Intelligent path planning and action collaboration method and system for robot
By constructing a three-dimensional dynamic environment model and implementing closed-loop cooperative control, the problem of insufficient robot path planning and motion coordination was solved, thereby improving obstacle avoidance safety and task execution stability in dynamic environments.
Patent Information
- Application Number
- CN202511656236.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-12
- Publication Date
- 2026-02-10
AI Technical Summary
Existing technologies suffer from insufficient robot path planning and motion coordination in complex scenarios, poor adaptability to dynamic environments, weak proactive dynamic obstacle avoidance, and weak path and motion coordination, resulting in high collision risk, high task interruption rate, and difficulty in meeting industrial-grade reliability requirements.
By constructing a three-dimensional dynamic environment model, combining multimodal sensor data and weighted Kalman filter fusion, feasible paths are generated and safety margins are optimized to achieve closed-loop collaborative control, accurately extract key nodes and generate coherent action sequences, and correct deviations in real time.
It significantly improves the robot's obstacle avoidance safety and path movement coordination efficiency in dynamic environments, reduces collision risks, and enhances robustness and task execution stability in complex scenarios.
Smart Images
Figure CN121500967A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of robotics technology, and in particular to an intelligent path planning and motion coordination method and system for robots. Background Technology
[0002] In complex scenarios such as warehousing and logistics, industrial manufacturing, and special rescue, the autonomous navigation of mobile robots relies on a collaborative process of "environmental perception - path planning - action execution." Currently, existing technologies in this field suffer from the following prominent problems:
[0003] (1) Insufficient dynamic environment adaptation: Path planning algorithms are mostly based on static environment assumptions or fixed rules. Even if some dynamic planning algorithms are involved, they can only passively respond to environmental changes and do not fully integrate key state parameters such as the type of dynamic target and real-time speed. This results in poor dynamic obstacle avoidance initiative and a high risk of collision in multi-target interaction scenarios. At the same time, the safety margin mostly adopts a preset fixed value and cannot be adaptively adjusted according to the speed difference of dynamic targets. Path redundancy occurs in low-speed scenarios and safety hazards are prominent in high-speed scenarios.
[0004] (2) Weak coordination between path and action: Path generation and robot joint action adaptation are often serial and independent processes. There is a lack of accurate extraction of key nodes of the path and insufficient matching of action parameters. Deviation correction in the execution stage depends on fixed threshold triggering. In complex dynamic scenarios, "false triggering" or "missed triggering" is likely to occur, resulting in poor action continuity and high task interruption rate.
[0005] These issues cause existing robots to deviate significantly from their intended trajectory under dynamic disturbances, making it difficult to meet the industrial-grade reliability requirements of scenarios such as warehouse AGV cluster operations and rubble rescue. Summary of the Invention
[0006] The purpose of this application is to provide an intelligent path planning and action coordination method and system for robots. Through three-dimensional dynamic environment modeling, safety margin optimization and closed-loop collaborative control, the dynamic obstacle avoidance safety, path action coordination efficiency and robustness in complex scenarios of robots are significantly improved.
[0007] To achieve the above objectives, this application provides an intelligent path planning and action coordination method for robots, comprising the following steps: S1: Receiving multimodal sensor data and constructing a three-dimensional dynamic environment model based on the multimodal sensor data; wherein, the multimodal sensor data includes at least: lidar data, vision sensor data, and infrared sensor data; the three-dimensional dynamic environment model contains three types of core information: static obstacles, dynamic targets, and temporary constraints; S2: Inputting the three-dimensional dynamic environment model into a preset dynamic environment response algorithm, which outputs a global preliminary path; introducing a safety margin to the global preliminary path. S3: Optimize local segments of the path to generate a feasible path; S4: Extract key nodes in the feasible path, call the robot joint motion database to match and adapt the actions, and generate a continuous action sequence through the inverse kinematics algorithm; S5: Compare the real-time sensor data collected by the robot during the execution of the continuous action sequence with the feasible path. If the deviation between the real-time sensor data and the feasible path exceeds the preset deviation threshold, the path fine-tuning and action parameter correction are triggered to achieve closed-loop coordination. If the deviation between the real-time sensor data and the feasible path does not exceed the preset deviation threshold, the robot executes the continuous action sequence normally until the end.
