Obstacle avoidance early warning method and system for robot

Abstract

The invention relates to the technical field of obstacle avoidance early warning, in particular to an obstacle avoidance early warning method and system for a robot. The method comprises the following steps: acquiring motion data and moving target point data of a robot; robot motion mode classification is conducted on the robot motion data, and a robot motion mode and a robot static mode are generated; performing environment detection on the moving target point data by using a built-in laser radar of the robot according to the static mode to obtain standard static environment detection data; and planning the path according to the standard static environment detection data, and generating initial running path planning data of the robot. According to the invention, the obstacle avoidance early warning precision and real-time performance of the robot in a complex environment are improved by accurately distinguishing static and dynamic objects, reconstructing the motion trail and planning and adjusting the path.

Description

Technical Field

[0001] The present invention relates to the technical field of obstacle avoidance and early warning, and in particular to an obstacle avoidance and early warning method and system for a robot. Background Art

[0002] Early robot obstacle avoidance mainly relied on simple sensors, such as ultrasonic sensors and infrared sensors, which can trigger obstacle avoidance actions by measuring the distance of obstacles. However, due to their limited perception range and inability to handle complex environmental changes, they have difficulty dealing with dynamic and uncertain obstacles. With the emergence of LiDAR and visual sensors, the accuracy and reliability of robot obstacle avoidance systems have been significantly improved. LiDAR can generate high-precision environmental maps and provide richer spatial information, while computer vision combined with deep learning technology can analyze image and video data, identify obstacles and predict their movement trajectories. These advances enable robots to better understand their surroundings and make obstacle avoidance decisions in complex environments. In recent years, the introduction of deep learning and artificial intelligence has further promoted the development of obstacle avoidance warning methods. Algorithms based on deep neural networks (DNNs) can be trained with a large amount of data to enable robots to learn autonomously and adapt to different environments. However, at present, robots often cannot effectively distinguish between static objects and dynamic objects, resulting in insufficient response of the obstacle avoidance system. At the same time, ignoring the analysis of the movement trajectory of dynamic objects can easily lead to missed obstacle avoidance opportunities or insufficient obstacle avoidance actions, resulting in weak autonomous obstacle avoidance capabilities and low responsiveness of robots. Summary of the invention

[0003] Based on this, it is necessary to provide an obstacle avoidance warning method and system for a robot to solve at least one of the above technical problems.

[0004] To achieve the above object, a robot obstacle avoidance warning method is provided, the method comprising the following steps: Step S1: obtaining robot motion data and moving target point data; classifying the robot motion data into robot motion modes to generate robot motion modes and robot static modes; performing environmental detection on the moving target point data using the robot's built-in laser radar according to the static mode to obtain standard static environmental detection data; Step S2: Plan the path according to the standard static environment detection data to generate the robot's initial operation path planning data; perform dynamic and static object recognition on the robot's initial operation path planning data using the robot's built-in visual camera based on the robot's motion mode to generate robot motion dynamic object recognition data and robot motion static object recognition data; perform static obstacle avoidance warning on the robot motion static object recognition data to generate static object obstacle avoidance warning data; adjust the path planning of the robot's initial operation path planning data using the static object obstacle avoidance warning data to generate robot operation path planning adjustment data; Step S3: reconstructing the dynamic object motion trajectory of the robot motion dynamic object recognition data based on the robot operation path planning adjustment data to generate a dynamic object motion reconstructed trajectory; performing trajectory intersection distance analysis on the robot operation path planning adjustment data and the dynamic object motion reconstructed trajectory to obtain a plane difference collision distance; performing dynamic obstacle avoidance warning based on the plane difference collision distance to generate dynamic object obstacle avoidance warning data; Step S4: encapsulate the static object obstacle avoidance warning data and the dynamic object obstacle avoidance warning data into an obstacle avoidance warning module to generate a robot obstacle avoidance warning module to perform the robot's obstacle avoidance warning operation.

[0005] The present invention can effectively distinguish whether the robot is in motion or static state by acquiring the robot's motion data and target point data and classifying the motion mode, and provide accurate initial data for subsequent obstacle avoidance decision-making and path planning. In addition, environmental detection by laser radar in static mode helps to obtain accurate surrounding environment information and ensure that the robot can effectively avoid obstacles in complex environments. Initial path planning data is generated based on static environment detection data to ensure that the initial path of the robot is reasonable and safe. By using a visual camera to identify dynamic and static objects in path planning, dynamic and static obstacles in the environment can be accurately identified, thereby improving the accuracy of obstacle avoidance. At the same time, the obstacle avoidance warning of static objects can adjust the path planning in real time to avoid collisions between the robot and static obstacles. By reconstructing the motion trajectory of dynamic objects, the future position of dynamic objects can be predicted, so that the robot can better predict and avoid dynamic obstacles and reduce the risk of collision. The trajectory intersection distance analysis and the calculation of the plane difference collision distance help the robot make accurate dynamic obstacle avoidance decisions in complex environments, thereby improving the timeliness and accuracy of obstacle avoidance. The obstacle avoidance warning data of static objects and dynamic objects are modularly packaged to form a comprehensive obstacle avoidance warning module, which improves the integration and executability of the system. This module can control the robot's obstacle avoidance warning operation in real time and effectively, ensuring that the robot can perform seamless and intelligent obstacle avoidance operations in complex environments. The four steps improve the robot's obstacle avoidance ability in dynamic and static environments by comprehensively utilizing multi-sensor data, intelligent path planning, object recognition and motion trajectory analysis, and ensure the accuracy and real-time performance of obstacle avoidance operations. Therefore, the present invention improves the accuracy and real-time performance of the robot's obstacle avoidance warning in complex environments by accurately distinguishing between static and dynamic objects, reconstructing motion trajectories and adjusting path planning.

[0006] Preferably, step S1 comprises the following steps: Step S11: Acquire robot motion data and moving target point data; Step S12: extracting motion speed information from the robot motion data to obtain robot motion speed information data; performing non-zero discrimination on the robot motion speed information data, and when the robot motion speed information data is non-zero, marking the robot in the first mode based on the robot motion speed information data to generate a robot motion mode; Step S13: when the robot motion speed information data is zero, the robot is marked as in the second mode based on the robot motion speed information data to generate a static mode of the robot; according to the static mode, the robot's built-in laser radar is used to detect the environment of the moving target point data to obtain static environment detection data; Step S14: performing data preprocessing on the static environment detection data to generate standard static environment detection data, wherein the data preprocessing includes data cleaning, data denoising, data missing value filling and data standardization.

[0007] The present invention accurately divides the robot operation into "motion mode" and "stationary mode" by distinguishing the non-zero information data of the robot's motion speed. This dynamic and static state marking mechanism helps to allocate resources more efficiently and optimize the system's computing and perception capabilities. In the stationary mode (step S13), the robot's built-in laser radar is used to detect the environment, which can avoid noise interference generated during the movement process, thereby improving the accuracy of environmental data collection and ensuring the reliability of the detection data. Data preprocessing (including data cleaning, denoising, missing value filling and standardization) is performed on the stationary environment detection data, which effectively eliminates redundancy, errors and inconsistencies in the environmental data, improves the quality of the detection data, and provides standardized high-quality data for subsequent processing. The design of steps S12 and S13 ensures that the robot can respond to changes in the motion state in real time. In the motion mode, the robot maintains motion efficiency in a fast marking manner; while in the stationary mode, it can make full use of the stationary state for high-precision environmental detection, providing stronger adaptability for operations in complex environments. Step S11 combines the acquisition of robot motion data and moving target point data to achieve the comprehensive utilization of multi-dimensional data. This design can provide higher flexibility and scalability in multi-target data processing, and provide support for efficient perception and decision-making of intelligent robots in dynamic environments.

[0008] Preferably, performing environmental detection on the moving target point data using a built-in laser radar of the robot according to the stationary mode includes: According to the static mode, the robot's built-in laser radar is used to collect environmental point cloud data of the moving target point data to obtain environmental point cloud data; the environmental point cloud data is roughly aligned to generate environmental point cloud rough alignment data; Perform point cloud fine registration on the rough registration data of the environmental point cloud to generate environmental point cloud fine registration data; perform multi-frame point cloud fusion on the environmental point cloud fine registration data to generate environmental global point cloud fusion data; Perform point cloud segmentation on the global point cloud fusion data of the environment to generate an environmental point cloud segmentation object; perform point cloud meshing on the global point cloud fusion data of the environment according to the environmental point cloud segmentation object to generate environmental point cloud mesh data; perform environmental object category labeling on the environmental point cloud mesh data to generate static environment detection data.

[0009] The present invention gradually improves the accuracy and consistency of point cloud data from point cloud data collection to rough registration and fine registration, and then to multi-frame fusion, ensuring that the generated global point cloud fusion data of the environment can accurately reflect the real characteristics of the target environment. Fusion of multiple time-series frame point cloud data into a global point cloud can not only cover a wider detection range, but also integrate spatiotemporal information, effectively make up for the lack of details of single-frame point cloud data, and provide complete data support for global modeling of the environment. The global point cloud data is segmented and gridded, and the complex three-dimensional point cloud data is structured to generate environmental point cloud grid data. This structured data is convenient for subsequent environmental analysis, target recognition and path planning. By labeling the environmental object categories of the point cloud grid data, static or dynamic objects in the environment are identified and classified, providing a semantic level of environmental understanding. This step helps the robot to achieve accurate positioning, obstacle avoidance and decision-making in a complex environment. The point cloud data after global fusion and gridding is suitable for three-dimensional visualization and analysis, which is convenient for operators or control systems to have a clear perception of the environment.

[0010] Preferably, step S2 comprises the following steps: Step S21: performing target distance plane calculation on the moving target point data according to the standard static environment detection data to obtain the target distance plane distance; performing path planning on the moving target point data according to the target distance plane distance to generate the robot's initial operation path planning data; Step S22: Based on the robot motion mode, using the robot's built-in visual camera to collect motion images of the robot's initial running path planning data to obtain the robot's motion image; performing continuous frame image screening on the robot's motion image to obtain a robot motion frame image set; Step S23: performing dynamic and static object recognition on the robot motion frame image set to generate robot motion dynamic object recognition data and robot motion static object recognition data; Step S24: Perform static obstacle avoidance warning on the robot's moving static object recognition data according to the robot's built-in ultrasonic wave, and generate static object obstacle avoidance warning data; perform path planning adjustment on the robot's initial operation path planning data through the static object obstacle avoidance warning data, and generate robot operation path planning adjustment data.

