a preset threshold, a low battery alarm is issued to prompt the operator to replace the battery or pause the task in time.

[0083] Compared with existing technologies:

[0084] 1. Addressing the issue of insufficient scanning accuracy: The adaptive attitude compensation module can compensate for attitude deviations caused by robot movement in real time, ensuring that the scanning device maintains a stable position and attitude, significantly improving the accuracy of 3D spatial scanning. For example, when navigating rugged mountainous terrain, this module can promptly adjust the angle and position of the LiDAR to avoid laser beam deviation, effectively reducing distortion and warping in the constructed 3D model.

[0085] 2. Addressing the issue of poor environmental adaptability: The multi-sensor fusion sensing module enhances the system's adaptability to complex environments by leveraging the complementary strengths of different sensors. In dimly lit underground caves, infrared sensors can replace cameras; in dense jungles, multi-sensor data fusion filters out noise and invalid information, ensuring accurate environmental feature extraction and reducing positioning and map building errors.

[0086] 3. Addressing the issue of decreased map accuracy due to dynamic environmental interference: The dynamic feature removal and static map optimization modules accurately identify and remove dynamic object features using deep learning algorithms, preventing their interference with the static map. Simultaneously, temporal optimization fills in map gaps, significantly improving the accuracy and reliability of 3D maps in dynamic scenes. For example, when scanning a factory workshop, the influence of moving robotic arms can be effectively eliminated, generating a 3D model that perfectly matches the static structure of the workshop.

[0087] 4. For the dynamic path planning and obstacle avoidance module: This module ensures that the robot autonomously selects the optimal path in complex environments, avoiding interruptions to the scanning task due to collisions with obstacles, thus improving the system's autonomy and safety. For example, in a rubble rescue scenario, it can automatically avoid collapsed walls and steel bars, continuously scanning passable areas.

[0088] 5. Addressing the issue of low real-time data processing: In the edge computing and cloud collaborative processing module, edge computing performs real-time preprocessing of data, reducing data transmission and processing pressure. The powerful computing capabilities of the cloud server ensure high-precision processing. The combination of the two significantly improves data processing speed, meeting the needs of real-time 3D spatial imaging. For example, when scanning large factory buildings, it can quickly process millions of point cloud data points to generate 3D spatial images in real time.

[0089] 6. Addressing the issue of high system energy consumption: The intelligent energy management module effectively reduces system energy consumption and extends working time by dynamically adjusting equipment power consumption and energy recovery. In long-term operation scenarios such as large-scale mine scanning, it can reduce the number of battery replacements, improve work efficiency, and avoid affecting the work process due to frequent battery swapping. Attached Figure Description

[0090] Figure 1 This is a schematic diagram of the overall framework of the present invention;

[0091] Figure 2 This is a schematic diagram of the framework of the dynamic feature removal and static map optimization module of the present invention;

[0092] Figure 3 This is a schematic diagram of the framework of the dynamic path planning and obstacle avoidance module of the present invention;

[0093] Figure 4 This is a schematic diagram of the edge computing and cloud collaborative processing module framework of the present invention;

[0094] Figure 5 This is a schematic diagram of the intelligent energy management module framework of the present invention. Detailed Implementation

[0095] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0096] This invention provides a SLAM-based real-world 3D spatial scanning imaging system using a quadruped robot. Please refer to [link / reference]. Figures 1-5 The method includes:

[0097] S1: First, collect the robot's posture information (such as tilt angle, swing amplitude, etc.) in real time, and calculate the angle and position that the scanning device needs to be adjusted based on the posture information. Then, adjust the scanning device in real time to counteract the posture deviation caused by the robot's movement and improve the scanning accuracy.