[0008] As described above, the sub-steps for constructing a 3D dynamic environment model based on multimodal sensor data are as follows: S11: Preprocess the multimodal sensor data to obtain preprocessed state data, which includes: 3D coordinates of static obstacles, contour data of static obstacles, environmental semantic data with category labels, and thermal location data of dynamic targets; S12: Perform weighted Kalman filter fusion on the preprocessed state data to obtain... S13: Based on the preprocessed state data, and the estimated environmental state after time-mapping; The environmental state estimate after time-mapping is used to construct a three-dimensional dynamic environment model containing three core types of information: static obstacles, dynamic targets, and temporary constraints, through dynamic target trajectory prediction and multi-dimensional environmental information integration.
[0009] As described above, the sub-steps for preprocessing multimodal sensor data to obtain preprocessed state data are as follows: S111: Perform noise filtering on the lidar data in the multimodal sensor data to obtain the three-dimensional coordinates and contour data of the static obstacle; S112: Perform edge smoothing on the visual sensor data in the multimodal sensor data to obtain environmental semantic data with category labels; S113: Perform temperature threshold calibration on the infrared sensor data in the multimodal sensor data, filter out environmental background thermal signals below the preset dynamic target thermal threshold, and obtain the thermal position data of the dynamic target; S114: Use the three-dimensional coordinates of the static obstacle, the contour data of the static obstacle, the environmental semantic data with category labels, and the thermal position data of the dynamic target as the preprocessed state data.
[0010] As described above, the sub-steps for noise filtering of the lidar data in the multimodal sensor data to obtain the three-dimensional coordinates and contour data of the static obstacle are as follows: S1111: Use Gaussian filtering to remove pulse noise from the original distance data in the lidar data to obtain clean distance data; S1112: Combine the clean distance data with the laser beam emission angle in the lidar data, generate three-dimensional point cloud data through spatial coordinate conversion, and use the three-dimensional point cloud data as the three-dimensional coordinates of the static obstacle; S1113: Based on the three-dimensional point cloud data, reconstruct the three-dimensional contour of the static obstacle in the environment, clarify the spatial size and position of the static obstacle, and use the spatial size and position of the static obstacle as the contour data of the static obstacle.
[0011] As mentioned above, among them, The expression for the estimated environmental state after time-mapping is: ; in, for The estimated environmental state after time-mapping; These are the weighting coefficients for the lidar data; This was obtained based on the three-dimensional coordinates and contour data of the static obstacle. The estimated state of the lidar after time-preprocessing; These are the weighting coefficients for the visual sensor data; Obtained from environmental semantic data with category labels Visual semantic state estimation value after time-mapping; These are the weighting coefficients for the infrared sensor data; To obtain from the thermal location data of dynamic targets Preprocessed infrared thermal state estimate at time step; for Kalman gain matrix at time step; for Actual observation data from the time sensor; for The time-observation matrix is used to map the state estimates to the observation space; for Prior state estimate at time step.
[0012] As mentioned above, based on the preprocessed state data, and The sub-steps for constructing a three-dimensional dynamic environment model containing three core types of information—static obstacles, dynamic targets, and temporary constraints—by integrating the estimated environmental state values after time-mapping and through dynamic target trajectory prediction and multi-dimensional environmental information integration are as follows: S131: Obtain from historical data The first two frames of the time-fusiond environmental state estimate are used, and the first two frames of data are compared with... The three frames of data, including the time-fusion environmental state estimate, are extracted to obtain the real-time position and velocity of the dynamic target in each frame. Based on the real-time position and velocity of the dynamic target in each frame, a predicted trajectory of the dynamic target within a preset time period is generated. The first two frames are: Time-fusion environmental state estimates and S132: Based on the three-dimensional coordinates of static obstacles in the preprocessed state data, the predicted trajectory of dynamic targets within a preset time period, and... The core data from the time-series fusion-based environmental state estimates are used to generate a multi-source information integration package; among which... The core data in the environmental state estimate after time-mapping includes: the fused size parameters of static obstacles, the attribute labels of static obstacles, and the attribute labels of dynamic targets; S133: Using the multi-source information integration package as input, a temporary avoidance area that changes with the speed of the dynamic target is generated as a temporary constraint; S134: The multi-source information integration package and the predicted trajectory of the dynamic target within a preset time period are spatially correlated and layer-fused with the generated temporary constraints to construct a three-dimensional dynamic environment model containing three types of core information: static obstacles, dynamic targets, and temporary constraints.