[0011] The present invention calculates the distance plane between the moving target point and the robot according to the standard static environment detection data, thereby providing an accurate geometric basis for path planning. The initial path is generated based on the target distance plane distance, ensuring the accuracy and rationality of the path planning, and providing an optimized initial navigation solution for the robot. The robot's built-in visual camera collects motion images, and the continuous frame images are screened to construct a motion frame image set, thereby reducing redundant data and image noise, and providing high-quality visual information for subsequent processing. The motion frame image set is used to classify objects, distinguish dynamic and static objects, and provide fine-grained environmental information for the robot's path planning and obstacle avoidance strategy. Dynamic object recognition data can be used to adjust the path in real time to avoid conflicts with moving targets. Static object recognition data provides basic support for static obstacle avoidance and improves the safety and reliability of navigation. Based on the static object recognition data, static object obstacle avoidance warning data is generated by a built-in ultrasonic sensor to effectively predict potential collision risks and provide timely warnings. The initial path is adjusted in real time according to the obstacle avoidance warning data, and optimized operation path planning data is generated to ensure the path safety and efficiency of the robot in a complex environment.

[0012] Preferably, step S23 includes the following steps: Step S231: confirming the image timestamp of the robot motion frame image set to obtain the robot motion frame image timestamp; filtering the robot motion frame image set with adjacent timestamp images by using the robot motion frame image timestamp to obtain the first frame image of the robot motion and the second frame image of the robot motion; Step S232: Overlapping the first frame image of the robot motion and the second frame image of the robot motion to generate a robot motion frame overlap image; segmenting the robot motion frame overlap image into motion change regions to generate robot motion change regions; Step S233: randomly screening the robot motion frame image set based on the timestamp of the robot motion frame image to obtain the third frame image of the robot motion; matching the robot motion change area with the third frame image of the robot motion to generate a matching area of ​​the third frame image of the robot motion; Step S234: performing regional movement position calculation on the robot motion change region and the robot motion third frame image matching region to obtain the change region movement position data; comparing the change region movement position data with a preset change movement position threshold, and when the change region movement position data is greater than or equal to the preset change movement position threshold, marking the region corresponding to the robot motion change region as robot motion dynamic object recognition data; Step S235: When the change area movement position data is less than the preset change movement position threshold, the area corresponding to the robot motion change area is marked as robot motion dynamic still object recognition data.

[0013] The present invention compares the continuous frame data in the robot movement through timestamp confirmation and adjacent frame screening, and accurately captures the dynamic change area. By generating the change area through frame overlap, the specific position and shape of the change in the environment can be clarified, and the accuracy of dynamic recognition can be improved. The introduction of random frame images and the change area for matching analysis ensures that the matching results of the dynamic change area have multi-angle and multi-frame references, which helps to reduce misjudgment and missed judgment. Based on the comparison of the moving position data of the change area with the threshold, the marking strategy is dynamically adjusted: when the movement amplitude of the area is greater than or equal to the preset threshold, it is accurately marked as a dynamic object; the area with a change amplitude lower than the threshold is marked as a static object to avoid dynamic misjudgment. This classification mechanism realizes the accurate distinction between dynamic and static areas and improves the robot's cognitive ability of the environment. The continuous change area is captured by adjacent time frames (the first frame and the second frame), and the random frame (the third frame) is used to match the change area to supplement it, forming a dynamic analysis closed loop in the time and space dimensions to ensure the comprehensiveness of the recognition results. The random frame matching results are combined with the moving position data to further refine the behavior description of the dynamic area, providing more reference value for path planning and obstacle avoidance. The threshold can be adjusted according to the specific scenario requirements, such as setting a higher threshold in a dynamic environment and a lower threshold in a static environment, thereby enhancing the system's scenario adaptability.

[0014] Preferably, step S24 includes the following steps: Step S241: collecting ultrasonic echo signals of the robot's moving static object recognition data according to the robot's built-in ultrasonic wave to generate robot echo signal data; performing time difference calculation processing on the robot echo signal data to generate distance information data; Step S242: Performing intensity analysis on the distance information data to generate signal reflection intensity data; performing feature clustering on the signal reflection intensity data to generate obstacle feature data; generating obstacle avoidance warning signals for the robot motion static object recognition data based on the obstacle feature data to obtain static object obstacle avoidance warning data; Step S243: adjusting the path planning of the robot's initial operation path planning data through the static object obstacle avoidance warning data to generate robot operation path planning adjustment data.

[0015] The present invention uses the built-in ultrasonic sensor of the robot to collect echo signals of static objects, and obtains accurate distance information through time difference calculation. Through accurate time difference analysis, the robot can calculate the distance to the obstacle in real time, enhance its perception of the surrounding environment, and ensure safe operation. The signal reflection intensity analysis is performed on the distance information data to identify the reflection characteristics of different objects, and the obstacle object feature data is generated through feature clustering. The signal reflection intensity combined with cluster analysis provides the robot with more diverse obstacle identification information, which is not limited to the distance of the object, but can also infer its material, size and other properties according to the reflection intensity of the object. This analysis can help the robot distinguish different types of obstacles (such as walls, pillars, furniture, etc.), so as to take more appropriate measures when avoiding obstacles. Obstacle avoidance warning signals are generated according to the obstacle object feature data, and the existence of static objects is fed back in time, thereby providing a basis for path adjustment. By generating obstacle avoidance warning signals in real time, the robot can identify and process static obstacles in time to avoid collision or obstacle obstruction. An active defense mechanism is provided for the robot to ensure that the robot passes smoothly in a complex environment. The initial operation path of the robot is dynamically adjusted according to the static object obstacle avoidance warning data to generate new path planning data. The robot can adjust the original path according to the real-time detected obstacle position and characteristics to avoid conflicts with obstacles and ensure the smooth progress of the task. The adjusted path planning not only ensures safety, but also improves the efficiency of robot movement and avoids unnecessary detours. Combining the precise distance measurement of ultrasound and the analysis of signal reflection intensity, it can respond quickly in dynamic or changing environments and adapt to different obstacle scenarios. The robot can respond flexibly according to the complexity of the environment and the type of obstacles to achieve efficient path planning.

[0016] Preferably, step S243 includes the following steps: Step S2431: segmenting the initial operation path planning data of the robot to generate path segment data; extracting key nodes from the path segment data to generate node feature data; performing feasibility evaluation on the node feature data to generate path evaluation data; Step S2432: Calculate the safety distance of the static object obstacle avoidance warning data according to the path evaluation data to generate safety boundary data; perform path constraints on the safety boundary data to generate constraint condition data; generate alternative paths for the constraint condition data to generate path plan data; Step S2433: perform path cost evaluation on the path plan data to generate path cost evaluation data; select the optimal path for the path plan data based on the path cost evaluation data to generate optimized path data; use the optimized path data to adjust the path planning of the robot's initial operation path planning data to generate robot operation path planning adjustment data.

[0017] The present invention divides the initial path of the robot into multiple sections, and makes the path planning more refined through segmentation, which is convenient for local adjustment and optimization. The key nodes in the path are extracted so that the robot can focus on the key path points and important turning points. This segmentation method can improve the efficiency of path adjustment and ensure that the robot can more accurately identify and deal with each important obstacle or path change point. By evaluating the feasibility of the path, it can be ensured that each path is operable to avoid route deviation or accidents caused by infeasible paths. The extraction of path segmentation and node features helps the robot to handle complex environments more intelligently and improve the flexibility and reliability of path planning. According to the obstacle avoidance warning data of static objects, the safe distance from the obstacle is calculated to ensure that the robot always maintains an appropriate safety boundary to avoid collision. By setting constraints, different alternative path plans are generated, which can provide more options for the robot and avoid task interruption due to the failure of a single path. The calculation of the safe distance ensures that the robot can reasonably avoid obstacles and reduce the possibility of accidents. The generation of alternative paths provides the robot with a variety of path planning options, which improves the ability to adapt to complex environments and the success rate of task completion. The cost of each candidate path is evaluated, taking into account factors such as travel time, energy consumption, and complexity, so as to select the path with the lowest cost. Based on the path cost evaluation data, the path with the best cost is selected to ensure that the robot can complete the task in the best way while ensuring safety. Cost evaluation ensures that the robot chooses the best path in terms of time, energy, cost, etc., thereby improving the efficiency of the robot's task completion.

[0018] Preferably, step S3 comprises the following steps: Step S31: performing dynamic object motion analysis on the robot motion dynamic object recognition data to obtain dynamic object motion analysis data; performing object motion orientation extraction on the dynamic object motion analysis data to generate dynamic object motion direction data; Step S32: reconstructing the dynamic object motion trajectory of the dynamic object motion analysis data according to the dynamic object motion direction data to generate a dynamic object motion reconstructed trajectory; Step S33: performing trajectory intersection judgment on the robot operation path planning adjustment data and the dynamic object motion reconstructed trajectory. When the robot operation path planning adjustment data and the dynamic object motion reconstructed trajectory intersect, the plane difference collision distance between the robot operation path planning adjustment data and the dynamic object motion reconstructed trajectory is calculated to obtain the plane difference collision distance; Step S34: Perform dynamic obstacle avoidance warning based on the plane difference collision distance to generate initial dynamic object obstacle avoidance warning data; segment the plane difference collision distance to generate collision distance segmentation data; adjust the warning volume of the initial dynamic object obstacle avoidance warning data through the collision distance segmentation data to generate dynamic object obstacle avoidance warning data.