[0098] S2: First, environmental and robot data are collected through multiple sensors; LiDAR is responsible for acquiring 3D point cloud data of the environment, which is suitable for various lighting conditions; high-definition cameras are used to collect texture and color information of the environment to improve the realism of the 3D model; infrared sensors can effectively detect objects in dimly lit or smoky environments; IMU is used to monitor the robot's motion state and position changes in real time; then, through multi-sensor data fusion algorithms, the data collected by different sensors are calibrated, fused, and complemented to improve the system's environmental perception capability and feature extraction accuracy in complex environments;

[0099] S3: Based on the dynamic feature recognition model of deep learning (such as the improved YOLOv8 combined with optical flow method), the fused point cloud data and image sequence are analyzed to identify moving dynamic targets (such as pedestrians, robotic arms, rolling stones, etc.) and remove their feature points in the point cloud; at the same time, the static environment features are optimized through the temporal consistency verification algorithm to fill the map blank areas caused by dynamic object occlusion and generate a pure static map without dynamic interference.

[0100] S4: Construct a real-time environmental map based on the optimized environmental data, and use the fast exploration random tree algorithm combined with the target area of ​​the scanning task to plan the optimal movement path. At the same time, by continuously monitoring dynamic or static obstacles in the environment, such as suddenly falling stones or temporary obstacles, adjust the path in real time to avoid collisions between the robot and obstacles.

[0101] S5: Performs real-time preprocessing on the scanned data (such as data compression, noise filtering, feature extraction, etc.) to reduce the amount of data transmitted and the complexity of subsequent processing. At the same time, the preprocessed data is transmitted to the cloud server via wireless network so that the cloud server can use its powerful computing resources to perform high-precision 3D map construction, data analysis and model optimization.

[0102] S6: Real-time monitoring of battery power, voltage, and current information enables dynamic adjustment of power consumption of each device based on system operating status (such as scanning mode, movement speed, etc.) and task requirements. Simultaneously, an energy recovery device is set up to convert some kinetic energy into electrical energy and store it in the battery when the robot is going downhill or decelerating, thereby extending the system's working time.

[0103] S3 includes the following steps:

[0104] S31: Receive the fused point cloud data and image sequence data output by S2, and simultaneously acquire the robot's motion state information;

[0105] S32: The improved YOLOv8 algorithm is used to perform target detection on the image sequence, identify possible dynamic target candidate regions, and combine optical flow method to calculate the motion vector of pixels in the image, mark the regions with obvious motion trajectories, fuse the two results, and preliminarily determine the dynamic feature regions.

[0106] S33: Based on the calibration relationship between the image and the point cloud, the initially identified dynamic feature regions are mapped to the point cloud data, the corresponding point cloud feature points are determined, and the motion consistency analysis of the point cloud feature points is performed. If the motion pattern conforms to the characteristics of the dynamic target, it is removed from the point cloud data.

[0107] S34: Analyze the point cloud data after removing dynamic features according to the time series, calculate the consistency of static feature points at the same spatial location at different times, and fill in the missing static feature regions at a certain time due to occlusion by dynamic objects by interpolation based on the distribution of static features before and after the time series.

[0108] S35: Integrate the verified static feature point cloud data, construct a clean static 3D map without dynamic interference, and output it to S4.

[0109] S4 includes the following steps:

[0110] S41: Receives the clean static map output by S3 and combines it with real-time collected sensor data (such as real-time point cloud of LiDAR) to construct a real-time environmental map containing the current environmental state.

[0111] S42: Based on the target area of ​​the scanning task and the robot's current position, use the fast exploration random tree algorithm to search for paths in the real-time environment map and plan an initial optimal path from the current position to the target area. This path must meet constraints such as the shortest distance and the highest safety.

[0112] S43: By continuously receiving sensor data, monitor whether new dynamic or static obstacles appear in the environment. For dynamic obstacles, track their movement trajectory and predict their position changes in the future.

[0113] S44: When an obstacle is detected that may conflict with the current planned path, a new path to avoid the obstacle is searched again in the real-time environment map based on the dynamic replanning mechanism of the fast exploration random tree algorithm. The new path needs to be smoothly transitioned with the original path to avoid drastic changes in the robot's motion state.

[0114] S45: Send the adjusted path instructions to the quadruped robot's motion control system to control the robot to move along the planned path. At the same time, collect the robot's actual motion trajectory in real time and compare it with the planned path. If there is a deviation, correct it in time.