[0013] As mentioned above, multi-sensor fusion data analysis technology is used to analyze the data from the first two frames and... The environmental state estimate after time-lapse fusion is extracted from these three frames of data to obtain the real-time position and velocity of the dynamic target in each frame.
[0014] As shown above, the preset dynamic environment response algorithm is the existing D algorithm, the existing RRT algorithm, or the improved A* algorithm; the expression for the improved A* algorithm is: ; in, For path points in the global preliminary path The overall assessment cost; The path points from the preset initial position to the global preliminary path for the robot to perform path planning tasks. The actual cost of moving; For path points in the global preliminary path The actual cost of moving to the preset destination of the robot's path planning task; For dynamic target risk weight coefficients; For path points in the global preliminary path Dynamic collision risk value; This refers to the predicted trajectory of a dynamic target in a three-dimensional dynamic environment model.
[0015] As mentioned above, the safety margin can be either the existing safety margin or the dynamic safety margin; the expression for the dynamic safety margin is: ;in, For dynamic safety margin; This is the speed influence coefficient; The real-time speed of the dynamic target; The basic safety distance is the minimum safe distance that the robot path must maintain between the robot and the dynamic target when the target is stationary.
[0016] This application also provides an intelligent path planning and motion coordination system for robots, comprising: a robot body, a sensor subsystem disposed on the robot body, and an intelligent path planning and motion coordination subsystem disposed on the robot body; wherein, the sensor subsystem is used to collect multimodal sensor data and send the multimodal sensor data to the intelligent path planning and motion coordination subsystem; the intelligent path planning and motion coordination subsystem is used to execute the above-mentioned intelligent path planning and motion coordination method for robots.
[0017] The beneficial effects achieved by this application are as follows: (1) The intelligent path planning and action coordination method and system for robots in this application can significantly improve the dynamic environment adaptability and obstacle avoidance safety of robot paths. Specifically, by constructing a three-dimensional dynamic environment model that includes static obstacles, dynamic targets and temporary constraints, the key state parameters such as the type and real-time speed of dynamic targets are accurately captured. Then, combined with dynamic safety margin, the global preliminary path is optimized, so that path planning changes from passively responding to environmental changes to actively adapting to the state of dynamic targets, which greatly reduces the collision risk in multi-target interaction scenarios.
[0018] (2) The intelligent path planning and action coordination method and system for robots in this application can significantly optimize the coordination efficiency of robot path and action. Specifically, it can accurately extract key nodes from feasible paths, call the robot joint motion database to match and adapt actions, and then generate a continuous action sequence through the inverse kinematics algorithm to achieve deep coordination between path planning and joint action, improve the smoothness and energy efficiency of robot movement, and optimize task execution efficiency.
[0019] (3) The intelligent path planning and action coordination method and system for robots in this application can significantly enhance the robustness of robots in complex dynamic scenarios. Specifically, during the execution of the robot's action sequence, the real-time data of the sensors is compared with the feasible path. If the deviation exceeds the limit, the path fine-tuning and action parameter correction are triggered. Relying on the closed-loop feedback and deviation correction mechanism, dynamic working condition interference can be effectively dealt with, and the stability of task execution and environmental adaptability can be improved. Attached Figure Description
[0020] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings.
[0021] Figure 1 A schematic diagram of one embodiment of an intelligent path planning and motion coordination system for robots; Figure 2 This is a flowchart of one embodiment of an intelligent path planning and motion coordination method for robots. Detailed Implementation
[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0023] like Figure 1 As shown, this application provides an intelligent path planning and motion coordination system for robots, including: a robot body, a sensor subsystem disposed on the robot body, and an intelligent path planning and motion coordination subsystem disposed on the robot body.