[0019] The present invention can identify the motion law of the object and extract key information such as speed, acceleration and other features by analyzing the motion of the dynamic object, which helps the robot to understand the motion state of the objects in the surrounding environment in real time. The dynamic object motion analysis helps the robot better predict the movement trend of the surrounding objects and take obstacle avoidance measures in advance. By extracting the motion direction data, the robot can respond more quickly and reduce the risk of collision with the dynamic object. Based on the object motion direction data, the motion trajectory of the dynamic object is reconstructed to generate an accurate dynamic object motion path. This step enables the robot to grasp the motion route of the dynamic object in real time, so as to make path planning adjustments. The reconstruction of the motion trajectory of the dynamic object ensures that the robot can accurately capture the changing trajectory of the dynamic object, thereby providing an accurate decision-making basis for subsequent obstacle avoidance. Through accurate trajectory reconstruction, the robot can predict the possibility of intersection with the dynamic object in advance and respond quickly. When the robot path and the motion trajectory of the dynamic object intersect, the collision distance is calculated and judged. In this way, the robot can accurately identify when a collision will occur and give an early warning. Through trajectory intersection judgment, the robot can perceive the potential collision risk in time and provide sufficient time to avoid or adjust the path. This step can ensure that the robot avoids collisions with moving objects and significantly improves the safety of operation. Based on the calculated plane difference collision distance, an obstacle avoidance warning signal for dynamic objects is generated. By further segmenting the collision distance, the warning volume is dynamically adjusted to ensure that the obstacle avoidance signal is significant enough so that the robot can better avoid obstacles. The generation of dynamic obstacle avoidance warnings allows the robot to respond promptly when encountering potential collisions, reducing the probability of collisions. By segmenting the collision distance, the robot can adjust the warning volume according to different situations to ensure that the warning signal can be effectively transmitted in various environments.

[0020] Preferably, reconstructing the motion trajectory of the dynamic object based on the dynamic object motion analysis data according to the dynamic object motion direction data includes: Perform time series segmentation on the dynamic object motion direction data to generate motion direction change points; perform dynamic object continuous position estimation on the dynamic object motion analysis data through the motion direction change points to generate dynamic object preliminary position trajectory data; The preliminary position trajectory data of the dynamic object is converted into three-dimensional trajectory to generate the three-dimensional data of the dynamic object position trajectory; the three-dimensional data of the dynamic object position trajectory is bound to the trajectory time sequence to generate the spatiotemporal motion trajectory of the dynamic object; the spatiotemporal motion trajectory of the dynamic object is classified and labeled to generate the dynamic object motion reconstruction trajectory.

[0021] The present invention extracts key change points of the direction of motion by segmenting the time series of the motion direction data of the dynamic object, and provides a basis for subsequent trajectory estimation. The extraction of the change points of the direction of motion helps the robot identify the key moments of the dynamic object change, avoiding the computational burden caused by too frequent data updates. By dividing the time series and identifying the change points, the robot can accurately grasp the movement changes of the dynamic object and improve the accurate prediction of the trajectory. Continuous position estimation is performed based on the change points of the motion direction, so as to obtain the preliminary trajectory information of the dynamic object. Through this process, the robot can understand the movement trend of the dynamic object earlier. Continuous position estimation provides the robot with the preliminary trajectory of the dynamic object, ensuring that the path planning of the robot is more stable in a complex environment. The generation of the preliminary trajectory can avoid misjudgment caused by data noise, and ensure that the robot's obstacle avoidance and dynamic object prediction are more accurate. The three-dimensional trajectory helps the robot better understand the position and movement of objects in the environment and improves the spatial perception ability. Through three-dimensional data, the robot can more accurately judge the actual motion range of the object and make corresponding obstacle avoidance decisions. Time series binding ensures that the robot can understand the position changes of the dynamic object in both time and space dimensions at the same time, and enhances the robot's ability to adapt to dynamic environments. The generation of spatiotemporal trajectories provides robots with more comprehensive dynamic object motion prediction capabilities, improving the timeliness and accuracy of responses. Trajectory classification and annotation help robots identify different motion modes of objects, making it easier to make targeted obstacle avoidance or path adjustments. Annotation and classification provide a clearer and more structured data model, making the motion trajectories of dynamic objects more efficient in processing and management.

[0022] In this specification, an obstacle avoidance warning system for a robot is provided, which is used to execute the above-mentioned obstacle avoidance warning method for a robot. The obstacle avoidance warning system for a robot includes: The motion mode classification module is used to obtain robot motion data and moving target point data; classify the robot motion mode of the robot motion data to generate the robot motion mode and the robot static mode; according to the static mode, the robot's built-in laser radar is used to detect the environment of the moving target point data to obtain standard static environment detection data; The static risk avoidance warning module is used to plan the path according to the standard static environment detection data and generate the robot's initial operation path planning data; based on the robot's motion mode, the robot's built-in visual camera is used to perform dynamic and static object recognition on the robot's initial operation path planning data to generate robot motion dynamic object recognition data and robot motion static object recognition data; static obstacle avoidance warning is performed on the robot's motion static object recognition data to generate static object obstacle avoidance warning data; path planning adjustment is performed on the robot's initial operation path planning data through the static object obstacle avoidance warning data to generate robot operation path planning adjustment data; The dynamic risk avoidance warning module is used to reconstruct the dynamic object motion trajectory of the robot motion dynamic object recognition data based on the robot motion path planning adjustment data to generate the dynamic object motion reconstruction trajectory; perform trajectory intersection distance analysis on the robot motion path planning adjustment data and the dynamic object motion reconstruction trajectory to obtain the plane difference collision distance; perform dynamic obstacle avoidance warning based on the plane difference collision distance to generate dynamic object obstacle avoidance warning data; The warning encapsulation module is used to encapsulate the static object obstacle avoidance warning data and the dynamic object obstacle avoidance warning data into an obstacle avoidance warning module to generate a robot obstacle avoidance warning module to perform the robot's obstacle avoidance warning operation.

[0023] The beneficial effect of the present invention is that by classifying the robot motion data, it is distinguished whether the robot is in motion mode or static mode, and accurate state information is provided for subsequent path planning and environmental perception. By using laser radar to detect the environment in static mode, standard static environment data can be generated, which provides a basis for the safe navigation of the robot in a static state and avoids collisions with static obstacles. The standard static environment detection data is used for path planning, and the dynamic and static objects are identified in combination with the robot motion mode, which improves the robot's obstacle avoidance ability in complex environments. Through the obstacle avoidance warning of static objects, the robot can adjust the path in real time to avoid static obstacles, ensuring the safety and reliability of the driving path. Based on the reconstruction of the path planning adjustment data and the motion trajectory of the dynamic object, the dynamic risk avoidance warning module can predict the future position of the dynamic obstacle, and evaluate the potential collision risk through the trajectory intersection distance analysis. The module can calculate the plane difference collision distance in real time in a dynamic environment, generate the dynamic object obstacle avoidance warning in time, enhance the robot's adaptability and reaction speed in a dynamic environment, and avoid collisions with dynamic obstacles. The obstacle avoidance warning data of static and dynamic objects are modularly packaged to form a unified obstacle avoidance warning system, which improves the overall coordination and operability of the system. This module ensures that the robot can consider both static and dynamic obstacles, and performs obstacle avoidance tasks through centralized warning, optimizes the robot's obstacle avoidance strategy, and improves the efficiency and accuracy of the obstacle avoidance process. Therefore, the present invention improves the accuracy and real-time performance of the robot's obstacle avoidance warning in complex environments by accurately distinguishing static and dynamic objects, reconstructing motion trajectories, and adjusting path planning. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 A schematic diagram of the steps of an obstacle avoidance warning method for a robot; Figure 2 for Figure 1 Detailed implementation steps of step S2 in the flowchart; Figure 3 for Figure 1Detailed implementation steps of step S3 in FIG. The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0025] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0026] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0027] It should be understood that although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit, and the term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0028] To achieve this, please refer to Figures 1 to 3 , an obstacle avoidance warning method for a robot, the method comprising the following steps: Step S1: obtaining robot motion data and moving target point data; classifying the robot motion data into robot motion modes to generate robot motion modes and robot static modes; performing environmental detection on the moving target point data using the robot's built-in laser radar according to the static mode to obtain standard static environmental detection data; Step S2: Plan the path according to the standard static environment detection data to generate the robot's initial operation path planning data; perform dynamic and static object recognition on the robot's initial operation path planning data using the robot's built-in visual camera based on the robot's motion mode to generate robot motion dynamic object recognition data and robot motion static object recognition data; perform static obstacle avoidance warning on the robot motion static object recognition data to generate static object obstacle avoidance warning data; adjust the path planning of the robot's initial operation path planning data using the static object obstacle avoidance warning data to generate robot operation path planning adjustment data; Step S3: reconstructing the dynamic object motion trajectory of the robot motion dynamic object recognition data based on the robot operation path planning adjustment data to generate a dynamic object motion reconstructed trajectory; performing trajectory intersection distance analysis on the robot operation path planning adjustment data and the dynamic object motion reconstructed trajectory to obtain a plane difference collision distance; performing dynamic obstacle avoidance warning based on the plane difference collision distance to generate dynamic object obstacle avoidance warning data; Step S4: encapsulate the static object obstacle avoidance warning data and the dynamic object obstacle avoidance warning data into an obstacle avoidance warning module to generate a robot obstacle avoidance warning module to perform the robot's obstacle avoidance warning operation.

[0029] The present invention can effectively distinguish whether the robot is in motion or static state by acquiring the robot's motion data and target point data and classifying the motion mode, and provide accurate initial data for subsequent obstacle avoidance decision-making and path planning. In addition, environmental detection by laser radar in static mode helps to obtain accurate surrounding environment information and ensure that the robot can effectively avoid obstacles in complex environments. Initial path planning data is generated based on static environment detection data to ensure that the initial path of the robot is reasonable and safe. By using a visual camera to identify dynamic and static objects in path planning, dynamic and static obstacles in the environment can be accurately identified, thereby improving the accuracy of obstacle avoidance. At the same time, the obstacle avoidance warning of static objects can adjust the path planning in real time to avoid collisions between the robot and static obstacles. By reconstructing the motion trajectory of dynamic objects, the future position of dynamic objects can be predicted, so that the robot can better predict and avoid dynamic obstacles and reduce the risk of collision. The trajectory intersection distance analysis and the calculation of the plane difference collision distance help the robot make accurate dynamic obstacle avoidance decisions in complex environments, thereby improving the timeliness and accuracy of obstacle avoidance. The obstacle avoidance warning data of static objects and dynamic objects are modularly packaged to form a comprehensive obstacle avoidance warning module, which improves the integration and executability of the system. This module can control the robot's obstacle avoidance warning operation in real time and effectively, ensuring that the robot can perform seamless and intelligent obstacle avoidance operations in complex environments. The four steps improve the robot's obstacle avoidance ability in dynamic and static environments by comprehensively utilizing multi-sensor data, intelligent path planning, object recognition and motion trajectory analysis, and ensure the accuracy and real-time performance of obstacle avoidance operations. Therefore, the present invention improves the accuracy and real-time performance of the robot's obstacle avoidance warning in complex environments by accurately distinguishing between static and dynamic objects, reconstructing motion trajectories and adjusting path planning.