[0115] S5 includes the following steps:

[0116] S51: Receives raw data (point cloud, image, attitude, etc.) collected by various sensors;

[0117] S52: Use point cloud feature-based compression algorithms (such as voxel grid downsampling) to compress redundant point cloud data, and use an adaptive resolution compression method for image data to reduce the amount of data;

[0118] S53: Use algorithms such as statistical filtering and radius filtering to remove noise points in point cloud data, smooth image data, and eliminate image noise;

[0119] S54: Extract key feature points (such as planes, edges, etc.) from preprocessed point cloud data, and extract feature descriptors (such as SIFT, ORB, etc.) from image data;

[0120] S55: Stores the pre-processed data locally temporarily and transmits it to the cloud server via wireless networks such as 5G / Wi-Fi;

[0121] S56: After receiving preprocessed data transmitted from the edge terminal, the cloud server performs time synchronization and spatial registration on data collected at different times and locations, and integrates them into a unified dataset.

[0122] S57: Utilizing the powerful computing resources of the cloud, a global optimization-based SLAM algorithm (such as BundleAdjustment) is used to process the integrated dataset and construct a high-precision global 3D map;

[0123] S58: Perform accuracy analysis and error assessment on the constructed 3D map to optimize SLAM algorithm parameters and sensor calibration parameters based on the analysis results. At the same time, train dynamic feature recognition model and path planning model based on historical data to improve model performance.

[0124] S59: Provide feedback on the optimized algorithm parameters, model, and high-precision 3D map to guide subsequent scanning imaging and path planning.

[0125] S6 includes the following steps:

[0126] S61: Real-time acquisition of battery power, voltage, current, and power consumption data of each device;

[0127] S62: Analyze the power consumption data of each device and establish a correlation model between device power consumption and working status (such as scanning frequency, movement speed, etc.) to predict the energy consumption required to complete the remaining tasks based on the current scanning task progress and environmental conditions.

[0128] S63: Develop a power consumption adjustment strategy based on the energy consumption prediction results and the current battery level;

[0129] When the battery is fully charged, control each device to operate in normal working mode to ensure scanning accuracy and efficiency;

[0130] When the battery is low, reduce the power consumption of non-critical equipment (such as reducing the camera frame rate and reducing the LiDAR scanning density), and reduce the robot's movement speed on flat terrain to reduce the power consumption of the drive motor.

[0131] S64: When the robot is going downhill or decelerating, the energy recovery device can be activated, the drive motor can be switched to generator mode, the kinetic energy can be converted into electrical energy, and the recovered electrical energy can be stored in the battery through the charging management circuit.

[0132] S65: Real-time feedback of battery status, energy consumption, and adjustment strategies to the robot control system. At the same time, when the battery level is lower than the preset threshold, a low battery alarm is issued to prompt the operator to replace the battery or pause the task in time.

[0133] A quadruped robot-based SLAM real-world 3D spatial scanning imaging system includes:

[0134] The adaptive posture compensation module is used to first collect the robot's posture information (such as tilt angle, swing amplitude, etc.) in real time, and calculate the angle and position that the scanning device needs to be adjusted based on the posture information. Then, the scanning device is adjusted in real time to counteract the posture deviation caused by the robot's movement and improve the scanning accuracy.

[0135] The multi-sensor fusion perception module is used to first collect environmental and robot data through multiple sensors; LiDAR is responsible for acquiring 3D point cloud data of the environment, which is suitable for various lighting conditions; high-definition cameras are used to collect texture and color information of the environment to improve the realism of the 3D model; infrared sensors can effectively detect objects in dimly lit or smoky environments; IMU is used to monitor the robot's motion state and position changes in real time; then, through multi-sensor data fusion algorithms, the data collected by different sensors are calibrated, fused, and complemented to improve the system's environmental perception capability and feature extraction accuracy in complex environments;

[0136] The dynamic feature removal and static map optimization module is used to analyze the fused point cloud data and image sequences based on the dynamic feature recognition model of deep learning (such as the improved YOLOv8 combined with optical flow method), identify moving dynamic targets (such as pedestrians, robotic arms, rolling stones, etc.), and remove their feature points in the point cloud; at the same time, through the temporal consistency verification algorithm, the static environment features are optimized to fill the map blank areas caused by dynamic object occlusion and generate a clean static map without dynamic interference.