[0024] The sensor subsystem is used to collect multimodal sensor data and send the multimodal sensor data to the intelligent path planning and action coordination subsystem.
[0025] Intelligent Path Planning and Motion Coordination Subsystem: This subsystem executes the intelligent path planning and motion coordination methods for robots, enabling full-process functionality from 3D dynamic environment modeling and path planning optimization to motion sequence generation and closed-loop collaborative control.
[0026] Furthermore, the sensor subsystem includes at least: lidar, vision sensor, infrared sensor, and data transmission unit.
[0027] Among them, the lidar is used to acquire lidar data and send it to the data transmission unit.
[0028] Visual sensor: Used to acquire visual sensor data and send it to the data transmission unit.
[0029] Infrared sensor: Used to acquire infrared sensor data and send it to the data transmission unit.
[0030] Data transmission unit: transmits LiDAR data, visual sensor data, and infrared sensor data as multimodal sensor data to the intelligent path planning and motion coordination subsystem.
[0031] Furthermore, the coordinate system used in the intelligent path planning and motion coordination system for robots in this application is the right-handed Cartesian coordinate system commonly used in the field of robotics. Specifically, the center of the robot base is the origin (point O), the X-axis extends along the robot's forward direction, the Y-axis is perpendicular to the X-axis and points to the left of the robot, and the Z-axis is perpendicular to the XY plane and points upward (i.e., vertical direction). Spatial parameters such as the velocity (v) of the dynamic target and the distance parameter of the safety margin in the intelligent path planning and motion coordination system for robots are all calculated based on this coordinate system, which can ensure the uniformity of parameter dimensions.
[0032] like Figure 2 As shown, this application provides an intelligent path planning and action coordination method for robots, including the following steps: S1: Receive multimodal sensor data and construct a three-dimensional dynamic environment model based on the multimodal sensor data; wherein, the multimodal sensor data includes at least: lidar data, vision sensor data and infrared sensor data; the three-dimensional dynamic environment model contains three core types of information: static obstacles, dynamic targets and temporary constraints.
[0033] Furthermore, the sub-steps for constructing a 3D dynamic environment model based on multimodal sensor data are as follows: S11: Preprocess the multimodal sensor data to obtain preprocessed state data, which includes: the three-dimensional coordinates of static obstacles, the contour data of static obstacles, environmental semantic data with category labels, and the thermal location data of dynamic targets.
[0034] Furthermore, the sub-steps for preprocessing the multimodal sensor data to obtain preprocessed state data are as follows: S111: Perform noise filtering on the lidar data in the multimodal sensor data to obtain the three-dimensional coordinates and contour data of the static obstacle.
[0035] Furthermore, the sub-steps for performing noise filtering on the lidar data in the multimodal sensor data to obtain the three-dimensional coordinates and contour data of the static obstacle are as follows: S1111: Gaussian filtering is used to remove impulse noise from the original distance data in the lidar data, resulting in clean distance data.
[0036] Specifically, in the raw distance data of LiDAR, impulse noise refers to abnormal distance values caused by environmental interference (such as dust, abnormal light reflection, sensor electrical noise, etc.), which manifest as abrupt data points that are seriously inconsistent with the actual static obstacle distance.
[0037] S1112: Combine the clean distance data with the laser beam emission angle in the lidar data, generate three-dimensional point cloud data through spatial coordinate conversion, and use the three-dimensional point cloud data as the three-dimensional coordinates of the static obstacle.
[0038] Specifically, 3D point cloud data consists of a massive number of (X,Y,Z) coordinate points. Step S1112 can be achieved using existing technology, so it will not be described in detail here.
[0039] S1113: Based on 3D point cloud data, reconstruct the 3D outline of static obstacles in the environment, clarify the spatial dimensions and location of static obstacles, and use the spatial dimensions and location of static obstacles as the outline data of static obstacles.
[0040] Specifically, Gaussian filtering is used to remove pulse noise from the LiDAR data, improving the accuracy of the output 3D coordinates and contour data of static obstacles. This provides a core basis for determining the location of static obstacles in 3D dynamic environment models. Static obstacles refer to objects in the robot's environment whose position, shape, and size remain fixed over a long period and do not move autonomously or change shape. They are fixed no-go zones that the robot needs to avoid first during path planning, such as machine tools and columns.