[0030] In the embodiment of the present invention, reference Figure 1 FIG. 1 is a schematic diagram of a process flow of an obstacle avoidance warning method for a robot according to the present invention. In this example, the obstacle avoidance warning method for a robot includes the following steps: Step S1: obtaining robot motion data and moving target point data; classifying the robot motion data into robot motion modes to generate robot motion modes and robot static modes; performing environmental detection on the moving target point data using the robot's built-in laser radar according to the static mode to obtain standard static environmental detection data; In the embodiment of the present invention, the robot's motion control system or inertial measurement unit (IMU) is used to obtain: the real-time position of the robot in two-dimensional or three-dimensional space, which is usually obtained through sensors (such as GPS, encoders or visual systems). The linear velocity and angular velocity of the robot at each moment. The acceleration data obtained by the IMU reflects the change of the robot's motion state. Describe the angle of the robot relative to the world coordinate system, such as pitch, roll and yaw angle, which are usually obtained through a gyroscope. Usually include the spatial position (coordinates), movement direction, speed and other information of the target, which can represent the dynamic position change of the target in the environment. For example, the global positioning coordinates (GPS) of the target or the relative coordinates in the local coordinate system. Including dynamic change information such as the speed, acceleration, direction of the target. Before pattern classification, the robot motion data is standardized and cleaned to remove noise and outliers, and fill in missing values. The sliding window method can usually be used to smooth the data to ensure the continuity and accuracy of the data. It can be judged whether the robot is moving based on the speed and acceleration thresholds of the robot. For example, if the speed is greater than a certain threshold, the robot is judged to be in motion; if the acceleration is less than a preset value and the speed is close to zero, the robot is judged to be stationary. Use classification models (such as decision trees, support vector machines, K-means clustering, etc.) to train the robot's motion data and automatically classify the robot's motion mode based on historical data. When the robot's linear velocity is greater than a certain threshold (such as 0.1m / s) and the acceleration is greater than a certain threshold (such as 0.5m / s²), it is determined to be in motion mode. When the robot's speed is lower than 0.1m / s and the acceleration is close to zero (such as ±0.05m / s²), it is determined to be in static mode. When the robot's speed is lower than 0.1m / s and the acceleration is close to zero (such as ±0.05m / s²), it is determined to be in static mode. In static mode, the robot uses the built-in laser radar (LiDAR) to scan and perceive the surrounding environment and collect the point cloud data reflected by the laser radar. The working principle of laser radar: the laser transmitter emits a laser beam, which hits the surface of the object and reflects. The receiver receives the reflected light signal and obtains the distance information of the object by calculating the difference between the emission time and the reception time. According to the environment in which the robot is located, the laser radar performs omnidirectional detection by rotating 360 degrees. LiDAR generates a large amount of point cloud data, each point represents the spatial coordinates (x, y, z) of a reflected signal. Remove noise and abnormal data from the point cloud and retain valid reflection points. Convert the point cloud data acquired by the LiDAR to the robot's coordinate system or global coordinate system. Point cloud data can be processed by clustering algorithms (such as Euclidean clustering) to identify surrounding obstacles, target points, etc. Based on the point cloud data collected by the LiDAR, the following types of data are generated: Each point contains position coordinates (x, y, z) and reflection intensity information to describe the environment around the robot.Obstacle positions are extracted from the point cloud, and irrelevant ground points can be filtered out through the ground detection algorithm. Target points are identified from the environmental point cloud, and the relative position and distance between the target and the robot are calculated for subsequent target tracking and obstacle avoidance. The generated point cloud data can be converted into a grid map, a 3D model or other standard format for subsequent processing and application.

[0031] Step S2: Plan the path according to the standard static environment detection data to generate the robot's initial operation path planning data; perform dynamic and static object recognition on the robot's initial operation path planning data using the robot's built-in visual camera based on the robot's motion mode to generate robot motion dynamic object recognition data and robot motion static object recognition data; perform static obstacle avoidance warning on the robot motion static object recognition data to generate static object obstacle avoidance warning data; adjust the path planning of the robot's initial operation path planning data using the static object obstacle avoidance warning data to generate robot operation path planning adjustment data; In an embodiment of the present invention, the point cloud data provided by the laser radar can be used to describe the environment around the robot, including obstacles, free space, target points, etc. The point cloud data is converted into an occupancy grid map or other models suitable for path planning. The obstacles and walkable areas of the environment can be represented by a grid-based method or a topological map, and common path planning algorithms such as the Dijkstra algorithm can be used for path planning from the current position of the robot to the target position. When planning the path, it is necessary to consider the obstacle information in the environment to avoid planning the obstacle area. A preliminary robot operation path is generated according to the environment model and the target position. The built-in visual camera of the robot is used to capture the environment image in real time to obtain more detailed environmental information. The image data captured by the camera will be processed differently according to the robot's motion mode (motion mode or static mode). In motion mode, the camera needs to identify dynamic objects (such as pedestrians, moving obstacles, etc.) in the area along the path in real time. In static mode, the camera can identify static objects (such as fixed obstacles such as walls and tables). Detect and track moving objects through image processing algorithms (such as background subtraction, optical flow method, YOLO and other target detection methods). Dynamic objects will affect the robot's path planning, and the path needs to be updated dynamically. Using visual cameras to identify fixed obstacles, deep learning target detection algorithms (such as YOLO and Faster R-CNN) can be used to identify static obstacles. Static objects remain unchanged, but affect the robot's current path, especially in static mode, these objects need to be avoided. For identified static objects (such as walls, tables and chairs, etc.), the robot determines whether it will collide with these objects based on the path planning data. Use collision detection algorithms, such as distance-based or bounding box-based methods, to determine whether the robot's current path will contact static objects. When the robot approaches a static obstacle, the algorithm calculates the potential collision risk and generates warning data. The warning data includes: the distance of the static object, the collision risk value (such as the contact risk), the warning time, etc. The static object obstacle avoidance warning data is used to determine whether the robot's initial running path needs to be adjusted. If the robot's current path is close to a static object and there is a risk of collision, the system will recalculate and adjust the path. Replan the local path through the RRT algorithm (rapidly expanding random tree) or local obstacle avoidance algorithm (such as Dynamic Window Approach) to avoid static obstacles. If a static obstacle is very large or in a critical position, the global path can be recalculated to avoid that area. Based on the static object obstacle avoidance warning data, the robot can choose a path to bypass the static object, or adjust the robot's speed, posture and other parameters to avoid collision with obstacles. Dynamic programming or reinforcement learning methods can be used to adjust the path in real time.

[0032] Step S3: reconstructing the dynamic object motion trajectory of the robot motion dynamic object recognition data based on the robot operation path planning adjustment data to generate a dynamic object motion reconstructed trajectory; performing trajectory intersection distance analysis on the robot operation path planning adjustment data and the dynamic object motion reconstructed trajectory to obtain a plane difference collision distance; performing dynamic obstacle avoidance warning based on the plane difference collision distance to generate dynamic object obstacle avoidance warning data; In an embodiment of the present invention, a kinematic model (such as a uniform linear motion model, an acceleration model, a Kalman filter model, etc.) is used to reconstruct the trajectory according to the current position, velocity, acceleration and other data of the dynamic object. The trajectory of the object in the future period of time is inferred using these data. The motion of a dynamic object is a time series process, so the position of the object at multiple moments in the future can be predicted by time series inference. This process requires the use of the object's velocity and acceleration data to infer its trajectory. Based on the derivation of the model and the time series inference, the trajectory of the dynamic object is reconstructed and used for subsequent path analysis and obstacle avoidance operations to generate the future motion trajectory of the dynamic object, which shows the future movement path of the object in space. This data will be compared with the robot path to help determine whether there is a collision risk. The robot path and the motion trajectory of the dynamic object are discretized into a series of points. Each point represents the position of the robot and the position of the dynamic object, and a new data point is generated at each time step. The spatial distance between each pair of robot path points and dynamic object trajectory points is calculated. All discrete points are compared point by point to find the shortest distance between the robot path and the motion trajectory of the dynamic object. The shortest distance between the robot path and the object trajectory is calculated by Euclidean distance, shortest distance or other related algorithms. This distance can be used to assess the risk of collision. Plane difference refers to the path difference calculated in a two-dimensional plane, taking into account the difference between the position of the object on the two-dimensional plane and the robot path. If this difference is small enough, it means that the paths of the two intersect and a collision will occur. A plane difference collision distance data set is generated, which contains the minimum distance between the robot path point and the dynamic object trajectory point. Through this data, it can be determined whether there is a potential collision risk. Set a safe distance threshold according to the actual situation. For example, when the plane difference collision distance is less than the set 1 meter, the system considers that the risk of collision is high and obstacle avoidance is required. According to the plane difference collision distance and the set threshold, it is determined whether there is a potential collision risk. If there is a collision risk, a dynamic obstacle avoidance warning is triggered. If the warning is triggered, the robot will start the obstacle avoidance strategy. Common obstacle avoidance strategies include: recalculating a safe path through local path planning algorithms (such as RRT, DWA, etc.) to enable the robot to avoid dynamic objects. According to the relative speed and distance of the dynamic object, the robot can choose to slow down, stop or detour to avoid collision with the object. During the robot's movement, the path is calculated and adjusted in real time to ensure that the robot always remains at a safe distance.

[0033] Step S4: encapsulate the static object obstacle avoidance warning data and the dynamic object obstacle avoidance warning data into an obstacle avoidance warning module to generate a robot obstacle avoidance warning module to perform the robot's obstacle avoidance warning operation.