[0137] The dynamic path planning and obstacle avoidance module is used to build a real-time environment map based on the optimized environment data, and to plan the optimal movement path by using a fast exploration random tree algorithm combined with the target area of ​​the scanning task. At the same time, by continuously monitoring dynamic or static obstacles in the environment, such as suddenly falling stones or temporary obstacles, the path is adjusted in real time to avoid collisions between the robot and obstacles.

[0138] The edge computing and cloud collaborative processing module is used to perform real-time preprocessing of scanned data (such as data compression, noise filtering, feature extraction, etc.), reducing the amount of data transmission and the complexity of subsequent processing. At the same time, the preprocessed data is transmitted to the cloud server via wireless network, so that the cloud server can use its powerful computing resources to perform high-precision 3D map construction, data analysis and model optimization.

[0139] The intelligent energy management module is used to monitor battery power, voltage, and current in real time. It can dynamically adjust the power consumption of each device according to the system's working status (such as scanning mode, movement speed, etc.) and task requirements. At the same time, an energy recovery device is set up to convert some kinetic energy into electrical energy and store it in the battery when the robot goes downhill or decelerates, thus extending the system's working time.

[0140] The dynamic feature removal and static map optimization module includes:

[0141] The data input module is used to receive fused point cloud data and image sequence data output by the multi-sensor fusion perception module, and at the same time acquire the robot's motion state information;

[0142] The dynamic feature preliminary identification module is used to perform target detection on image sequences using the improved YOLOv8 algorithm, identify possible dynamic target candidate regions, and combine optical flow method to calculate the motion vector of pixels in the image, mark the regions with obvious motion trajectories, and fuse the two results to preliminarily determine the dynamic feature regions.

[0143] The point cloud dynamic feature matching and removal module is used to map the initially identified dynamic feature regions to the point cloud data according to the calibration relationship between the image and the point cloud, determine the corresponding point cloud feature points, and perform motion consistency analysis on the point cloud feature points. If their motion pattern conforms to the characteristics of the dynamic target, they are removed from the point cloud data.

[0144] The static feature temporal consistency verification module is used to analyze the point cloud data after removing dynamic features according to the time series, calculate the consistency of static feature points at the same spatial location at different times, and fill in the missing static feature regions at a certain time due to dynamic object occlusion by interpolation based on the static feature distribution before and after the time.

[0145] The clean static map generation module is used to integrate the verified static feature point cloud data, construct a clean static 3D map without dynamic interference, and output it to the dynamic path planning and obstacle avoidance module.

[0146] The dynamic path planning and obstacle avoidance module includes:

[0147] The environment map loading module is used to receive the clean static map output by the dynamic feature removal and static map optimization module, and combine it with real-time collected sensor data (such as real-time point cloud of LiDAR) to construct a real-time environment map containing the current environmental state.

[0148] The initial path planning module is used to search for a path in the real-time environment map based on the target area of ​​the scanning task and the robot's current position using a fast exploration random tree algorithm, and to plan an initial optimal path from the current position to the target area. This path must meet constraints such as the shortest distance and the highest safety.

[0149] The obstacle real-time monitoring module is used to continuously receive sensor data to monitor whether new dynamic or static obstacles appear in the environment. For dynamic obstacles, it tracks their movement trajectory and predicts their position changes in the future.

[0150] The path dynamic adjustment module is used to search for a new path to avoid the obstacle in the real-time environment map when an obstacle is detected that may conflict with the current planned path. This is based on the dynamic replanning mechanism of the fast exploration random tree algorithm. The new path needs to be smoothly transitioned from the original path to avoid drastic changes in the robot's motion state.

[0151] The path execution and feedback module is used to send the adjusted path instructions to the quadruped robot's motion control system to control the robot to move along the planned path. At the same time, it collects the robot's actual motion trajectory in real time, compares it with the planned path, and corrects it in time if there is a deviation.