[0034] In an embodiment of the present invention, static object obstacle avoidance warning data and dynamic object obstacle avoidance warning data are encapsulated in a standardized manner. Data structures (such as JSON, XML or custom data classes) can be used to organize these data so that they can be quickly accessed and processed in the module. The encapsulated static object obstacle avoidance warning data and dynamic object obstacle avoidance warning data are integrated into a robot obstacle avoidance warning module to perform obstacle avoidance warning work. The design and development of the robot obstacle avoidance warning module mainly include the following steps: The robot obstacle avoidance warning module needs to design an input interface to receive data packets from different sources, such as environmental monitoring systems, sensor networks, robot status monitoring, etc. After the data packet is input, the obstacle avoidance module will parse the warning information of static objects and dynamic objects through the data processing unit. The processing logic includes: static obstacle detection: analyze static object obstacle avoidance warning data, check whether the robot path collides with static obstacles, and generate relevant obstacle avoidance instructions. Dynamic obstacle detection: analyze dynamic object obstacle avoidance warning data, check the intersection of the robot path and the dynamic object trajectory, and generate obstacle avoidance decisions. According to the obstacle avoidance warning information of static objects and dynamic objects, the robot obstacle avoidance warning module needs to execute the corresponding decision logic: if the warning data of static objects or dynamic objects indicates a collision risk, the module will generate an obstacle avoidance command according to the preset obstacle avoidance strategy (such as replanning the path, adjusting the speed, etc.). In the case of multiple obstacle avoidance tasks, the module will determine the priority obstacle avoidance tasks based on factors such as the priority, distance, and speed of the collision risk. If multiple obstacle avoidance tasks are triggered at the same time, the module will use algorithms to perform collaborative obstacle avoidance to ensure that the robot avoids collisions with multiple obstacles. The obstacle avoidance instructions generated by the decision engine will be sent to the robot control system. These instructions include path adjustment, speed control, posture adjustment, etc., to ensure that the robot can operate safely in a dynamic environment.

[0035] Preferably, step S1 comprises the following steps: Step S11: Acquire robot motion data and moving target point data; Step S12: extracting motion speed information from the robot motion data to obtain robot motion speed information data; performing non-zero discrimination on the robot motion speed information data, and when the robot motion speed information data is non-zero, marking the robot in the first mode based on the robot motion speed information data to generate a robot motion mode; Step S13: when the robot motion speed information data is zero, the robot is marked as in the second mode based on the robot motion speed information data to generate a static mode of the robot; according to the static mode, the robot's built-in laser radar is used to detect the environment of the moving target point data to obtain static environment detection data; Step S14: performing data preprocessing on the static environment detection data to generate standard static environment detection data, wherein the data preprocessing includes data cleaning, data denoising, data missing value filling and data standardization.

[0036] In an embodiment of the present invention, real-time motion data is obtained from the motion control system of the robot, which generally includes the current position coordinates (x, y, z), motion direction, and target point coordinates of the robot. For example, the robot has the following data in spatial coordinates: (x0, y0, z0) = (2.5, 3.0, 1.2). Information about the target point of the robot's movement is collected, for example, the spatial position of the target point (xt, yt, zt) and the distance from the target. Assume that the target point position is (xt, yt, zt) = (5.0, 7.5, 1.2). The current speed of the robot is calculated based on the time interval. Assuming that the robot updates its position once per second, the current position change of the robot is Δx = 0.1 meters, Δy = 0.2 meters, and Δz = 0.0 meters, then the robot speed can be calculated as: 0.2236 meters / second; if v≠0, the robot is in a moving state. If the speed information data is non-zero, a robot motion mode flag (e.g., "motion mode") is generated. If the robot motion speed information v=0 (for example, the robot is stationary), the robot enters the second mode and generates a robot stationary mode mark (for example, "stationary mode"). In stationary mode, the robot's built-in laser radar system is used to detect the surrounding environment. Assuming that the laser radar scanning range is 360° and the maximum detection distance is 10 meters, the generated stationary environment detection data includes the position and distance of the surrounding obstacles, such as: obstacle position = [(3.5,5.0), (4.0,3.2), (6.0,7.8)], where each pair of brackets represents the spatial position of an obstacle. Remove invalid or abnormal values ​​from the environmental detection data, such as data whose detection distance exceeds the valid range. Assuming that the distance of a certain detection point is 15 meters, which exceeds the maximum effective detection distance of 10 meters, the data needs to be cleared. Apply filtering algorithms (such as median filtering, Kalman filtering, etc.) to remove noise from the laser radar data. Assuming that some distance data is interfered by noise, such as the detection data near 5.0 meters has a large deviation, after filtering, it is corrected to 4.8 meters. Interpolate missing environmental data. Common methods include linear interpolation and nearest neighbor interpolation. Assume that the detection data in a certain direction is missing, and fill in the missing values ​​through linear interpolation. Standardize the processed data so that all data are in the same scale range (for example, normalize to the range [0, 1]).

[0037] Preferably, performing environmental detection on the moving target point data using a built-in laser radar of the robot according to the stationary mode includes: According to the static mode, the robot's built-in laser radar is used to collect environmental point cloud data of the moving target point data to obtain environmental point cloud data; the environmental point cloud data is roughly aligned to generate environmental point cloud rough alignment data; Perform point cloud fine registration on the rough registration data of the environmental point cloud to generate environmental point cloud fine registration data; perform multi-frame point cloud fusion on the environmental point cloud fine registration data to generate environmental global point cloud fusion data; Perform point cloud segmentation on the global point cloud fusion data of the environment to generate an environmental point cloud segmentation object; perform point cloud meshing on the global point cloud fusion data of the environment according to the environmental point cloud segmentation object to generate environmental point cloud mesh data; perform environmental object category labeling on the environmental point cloud mesh data to generate static environment detection data.

[0038] In an embodiment of the present invention, the robot uses a laser radar to scan the surrounding environment in a stationary mode to obtain environmental point cloud data. Assuming that the scanning range of the laser radar is 360° and the maximum detection distance is 10 meters, the collected environmental point cloud data generally includes the position coordinates (x, y, z) of each point in space and the reflection intensity of each point. For example, the environmental point cloud data collected by the laser radar (partial example) is as follows: point cloud data = {(x1, y1, z1), (x2, y2, z2), ..., (xn, yn, zn)}, where each point (xi, yi, zi) represents the position of a laser radar detection point. Since the environmental point cloud data contains noise or the scanning angle is inconsistent, a rough alignment is first performed to preliminarily align the point cloud data collected at different angles or positions. Use a feature-based alignment method (such as SIFT and SURF feature matching) to preliminarily align multiple point cloud data frames. Multiple point cloud data sets come from different viewpoints or time periods. After coarse registration, the coarse registration data obtained will be: Coarse registration point cloud data = {(x1′, y1′, z1′), (x2′, y2′, z2′), ..., (xm′, ym′, zm′)}, where the coordinates of each point are the results of coarse registration. Use the iterative closest point (ICP) algorithm or its variants to perform fine registration on the coarsely registered point cloud. The ICP algorithm minimizes the distance error between point clouds, so that point cloud data collected from different viewpoints or positions can be accurately aligned. In the fine registration process, it is assumed that the point cloud data has undergone slight translation and rotation in the x and y axis directions. The point cloud data after fine registration is as follows: Fine registration point cloud data = {(x1′′, y1′′, z1′′), (x2′′, y2′′, z2′′), ..., (xk′′, yk′′, zk′′)} These point cloud data correspond to the same coordinate system more accurately. The multiple precisely aligned point cloud data are fused into a global environmental point cloud. Common methods are weighted averaging based on overlapping areas or global optimization methods (such as joint optimization) to fuse different frame data. The existing precisely aligned point cloud data is multiple frames, and the fused global point cloud data will be: global point cloud data = {(x1′′′′, y1′′′′, z1′′′′), (x2′′′′, y2′′′′, z2′′′′), ..., (xp′′′′, yp′′′′′, zp′′′′)} This dataset contains the point cloud information of all frames and is data in a global unified coordinate system. The global point cloud is processed by a point cloud segmentation algorithm (for example, based on region growing, Euclidean clustering, etc.) to identify different environmental objects (such as walls, obstacles, ground, etc.). Each object or area is assigned to an independent point cloud object.After point cloud segmentation, the following objects are generated: Segmented object = {O1 = {(x1, y1, z1), (x2, y2, z2), ...}, O2 = {(x3, y3, z3), ...}, ...} Each Oi represents an independent object point cloud object, representing an obstacle or other environmental element. The segmented point cloud data is converted into a mesh model, usually using a triangular mesh (such as Delaunay triangulation) or a voxel mesh (Voxel) method to generate a surface or volume model of each object. After a segmented object O1 is meshed, the generated mesh data is as follows: Meshed point cloud data = {(x1′, y1′, z1′), (x2′, y2′, z2′), ...} This mesh data is used for higher-level analysis and visualization. Object classification and annotation are performed on the meshed point cloud data through machine learning or rule-based classification methods. For example, objects such as walls, tables, chairs, etc. can be annotated, or obstacles can be distinguished from walkable areas. The classification result is: Labeling result = {(O1, wall), (O2, table), (O3, chair)}, where each object Oi is labeled as a corresponding category. The final static environment detection data includes all labeled object category information and the grid data of the environment point cloud. This data can be used for subsequent path planning, obstacle detection and other tasks.

[0039] As an example of the present invention, refer to Figure 2 As shown, in this example, step S2 includes: Step S21: performing target distance plane calculation on the moving target point data according to the standard static environment detection data to obtain the target distance plane distance; performing path planning on the moving target point data according to the target distance plane distance to generate the robot's initial operation path planning data; Step S22: Based on the robot motion mode, using the robot's built-in visual camera to collect motion images of the robot's initial running path planning data to obtain the robot's motion image; performing continuous frame image screening on the robot's motion image to obtain a robot motion frame image set; Step S23: performing dynamic and static object recognition on the robot motion frame image set to generate robot motion dynamic object recognition data and robot motion static object recognition data; Step S24: Perform static obstacle avoidance warning on the robot's moving static object recognition data according to the robot's built-in ultrasonic wave, and generate static object obstacle avoidance warning data; perform path planning adjustment on the robot's initial operation path planning data through the static object obstacle avoidance warning data, and generate robot operation path planning adjustment data.