[0152] The edge computing and cloud collaborative processing module includes:

[0153] The data acquisition module is used to receive raw data (point cloud, image, attitude, etc.) collected by various sensors;

[0154] The data compression module is used to compress redundant point cloud data using point cloud feature-based compression algorithms (such as voxel grid downsampling) and to reduce the amount of data by using an adaptive resolution compression method for image data.

[0155] The noise filtering module is used to remove noise points in point cloud data using algorithms such as statistical filtering and radius filtering, and to smooth image data to eliminate image noise.

[0156] The feature extraction module is used to extract key feature points (such as planes, edges, etc.) from preprocessed point cloud data and feature descriptors (such as SIFT, ORB, etc.) from image data.

[0157] The local storage and transmission module is used to temporarily store the preprocessed data locally and transmit it to the cloud server via wireless networks such as 5G / Wi-Fi.

[0158] The data receiving and integration module is used to receive preprocessed data transmitted from the edge terminal on the cloud server, and then perform time synchronization and spatial registration on data collected at different times and locations, integrating them into a unified dataset.

[0159] The high-precision 3D map building module is used to utilize the powerful computing resources in the cloud and employ a globally optimized SLAM algorithm (such as BundleAdjustment) to process the integrated dataset and build a high-precision global 3D map.

[0160] The data analysis and model optimization module is used to perform accuracy analysis and error assessment on the constructed 3D map, so as to optimize the SLAM algorithm parameters and sensor calibration parameters based on the analysis results. At the same time, it trains dynamic feature recognition model and path planning model based on historical data to improve model performance.

[0161] The results feedback module is used to provide feedback on the optimized algorithm parameters, model, and high-precision 3D map to guide subsequent scanning imaging and path planning.

[0162] The intelligent energy management module includes:

[0163] The energy status monitoring module is used to collect real-time data on battery charge, voltage, current, and power consumption of various devices.

[0164] The power consumption analysis and prediction module is used to analyze the power consumption data of each device and establish a correlation model between device power consumption and working status (such as scanning frequency, movement speed, etc.) to predict the energy consumption required to complete the remaining tasks based on the current scanning task progress and environmental conditions.

[0165] The dynamic power consumption adjustment module is used to formulate power consumption adjustment strategies based on energy consumption prediction results and the current battery level.

[0166] When the battery is fully charged, control each device to operate in normal working mode to ensure scanning accuracy and efficiency;

[0167] When the battery is low, reduce the power consumption of non-critical equipment (such as reducing the camera frame rate and reducing the LiDAR scanning density), and reduce the robot's movement speed on flat terrain to reduce the power consumption of the drive motor.

[0168] The energy recovery control module is used to activate the energy recovery device when the robot is going downhill or decelerating, switch the drive motor to generator mode, convert kinetic energy into electrical energy, and store the recovered electrical energy in the battery through the charging management circuit.

[0169] The status feedback and alarm module is used to provide real-time feedback on battery status, energy consumption, and adjustment strategies to the robot control system. At the same time, when the battery level is lower than a preset threshold, a low battery alarm is issued to prompt the operator to replace the battery or pause the task in time.