[0040] In an embodiment of the present invention, the plane distance of the target point is obtained by calculating the distance between each target point and the current position of the robot according to the environmental data collected in the robot's static mode. These data will help define the feasible path range of the robot and ensure that the planned path avoids obstacles. Using the acquired target distance plane information, a path planning algorithm (such as Dijkstra algorithm) is used to calculate a shortest path from the starting point to the target point for the robot. The path planning data includes multiple points that the robot needs to pass through, helping the robot to achieve optimal travel. During the movement, the robot collects images through its built-in visual camera. These images include the robot's surrounding environment and changes on the path, which are used for subsequent dynamic object recognition. The collected image frames are screened to eliminate invalid or repeated images, and only those image frames that reflect the robot's movement process are retained. Through screening, a continuous set of motion images is obtained, and these images are used for subsequent object recognition and dynamic analysis. According to the screened image frames, static objects that do not move in the environment are identified and marked through background modeling or image difference and other technologies. These static objects are usually obstacles on the path and need to be avoided by the robot. By analyzing the motion changes in the image frames, dynamic objects such as other moving robots or pedestrians are identified and marked, which may affect the robot's path and therefore require special attention. The robot's ultrasonic sensor is used to measure the distance to static objects. When the robot approaches a static object and the distance is less than the set threshold, the system will issue an obstacle avoidance warning to remind the robot to take obstacle avoidance measures. The robot's path planning is adjusted based on the obstacle avoidance warning data. If a static object is found to be blocking the current path, the system will recalculate and adjust the robot's route to ensure that the robot can avoid obstacles and reach the target point safely.

[0041] Preferably, step S23 includes the following steps: Step S231: confirming the image timestamp of the robot motion frame image set to obtain the robot motion frame image timestamp; filtering the robot motion frame image set with adjacent timestamp images by using the robot motion frame image timestamp to obtain the first frame image of the robot motion and the second frame image of the robot motion; Step S232: Overlapping the first frame image of the robot motion and the second frame image of the robot motion to generate a robot motion frame overlap image; segmenting the robot motion frame overlap image into motion change regions to generate robot motion change regions; Step S233: randomly screening the robot motion frame image set based on the timestamp of the robot motion frame image to obtain the third frame image of the robot motion; matching the robot motion change area with the third frame image of the robot motion to generate a matching area of ​​the third frame image of the robot motion; Step S234: performing regional movement position calculation on the robot motion change region and the robot motion third frame image matching region to obtain the change region movement position data; comparing the change region movement position data with a preset change movement position threshold, and when the change region movement position data is greater than or equal to the preset change movement position threshold, marking the region corresponding to the robot motion change region as robot motion dynamic object recognition data; Step S235: When the change area movement position data is less than the preset change movement position threshold, the area corresponding to the robot motion change area is marked as robot motion dynamic still object recognition data.

[0042] In an embodiment of the present invention, the robot captures continuous motion images through its built-in visual camera, and the system timestamps each frame of the image. The timestamp indicates the specific moment when the image is captured, which helps to analyze the time difference between the images later. According to the timestamp information, two adjacent frames of images, namely the first frame image and the second frame image, are selected from the robot motion frame image set. The purpose of the screening is to ensure that the time difference between the two frames of images is small and can reflect the slight changes in the robot motion. The selected first frame image and the second frame image of the robot motion are overlapped and compared, and the two images are usually aligned using an image registration algorithm (such as SIFT, SURF, ORB, etc.). The overlapped image can more clearly show the motion changes of the robot between the two frames of images. By analyzing the difference after the images are overlapped, the motion change area is segmented. The change area is usually caused by the robot motion or the change of objects in the environment, and these change areas contain information about dynamic objects. In the robot motion frame image set, random screening is performed based on the image timestamp, and the third frame image is selected. The purpose of selecting the third frame image is to increase the diversity of samples and reduce the deviation of dynamic object recognition. The robot motion change area is matched with the third frame image. Through feature matching algorithms (such as ORB, KLT tracking, etc.), find the area corresponding to the change area in the third frame image, that is, the matching area of ​​the robot motion change area. Calculate the moving position of the change area in the third frame image. Usually, the moving distance of the change area is obtained by calculating the displacement of the feature points in the image. This calculation can reflect the position change of the change area during the robot's movement. Compare the calculated change area moving position data with the preset change moving position threshold. If the displacement of the change area is greater than or equal to the preset threshold, it means that the area belongs to a dynamic object, otherwise the area is considered to be a static object. When the moving position of the change area is greater than or equal to the preset threshold, it means that the area belongs to a dynamic object, such as a moving person, other robots or obstacles. At this time, the area is marked as a dynamic object, and the corresponding robot motion dynamic object recognition data is generated. If the displacement of the change area is less than the preset threshold, the area is considered to be a static object, such as surrounding buildings, fixed facilities, etc. Mark the area as a static object, and generate the corresponding robot motion dynamic static object recognition data.

[0043] Preferably, step S24 includes the following steps: Step S241: collecting ultrasonic echo signals of the robot's moving static object recognition data according to the robot's built-in ultrasonic wave to generate robot echo signal data; performing time difference calculation processing on the robot echo signal data to generate distance information data; Step S242: Performing intensity analysis on the distance information data to generate signal reflection intensity data; performing feature clustering on the signal reflection intensity data to generate obstacle feature data; generating obstacle avoidance warning signals for the robot motion static object recognition data based on the obstacle feature data to obtain static object obstacle avoidance warning data; Step S243: adjusting the path planning of the robot's initial operation path planning data through the static object obstacle avoidance warning data to generate robot operation path planning adjustment data.

[0044] In an embodiment of the present invention, the robot uses a built-in ultrasonic sensor to sense the environment of the identified static object. The ultrasonic transmitting module transmits an ultrasonic signal to the target object, and collects the ultrasonic echo signal through the receiving module to generate corresponding echo signal data. The echo signal data contains information such as the time, amplitude, and frequency generated by the ultrasonic wave reflected by the surface of the object during the transmission process. According to the time difference between ultrasonic emission and reception, the distance between the object and the robot is calculated by the formula D=(v·Δt)÷2. D is the target distance; v is the propagation speed of ultrasonic waves in the air; Δt is the time difference between ultrasonic emission and reception. The calculated distance information is recorded as distance information data for subsequent obstacle feature analysis. The echo signal data is subjected to intensity analysis to extract the signal intensity information of ultrasonic reflection and generate signal reflection intensity data. The signal reflection intensity can reflect the material properties and shape of the surface of the object, such as smoothness and hardness. A clustering algorithm (such as K-means or density clustering algorithm) is used to perform feature classification on the signal reflection intensity data, extract features such as the size, shape, and material of the obstacle, and generate obstacle object feature data. The result of feature clustering can distinguish different types of static obstacles (such as walls, pillars or furniture) to provide support for obstacle avoidance strategies. Based on the feature data of obstacle objects and combined with the robot motion static object recognition data, the distribution position and characteristics of obstacles are analyzed. Obstacle avoidance warning signals are generated through obstacle avoidance strategies to generate static object obstacle avoidance warning data. The warning data contains the location, distance and characteristics of obstacles, providing input for path planning adjustment. According to the static object obstacle avoidance warning data, potential collision risk areas in the current robot's initial operation path planning data are identified. Local adjustments are made to the path segments that collide with obstacles, and obstacle avoidance algorithms (such as DWA dynamic window algorithm) are used to recalculate alternative paths. Under the premise of maintaining task efficiency and safety, paths with fewer obstacles or paths where obstacle characteristics have the least impact on robot actions are given priority, and new robot operation path planning adjustment data is generated. The adjusted path data is fed back to the robot's motion control module in real time to guide the robot's subsequent actual movement.

[0045] Preferably, step S243 includes the following steps: Step S2431: segmenting the initial operation path planning data of the robot to generate path segment data; extracting key nodes from the path segment data to generate node feature data; performing feasibility evaluation on the node feature data to generate path evaluation data; Step S2432: Calculate the safety distance of the static object obstacle avoidance warning data according to the path evaluation data to generate safety boundary data; perform path constraints on the safety boundary data to generate constraint condition data; generate alternative paths for the constraint condition data to generate path plan data; Step S2433: perform path cost evaluation on the path plan data to generate path cost evaluation data; select the optimal path for the path plan data based on the path cost evaluation data to generate optimized path data; use the optimized path data to adjust the path planning of the robot's initial operation path planning data to generate robot operation path planning adjustment data.

[0046] In an embodiment of the present invention, the initial operation path planning data of the robot is segmented according to the length and turning points of the path to generate multiple path segment data. Key nodes are extracted in each path segment, including the starting point, the end point and the important turning point, and the geometric features of the nodes (such as coordinate position, direction change) are recorded and the node feature data is generated. The node feature data is analyzed using an environment-based feasibility assessment model to assess whether the node has obstacle threats or inaccessible areas, and path assessment data is generated. The assessment criteria include the spatial safety of the node from the obstacle, the path smoothness, and whether the robot's motion ability meets the requirements. According to the path assessment data, combined with the static object obstacle avoidance warning data, the safe distance between each node in the path and the obstacle is calculated to generate safety boundary data. The calculation of the safety boundary is based on the size, speed and environmental characteristics of the robot to ensure that the robot will not collide with obstacles during operation. Using the safety boundary data, constraints are set for path planning to generate constraint data. Constraints include boundary areas to avoid obstacles and spatial ranges that limit the direction of robot movement. Under the premise of satisfying the constraints, multiple alternative path plans are generated based on the graph search algorithm (such as the Dijkstra algorithm), and the spatial position, number of turns, path length and other information of the path are recorded to form the path plan data. The path plan data is evaluated for cost, and the advantages and disadvantages of each path are evaluated using the comprehensive path cost function: C=α·L+β·T+γ·S; C: comprehensive path cost; L: path length; T: number of path turns; S: path safety (distance from obstacles); α, β, γ: weight coefficients used to adjust the priority of different factors. According to the result of the cost function, the path cost evaluation data is generated. The path with the lowest cost is selected from the path plan data to generate the optimized path data. The robot's initial operation path planning data is adjusted using the optimized path data to generate the final robot operation path planning adjustment data. The adjusted path has higher safety and operation efficiency, and avoids the interference of static obstacles.