[0170] Although the present invention has been described above with reference to embodiments, various modifications can be made and components can be replaced with equivalents without departing from the scope of the invention. In particular, as long as there is no structural conflict, the features in the disclosed embodiments can be combined with each other in any manner. The lack of an exhaustive description of these combinations in this specification is merely for the sake of brevity and resource conservation. Therefore, the present invention is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A SLAM-based real-world 3D spatial scanning imaging method using a quadruped robot, characterized in that, The method includes: S1: First, the robot's posture information is collected in real time, and the angle and position that the scanning device needs to be adjusted are calculated based on the posture information, and then the scanning device is adjusted in real time; S2: First, environmental and robot data are collected through multiple sensors. Then, a multi-sensor data fusion algorithm is used to calibrate, fuse, and complement the data collected by different sensors. S3: Based on the dynamic feature recognition model of deep learning, the fused point cloud data and image sequence are analyzed to identify moving dynamic targets and remove their feature points in the point cloud; at the same time, the static environment features are optimized through the temporal consistency verification algorithm to fill the map blank areas caused by dynamic object occlusion and generate a pure static map without dynamic interference. S4: Construct a real-time environmental map based on the optimized environmental data, and use the fast exploration random tree algorithm combined with the target area of ​​the scanning task to plan the optimal movement path. At the same time, adjust the path in real time by continuously monitoring dynamic or static obstacles in the environment to avoid collisions between the robot and obstacles. S5: Performs real-time preprocessing on scanned data to reduce the amount of data transmitted and the complexity of subsequent processing. At the same time, it transmits the preprocessed data to the cloud server via wireless network, enabling the cloud server to utilize its powerful computing resources for high-precision 3D map construction, data analysis, and model optimization. S6: Real-time monitoring of battery information enables dynamic adjustment of power consumption of each device based on system working status and task requirements. At the same time, an energy recovery device is set up to convert some kinetic energy into electrical energy and store it in the battery when the robot goes downhill or decelerates, thus extending the system's working time. The S3 includes the following steps: S31: Receive the fused point cloud data and image sequence data output by S2, and simultaneously acquire the robot's motion state information; S32: The improved YOLOv8 algorithm is used to perform target detection on the image sequence, identify possible dynamic target candidate regions, and combine optical flow method to calculate the motion vector of pixels in the image, mark the regions with obvious motion trajectories, fuse the two results, and preliminarily determine the dynamic feature regions. S33: Based on the calibration relationship between the image and the point cloud, the initially identified dynamic feature regions are mapped to the point cloud data, the corresponding point cloud feature points are determined, and the motion consistency analysis of the point cloud feature points is performed. If the motion pattern conforms to the characteristics of the dynamic target, it is removed from the point cloud data. S34: Analyze the point cloud data after removing dynamic features according to the time series, calculate the consistency of static feature points at the same spatial location at different times, and fill in the missing static feature regions at a certain time due to occlusion by dynamic objects by interpolation based on the distribution of static features before and after the time series. S35: Integrate the verified static feature point cloud data, construct a clean static 3D map without dynamic interference, and output it to S4.

2. The SLAM real-world three-dimensional spatial scanning imaging method based on a quadruped robot according to claim 1, characterized in that, The S4 includes the following steps: S41: Receive the clean static map output by S3, and combine it with the real-time collected sensor data to construct a real-time environmental map containing the current environmental state. S42: Based on the target area of ​​the scanning task and the robot's current position, use the fast exploration random tree algorithm to search for paths in the real-time environment map and plan an initial optimal path from the current position to the target area. S43: By continuously receiving sensor data, monitor whether new dynamic or static obstacles appear in the environment. For dynamic obstacles, track their movement trajectory and predict their position changes in the future. S44: When an obstacle is detected that may conflict with the current planned path, a new path to avoid the obstacle is searched again in the real-time environment map based on the dynamic replanning mechanism of the fast exploration random tree algorithm. S45: Send the adjusted path instructions to the quadruped robot's motion control system to control the robot to move along the planned path. At the same time, collect the robot's actual motion trajectory in real time and compare it with the planned path. If there is a deviation, correct it in time.

3. The SLAM real-world three-dimensional spatial scanning imaging method based on a quadruped robot according to claim 1, characterized in that, S5 includes the following steps: S51: Receive the raw data collected by each sensor; S52: A point cloud feature-based compression algorithm is used to compress redundant point cloud data, and an adaptive resolution compression method is used for image data to reduce the amount of data. S53: Use algorithms to remove noise points from point cloud data, smooth image data, and eliminate image noise; S54: Extract key feature points from preprocessed point cloud data and extract feature descriptors from image data; S55: Stores the preprocessed data locally temporarily and transmits it to the cloud server via wireless network; S56: After receiving preprocessed data transmitted from the edge terminal, the cloud server performs time synchronization and spatial registration on data collected at different times and locations, and integrates them into a unified dataset. S57: Utilizing the powerful computing resources of the cloud, it uses a globally optimized SLAM algorithm to process the integrated dataset and construct a high-precision global 3D map. S58: Perform accuracy analysis and error assessment on the constructed 3D map to optimize SLAM algorithm parameters and sensor calibration parameters based on the analysis results. At the same time, train dynamic feature recognition model and path planning model based on historical data to improve model performance. S59: Provide feedback on the optimized algorithm parameters, model, and high-precision 3D map to guide subsequent scanning imaging and path planning.

4. The SLAM real-world three-dimensional spatial scanning imaging method based on a quadruped robot according to claim 1, characterized in that, The S6 includes the following steps: S61: Real-time acquisition of battery power, voltage, current and power consumption data of each device; S62: Analyze the power consumption data of each device and establish a correlation model between device power consumption and working status to predict the energy consumption required to complete the remaining tasks based on the current scanning task progress and environmental conditions. S63: Develop a power consumption adjustment strategy based on the energy consumption prediction results and the current battery level; S64: When the robot is going downhill or decelerating, the energy recovery device can be activated, the drive motor can be switched to generator mode, the kinetic energy can be converted into electrical energy, and the recovered electrical energy can be stored in the battery through the charging management circuit. S65: Real-time feedback of battery status, energy consumption, and adjustment strategies to the robot control system. At the same time, when the battery level is lower than the preset threshold, a low battery alarm is issued to prompt the operator to replace the battery or pause the task in time.

5. A SLAM-based real-world three-dimensional spatial scanning imaging system based on a quadruped robot, characterized in that, include: The adaptive posture compensation module is used to first collect the robot's posture information in real time, calculate the angle and position that the scanning device needs to be adjusted based on the posture information, and then adjust the scanning device in real time. The multi-sensor fusion perception module is used to first collect environmental and robot data through multiple sensors, and then to calibrate, fuse and complement the data collected by different sensors through a multi-sensor data fusion algorithm. The dynamic feature removal and static map optimization module is used to analyze the fused point cloud data and image sequences based on the deep learning dynamic feature recognition model, identify moving dynamic targets, and remove their feature points in the point cloud. Meanwhile, by using a time-series consistency verification algorithm, the static environmental features are optimized to fill in the map blank areas caused by dynamic object occlusion, generating a pure static map without dynamic interference. The dynamic path planning and obstacle avoidance module is used to build a real-time environment map based on the optimized environment data, and to plan the optimal movement path by using a fast exploration random tree algorithm combined with the target area of ​​the scanning task. At the same time, by continuously monitoring dynamic or static obstacles in the environment, the path is adjusted in real time to avoid collisions between the robot and obstacles. The edge computing and cloud collaborative processing module is used to preprocess the scanned data in real time, reducing the amount of data transmission and the complexity of subsequent processing. At the same time, the preprocessed data is transmitted to the cloud server via wireless network, so that the cloud server can use its powerful computing resources to perform high-precision 3D map construction, data analysis and model optimization. The intelligent energy management module is used to monitor battery information in real time, so that it can dynamically adjust the power consumption of each device according to the system's working status and task requirements. At the same time, an energy recovery device is set up to convert some of the kinetic energy into electrical energy and store it in the battery when the robot goes downhill or decelerates, thus extending the system's working time. The dynamic feature removal and static map optimization module includes: a data input module, used to receive fused point cloud data and image sequence data output by the multi-sensor fusion perception module, and simultaneously acquire the robot's motion state information; The dynamic feature preliminary identification module is used to perform target detection on image sequences using the improved YOLOv8 algorithm, identify possible dynamic target candidate regions, and combine optical flow method to calculate the motion vector of pixels in the image, mark the regions with obvious motion trajectories, and fuse the two results to preliminarily determine the dynamic feature regions. The point cloud dynamic feature matching and removal module is used to map the initially identified dynamic feature regions to the point cloud data according to the calibration relationship between the image and the point cloud, determine the corresponding point cloud feature points, and perform motion consistency analysis on the point cloud feature points. If their motion pattern conforms to the characteristics of the dynamic target, they are removed from the point cloud data. The static feature temporal consistency verification module is used to analyze the point cloud data after removing dynamic features according to the time series, calculate the consistency of static feature points at the same spatial location at different times, and fill in the missing static feature regions at a certain time due to dynamic object occlusion by interpolation based on the static feature distribution before and after the time. The clean static map generation module is used to integrate the verified static feature point cloud data, construct a clean static 3D map without dynamic interference, and output it to the dynamic path planning and obstacle avoidance module.