[0047] As an example of the present invention, refer to Figure 3 As shown, in this example, step S3 includes: Step S31: performing dynamic object motion analysis on the robot motion dynamic object recognition data to obtain dynamic object motion analysis data; performing object motion orientation extraction on the dynamic object motion analysis data to generate dynamic object motion direction data; Step S32: reconstructing the dynamic object motion trajectory of the dynamic object motion analysis data according to the dynamic object motion direction data to generate a dynamic object motion reconstructed trajectory; Step S33: performing trajectory intersection judgment on the robot operation path planning adjustment data and the dynamic object motion reconstructed trajectory. When the robot operation path planning adjustment data and the dynamic object motion reconstructed trajectory intersect, the plane difference collision distance between the robot operation path planning adjustment data and the dynamic object motion reconstructed trajectory is calculated to obtain the plane difference collision distance; Step S34: Perform dynamic obstacle avoidance warning based on the plane difference collision distance to generate initial dynamic object obstacle avoidance warning data; segment the plane difference collision distance to generate collision distance segmentation data; adjust the warning volume of the initial dynamic object obstacle avoidance warning data through the collision distance segmentation data to generate dynamic object obstacle avoidance warning data.

[0048] In an embodiment of the present invention, the robot motion dynamic object recognition data is analyzed by using a target tracking algorithm (such as Kalman filtering, optical flow method or deep learning target detection model), and the motion parameters of the dynamic object, including speed, acceleration and position change, are extracted to generate dynamic object motion analysis data. The dynamic object motion analysis data includes the motion characteristics of the time series, forming a feature data set describing the motion state of the object. According to the dynamic object motion analysis data, the motion trajectory of the dynamic object is calculated, and its motion direction is extracted to generate dynamic object motion direction data. The motion direction is represented by an angle (such as an angle with the ground coordinate system) or a vector form for further trajectory reconstruction. According to the dynamic object motion direction data, the dynamic object motion analysis data is reconstructed to generate a continuous motion path, that is, a dynamic object motion reconstructed trajectory. The discrete motion data of the dynamic object is processed continuously by using a trajectory interpolation or fitting algorithm based on a time series (such as spline interpolation, Bézier curve fitting) to ensure that the trajectory is smooth and consistent with the actual motion. The dynamic object motion reconstructed trajectory is stored in the form of a path curve, including a spatial coordinate sequence and a timestamp, which can intuitively represent the moving trajectory of the dynamic object in the environment. The robot's path planning adjustment data and the dynamic object's motion reconstructed trajectory are spatially intersected to check whether the two paths have an intersection. Intersection determination can be achieved through geometric methods (such as line segment intersection test) or point cloud-based spatial interaction detection. When the robot's path intersects the dynamic object's trajectory, the shortest distance between the two trajectories is calculated, that is, the plane difference collision distance. The plane difference collision distance is expressed by the following formula: ; D: plane difference collision distance; Pr(t): position point in the robot path planning adjustment data; Pd(t): position point in the dynamic object motion reconstructed trajectory; : Euclidean distance. Based on the plane difference collision distance, the collision risk level is judged and the initial dynamic object obstacle avoidance warning data is generated. When the collision distance is lower than the set safety threshold, the dynamic obstacle avoidance warning is triggered. The plane difference collision distance is segmented into multiple risk levels (such as low risk, medium risk, and high risk) to generate collision distance segmentation data. According to the collision distance segmentation data, the warning volume of the initial dynamic object obstacle avoidance warning data is adjusted. The volume adjustment rules are as follows: high risk (extremely close distance): the highest volume; medium risk (close distance): medium volume; low risk (far distance): low volume. Combined with the warning signal after the volume adjustment, the dynamic object obstacle avoidance warning data is finally generated to remind the robot to avoid obstacles in real time.

[0049] Preferably, reconstructing the motion trajectory of the dynamic object based on the dynamic object motion analysis data according to the dynamic object motion direction data includes: Perform time series segmentation on the dynamic object motion direction data to generate motion direction change points; perform dynamic object continuous position estimation on the dynamic object motion analysis data through the motion direction change points to generate dynamic object preliminary position trajectory data; The preliminary position trajectory data of the dynamic object is converted into three-dimensional trajectory to generate the three-dimensional data of the dynamic object position trajectory; the three-dimensional data of the dynamic object position trajectory is bound to the trajectory time sequence to generate the spatiotemporal motion trajectory of the dynamic object; the spatiotemporal motion trajectory of the dynamic object is classified and labeled to generate the dynamic object motion reconstruction trajectory.

[0050] In an embodiment of the present invention, the motion direction data of the dynamic object is analyzed according to the time series, and the significant change points of the motion direction are identified to generate the motion direction change point data. The sliding window detection or statistical change detection algorithm (such as CUSUM method, Kalman filter) is used to analyze the change trend of the motion direction, and the time node of the direction change is extracted. The motion direction change point is used to continuously estimate the position of the dynamic object motion analysis data to generate the preliminary position trajectory data of the dynamic object. The estimation method is based on the motion direction vector and time series interpolation. The interpolation algorithm (such as linear interpolation and spline interpolation) is applied to estimate the motion position of the dynamic object between the direction change points to ensure the continuity and smoothness of the trajectory. The preliminary position trajectory data of the dynamic object is promoted to a three-dimensional space trajectory to generate the three-dimensional data of the dynamic object position trajectory. The realization of three-dimensionalization needs to be combined with the actual height or depth information of the dynamic object. Use a depth sensor (such as a lidar or an RGB-D camera) to collect depth data and map the two-dimensional plane trajectory to a three-dimensional space; for the missing depth information, the trajectory height fitting algorithm or the assumed height model is used to complete it. According to the time series of the three-dimensional data of the position trajectory of dynamic objects, an accurate timestamp is added to each trajectory point to ensure the consistency and integrity of the trajectory data in the spatial and temporal dimensions. The spatiotemporal motion trajectory of dynamic objects is analyzed to identify their motion modes (such as linear motion, curved motion, uniform motion, etc.), and they are classified and labeled based on the characteristics of the motion trajectory, and finally the dynamic object motion reconstruction trajectory is generated. Classification algorithms (such as KNN, decision trees or deep learning models) are used to classify the motion patterns of trajectory data; the classification results are labeled in combination with domain knowledge (such as classification as "pedestrian trajectory" and "vehicle trajectory").

[0051] In this specification, an obstacle avoidance warning system for a robot is provided, which is used to execute the above-mentioned obstacle avoidance warning method for a robot. The obstacle avoidance warning system for a robot includes: The motion mode classification module is used to obtain robot motion data and moving target point data; classify the robot motion mode of the robot motion data to generate the robot motion mode and the robot static mode; according to the static mode, the robot's built-in laser radar is used to detect the environment of the moving target point data to obtain standard static environment detection data; The static risk avoidance warning module is used to plan the path according to the standard static environment detection data and generate the robot's initial operation path planning data; based on the robot's motion mode, the robot's built-in visual camera is used to perform dynamic and static object recognition on the robot's initial operation path planning data to generate robot motion dynamic object recognition data and robot motion static object recognition data; static obstacle avoidance warning is performed on the robot's motion static object recognition data to generate static object obstacle avoidance warning data; path planning adjustment is performed on the robot's initial operation path planning data through the static object obstacle avoidance warning data to generate robot operation path planning adjustment data; The dynamic risk avoidance warning module is used to reconstruct the dynamic object motion trajectory of the robot motion dynamic object recognition data based on the robot motion path planning adjustment data to generate the dynamic object motion reconstruction trajectory; perform trajectory intersection distance analysis on the robot motion path planning adjustment data and the dynamic object motion reconstruction trajectory to obtain the plane difference collision distance; perform dynamic obstacle avoidance warning based on the plane difference collision distance to generate dynamic object obstacle avoidance warning data; The warning encapsulation module is used to encapsulate the static object obstacle avoidance warning data and the dynamic object obstacle avoidance warning data into an obstacle avoidance warning module to generate a robot obstacle avoidance warning module to perform the robot's obstacle avoidance warning operation.

[0052] The beneficial effect of the present invention is that by classifying the robot motion data, it is distinguished whether the robot is in motion mode or static mode, and accurate state information is provided for subsequent path planning and environmental perception. By using laser radar to detect the environment in static mode, standard static environment data can be generated, which provides a basis for the safe navigation of the robot in a static state and avoids collisions with static obstacles. The standard static environment detection data is used for path planning, and the dynamic and static objects are identified in combination with the robot motion mode, which improves the robot's obstacle avoidance ability in complex environments. Through the obstacle avoidance warning of static objects, the robot can adjust the path in real time to avoid static obstacles, ensuring the safety and reliability of the driving path. Based on the reconstruction of the path planning adjustment data and the motion trajectory of the dynamic object, the dynamic risk avoidance warning module can predict the future position of the dynamic obstacle, and evaluate the potential collision risk through the trajectory intersection distance analysis. The module can calculate the plane difference collision distance in real time in a dynamic environment, generate the dynamic object obstacle avoidance warning in time, enhance the robot's adaptability and reaction speed in a dynamic environment, and avoid collisions with dynamic obstacles. The obstacle avoidance warning data of static and dynamic objects are modularly packaged to form a unified obstacle avoidance warning system, which improves the overall coordination and operability of the system. This module ensures that the robot can consider both static and dynamic obstacles, and performs obstacle avoidance tasks through centralized warning, optimizes the robot's obstacle avoidance strategy, and improves the efficiency and accuracy of the obstacle avoidance process. Therefore, the present invention improves the accuracy and real-time performance of the robot's obstacle avoidance warning in complex environments by accurately distinguishing static and dynamic objects, reconstructing motion trajectories, and adjusting path planning.