6. The SLAM real-time three-dimensional spatial scanning imaging system based on a quadruped robot according to claim 5, characterized in that, The dynamic path planning and obstacle avoidance module includes: an environment map loading module, which receives a clean static map output by the dynamic feature removal and static map optimization module, and combines it with real-time collected sensor data to construct a real-time environment map containing the current environmental state. The initial path planning module is used to search for a path in the real-time environment map based on the target area of ​​the scanning task and the robot's current position, and to plan an initial optimal path from the current position to the target area. The obstacle real-time monitoring module is used to continuously receive sensor data to monitor whether new dynamic or static obstacles appear in the environment. For dynamic obstacles, it tracks their movement trajectory and predicts their position changes in the future. The path dynamic adjustment module is used to search for a new path to avoid the obstacle in the real-time environment map when an obstacle is detected that may conflict with the current planned path. This is based on the dynamic replanning mechanism of the fast exploration random tree algorithm. The path execution and feedback module is used to send the adjusted path instructions to the quadruped robot's motion control system to control the robot to move along the planned path. At the same time, it collects the robot's actual motion trajectory in real time, compares it with the planned path, and corrects it in time if there is a deviation.

7. The SLAM real-time three-dimensional spatial scanning imaging system based on a quadruped robot according to claim 5, characterized in that, The edge computing and cloud collaborative processing module includes: a data acquisition module, used to receive raw data collected by various sensors; The data compression module is used to compress redundant point cloud data using a compression algorithm based on point cloud features, and to reduce the amount of data by using an adaptive resolution compression method for image data. The noise filtering module is used to remove noise points from point cloud data using algorithms, smooth image data, and eliminate image noise. The feature extraction module is used to extract key feature points from preprocessed point cloud data and feature descriptors from image data. The local storage and transmission module is used to temporarily store the preprocessed data locally and transmit it to the cloud server via a wireless network; The data receiving and integration module is used to receive preprocessed data transmitted from the edge terminal on the cloud server, and then perform time synchronization and spatial registration on data collected at different times and locations, integrating them into a unified dataset. The high-precision 3D map building module is used to utilize the powerful computing resources in the cloud and employ a globally optimized SLAM algorithm to process the integrated dataset and build a high-precision global 3D map. The data analysis and model optimization module is used to perform accuracy analysis and error assessment on the constructed 3D map, so as to optimize the SLAM algorithm parameters and sensor calibration parameters based on the analysis results. At the same time, it trains dynamic feature recognition model and path planning model based on historical data to improve model performance. The results feedback module is used to provide feedback on the optimized algorithm parameters, model, and high-precision 3D map to guide subsequent scanning imaging and path planning.

8. The SLAM real-time three-dimensional spatial scanning imaging system based on a quadruped robot according to claim 5, characterized in that, The intelligent energy management module includes: an energy status monitoring module, used to collect real-time data on battery charge, voltage, current, and power consumption of each device; The power consumption analysis and prediction module is used to analyze the power consumption data of each device, establish a correlation model between device power consumption and working status, and predict the energy consumption required to complete the remaining tasks based on the current scanning task progress and environmental conditions. The dynamic power consumption adjustment module is used to formulate power consumption adjustment strategies based on energy consumption prediction results and the current battery level. The energy recovery control module is used to activate the energy recovery device when the robot is going downhill or decelerating, switch the drive motor to generator mode, convert kinetic energy into electrical energy, and store the recovered electrical energy in the battery through the charging management circuit. The status feedback and alarm module is used to provide real-time feedback on battery status, energy consumption, and adjustment strategies to the robot control system. At the same time, when the battery level is lower than a preset threshold, a low battery alarm is issued to prompt the operator to replace the battery or pause the task in time.

Images