[0053] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0054] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. An obstacle avoidance warning method for a robot, characterized in that: The following steps are involved: Step S1: obtaining robot motion data and moving target point data; classifying the robot motion data into robot motion modes to generate robot motion modes and robot static modes; performing environmental detection on the moving target point data using the robot's built-in laser radar according to the static mode to obtain standard static environmental detection data; Step S2: Plan the path according to the standard static environment detection data to generate the robot's initial operation path planning data; based on the robot's motion mode, use the robot's built-in visual camera to perform dynamic and static object recognition on the robot's initial operation path planning data to generate robot motion dynamic object recognition data and robot motion static object recognition data; Perform static obstacle avoidance warning on the robot motion static object recognition data to generate static object obstacle avoidance warning data; The path planning data of the robot's initial operation path planning is adjusted through the static object obstacle avoidance warning data to generate the robot's operation path planning adjustment data; Step S3: reconstructing the dynamic object motion trajectory of the robot motion dynamic object recognition data based on the robot motion path planning adjustment data to generate a dynamic object motion reconstructed trajectory; performing trajectory intersection distance analysis on the robot motion path planning adjustment data and the dynamic object motion reconstructed trajectory to obtain a plane difference collision distance; Perform dynamic obstacle avoidance warning based on plane difference collision distance and generate dynamic object obstacle avoidance warning data; Step S4: encapsulate the static object obstacle avoidance warning data and the dynamic object obstacle avoidance warning data into an obstacle avoidance warning module to generate a robot obstacle avoidance warning module to perform the robot's obstacle avoidance warning operation.

2. The obstacle avoidance warning method for a robot according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: Acquire robot motion data and moving target point data; Step S12: extracting motion speed information from the robot motion data to obtain robot motion speed information data; performing non-zero discrimination on the robot motion speed information data, and when the robot motion speed information data is non-zero, marking the robot in the first mode based on the robot motion speed information data to generate a robot motion mode; Step S13: when the robot motion speed information data is zero, the robot is marked as in the second mode based on the robot motion speed information data to generate a static mode of the robot; according to the static mode, the robot's built-in laser radar is used to detect the environment of the moving target point data to obtain static environment detection data; Step S14: performing data preprocessing on the static environment detection data to generate standard static environment detection data, wherein the data preprocessing includes data cleaning, data denoising, data missing value filling and data standardization.

3. The obstacle avoidance warning method for a robot according to claim 2, characterized in that: According to the static mode, the robot's built-in laser radar is used to detect the moving target point data, including: According to the static mode, the robot's built-in laser radar is used to collect environmental point cloud data of the moving target point data to obtain environmental point cloud data; the environmental point cloud data is roughly aligned to generate environmental point cloud rough alignment data; Perform point cloud fine registration on the rough registration data of the environmental point cloud to generate environmental point cloud fine registration data; perform multi-frame point cloud fusion on the environmental point cloud fine registration data to generate environmental global point cloud fusion data; Perform point cloud segmentation on the global point cloud fusion data of the environment to generate an environmental point cloud segmentation object; perform point cloud meshing on the global point cloud fusion data of the environment according to the environmental point cloud segmentation object to generate environmental point cloud mesh data; perform environmental object category labeling on the environmental point cloud mesh data to generate static environment detection data.

4. The obstacle avoidance warning method for a robot according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: performing target distance plane calculation on the moving target point data according to the standard static environment detection data to obtain the target distance plane distance; performing path planning on the moving target point data according to the target distance plane distance to generate the robot's initial operation path planning data; Step S22: Based on the robot motion mode, using the robot's built-in visual camera to collect motion images of the robot's initial running path planning data to obtain the robot's motion image; performing continuous frame image screening on the robot's motion image to obtain a robot motion frame image set; Step S23: performing dynamic and static object recognition on the robot motion frame image set to generate robot motion dynamic object recognition data and robot motion static object recognition data; Step S24: Perform static obstacle avoidance warning on the robot's moving static object recognition data according to the robot's built-in ultrasonic wave, and generate static object obstacle avoidance warning data; perform path planning adjustment on the robot's initial operation path planning data through the static object obstacle avoidance warning data, and generate robot operation path planning adjustment data.

5. The obstacle avoidance warning method for a robot according to claim 4, characterized in that: Step S23 includes the following steps: Step S231: confirming the image timestamp of the robot motion frame image set to obtain the robot motion frame image timestamp; filtering the robot motion frame image set with adjacent timestamp images by using the robot motion frame image timestamp to obtain the first frame image of the robot motion and the second frame image of the robot motion; Step S232: Overlapping the first frame image of the robot motion and the second frame image of the robot motion to generate a robot motion frame overlap image; segmenting the robot motion frame overlap image into motion change regions to generate robot motion change regions; Step S233: randomly screening the robot motion frame image set based on the timestamp of the robot motion frame image to obtain the third frame image of the robot motion; matching the robot motion change area with the third frame image of the robot motion to generate a matching area of ​​the third frame image of the robot motion; Step S234: performing regional movement position calculation on the robot motion change region and the robot motion third frame image matching region to obtain the change region movement position data; comparing the change region movement position data with a preset change movement position threshold, and when the change region movement position data is greater than or equal to the preset change movement position threshold, marking the region corresponding to the robot motion change region as robot motion dynamic object recognition data; Step S235: When the change area movement position data is less than the preset change movement position threshold, the area corresponding to the robot motion change area is marked as robot motion dynamic still object recognition data.

6. The obstacle avoidance warning method for a robot according to claim 4, characterized in that: Step S24 includes the following steps: Step S241: collecting ultrasonic echo signals of the robot's moving static object recognition data according to the robot's built-in ultrasonic wave to generate robot echo signal data; performing time difference calculation processing on the robot echo signal data to generate distance information data; Step S242: Performing intensity analysis on the distance information data to generate signal reflection intensity data; performing feature clustering on the signal reflection intensity data to generate obstacle feature data; generating obstacle avoidance warning signals for the robot motion static object recognition data based on the obstacle feature data to obtain static object obstacle avoidance warning data; Step S243: adjusting the initial operation path planning data of the robot by using the static object obstacle avoidance warning data to generate the robot operation path planning adjustment data.

7. The obstacle avoidance warning method for a robot according to claim 6, characterized in that: Step S243 includes the following steps: Step S2431: segmenting the initial operation path planning data of the robot to generate path segment data; extracting key nodes from the path segment data to generate node feature data; performing feasibility evaluation on the node feature data to generate path evaluation data; Step S2432: Calculate the safety distance of the static object obstacle avoidance warning data according to the path evaluation data to generate safety boundary data; perform path constraints on the safety boundary data to generate constraint condition data; generate alternative paths for the constraint condition data to generate path plan data; Step S2433: perform path cost evaluation on the path plan data to generate path cost evaluation data; select the optimal path for the path plan data based on the path cost evaluation data to generate optimized path data; use the optimized path data to adjust the path planning of the robot's initial operation path planning data to generate robot operation path planning adjustment data.

8. The obstacle avoidance warning method for a robot according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: performing dynamic object motion analysis on the robot motion dynamic object recognition data based on the robot operation path planning adjustment data to obtain dynamic object motion analysis data; performing object motion orientation extraction on the dynamic object motion analysis data to generate dynamic object motion direction data; Step S32: reconstructing the dynamic object motion trajectory of the dynamic object motion analysis data according to the dynamic object motion direction data to generate a dynamic object motion reconstructed trajectory; Step S33: performing trajectory intersection judgment on the robot operation path planning adjustment data and the dynamic object motion reconstructed trajectory. When the robot operation path planning adjustment data and the dynamic object motion reconstructed trajectory intersect, the plane difference collision distance between the robot operation path planning adjustment data and the dynamic object motion reconstructed trajectory is calculated to obtain the plane difference collision distance; Step S34: Perform dynamic obstacle avoidance warning based on the plane difference collision distance to generate initial dynamic object obstacle avoidance warning data; segment the plane difference collision distance to generate collision distance segmentation data; adjust the warning volume of the initial dynamic object obstacle avoidance warning data through the collision distance segmentation data to generate dynamic object obstacle avoidance warning data.

9. The obstacle avoidance warning method for a robot according to claim 8, characterized in that: Reconstructing the motion trajectory of a dynamic object based on the motion analysis data of the dynamic object according to the motion direction data of the dynamic object includes: Perform time series segmentation on the dynamic object motion direction data to generate motion direction change points; perform dynamic object continuous position estimation on the dynamic object motion analysis data through the motion direction change points to generate dynamic object preliminary position trajectory data; The preliminary position trajectory data of the dynamic object is converted into three-dimensional trajectory to generate the three-dimensional data of the dynamic object position trajectory; the three-dimensional data of the dynamic object position trajectory is bound to the trajectory time sequence to generate the spatiotemporal motion trajectory of the dynamic object; the spatiotemporal motion trajectory of the dynamic object is classified and labeled to generate the dynamic object motion reconstruction trajectory.

10. An obstacle avoidance warning system for a robot, characterized in that: Used to execute the obstacle avoidance warning method for a robot as claimed in claim 1, the obstacle avoidance warning system for a robot comprises: The motion mode classification module is used to obtain robot motion data and moving target point data; classify the robot motion mode of the robot motion data to generate the robot motion mode and the robot static mode; according to the static mode, the robot's built-in laser radar is used to detect the environment of the moving target point data to obtain standard static environment detection data; The static risk avoidance warning module is used to plan the path according to the standard static environment detection data and generate the robot's initial operation path planning data; based on the robot's motion mode, the robot's built-in visual camera is used to perform dynamic and static object recognition on the robot's initial operation path planning data to generate robot motion dynamic object recognition data and robot motion static object recognition data; static obstacle avoidance warning is performed on the robot's motion static object recognition data to generate static object obstacle avoidance warning data; path planning adjustment is performed on the robot's initial operation path planning data through the static object obstacle avoidance warning data to generate robot operation path planning adjustment data; The dynamic risk avoidance warning module is used to reconstruct the dynamic object motion trajectory of the robot motion dynamic object recognition data based on the robot motion path planning adjustment data to generate the dynamic object motion reconstruction trajectory; perform trajectory intersection distance analysis on the robot motion path planning adjustment data and the dynamic object motion reconstruction trajectory to obtain the plane difference collision distance; perform dynamic obstacle avoidance warning based on the plane difference collision distance to generate dynamic object obstacle avoidance warning data; The warning encapsulation module is used to encapsulate the static object obstacle avoidance warning data and the dynamic object obstacle avoidance warning data into an obstacle avoidance warning module to generate a robot obstacle avoidance warning module to perform the robot's obstacle avoidance warning operation.

Images