[0041] S112: Perform edge smoothing processing on the visual sensor data in the multimodal sensor data to obtain environmental semantic data with category labels.
[0042] Furthermore, the sub-steps for performing edge smoothing processing on the visual sensor data in the multimodal sensor data to obtain environmental semantic data with category labels are as follows: S1121: Use a semantic segmentation algorithm to classify the image pixels in the visual sensor data, obtain the labeled category of each image pixel, and generate an initial semantic segmentation mask based on the labeled category.
[0043] Specifically, the specific categories of semantic segmentation algorithms are set according to actual needs, such as DeepLab and Mask R-CNN. The specific content of the labeled categories is also set according to actual needs, such as categories like pedestrians, AGVs, walls, and the ground. Generating the initial semantic segmentation mask based on the labeled categories refers to the region division in the visual sensor data for different labeled categories.
[0044] S1122: Use morphological filtering or Gaussian blurring algorithm to smooth the edge regions in the initial semantic segmentation mask to obtain environmental semantic data with category labels.
[0045] Specifically, to address the jagged distortion of object edges in the initial semantic segmentation mask, morphological filtering (such as dilation-erosion operation) or Gaussian blurring algorithm is used to smooth the edge regions, making the semantic boundaries of objects more closely resemble the real shape (such as changing the pedestrian outline from jagged to a natural curve), resulting in environmental semantic data with category labels. The environmental semantic data with category labels includes the category information of each object (such as pedestrian), the smoothed spatial contour, and the pixel coordinates of the corresponding image region, which can provide key semantic information of dynamic targets for the 3D dynamic environment model.
[0046] S113: Perform temperature threshold calibration on the infrared sensor data in the multimodal sensor data, filter out the environmental background thermal signals that are lower than the preset dynamic target thermal threshold, and obtain the thermal position data of the dynamic target.
[0047] Specifically, the raw thermal data from the infrared sensor data is filtered out from environmental background thermal signals (such as low-temperature signals from the ground or fixed equipment) that are below the thermal threshold of the dynamic target, thus obtaining the thermal location data of the dynamic target. Thermal location data refers to the real-time thermal coordinates of dynamic targets such as humans and operating equipment in space, which can provide core basis for the three-dimensional dynamic environment model to clarify the movement state and position of the dynamic target.
[0048] The specific value of the dynamic target thermal threshold is set according to actual needs. In this application, the preferred value is 30-40℃.
[0049] S114: Use the three-dimensional coordinates of static obstacles, the contour data of static obstacles, the environmental semantic data with category labels, and the thermal location data of dynamic targets as preprocessed state data.
[0050] S12: Perform weighted Kalman filtering fusion on the preprocessed state data to obtain... The estimated environmental state after time-mapping.
[0051] Specifically, The environmental state estimate after time-mapping fusion contains complete environmental information such as obstacle positions, dynamic target coordinates, and semantic categories, which can provide core basis for the construction of 3D dynamic environment models and trajectory prediction of dynamic targets.
[0052] Furthermore, The expression for the estimated environmental state after time-mapping is: ; in, for The estimated environmental state after time-mapping; These are the weighting coefficients for the lidar data; This was obtained based on the three-dimensional coordinates and contour data of the static obstacle. The estimated state of the lidar after time-preprocessing; These are the weighting coefficients for the visual sensor data; Obtained from environmental semantic data with category labels Visual semantic state estimation value after time-mapping; These are the weighting coefficients for the infrared sensor data; To obtain from the thermal location data of dynamic targets Preprocessed infrared thermal state estimate at time step; for Kalman gain matrix at time step; for Actual observation data from the time sensor; for The time-observation matrix is used to map the state estimates to the observation space; for Prior state estimate at time step.
[0053] Specifically, The specific value is set according to actual needs, and the preferred value in this application is: , This is used to highlight the distance measurement accuracy advantage of lidar data. The specific value is set according to actual needs, and the preferred value in this application is: , Advantages of semantic recognition capabilities for matching visual sensor data. The specific value is set according to actual needs, and the preferred value in this application is: , Advantages of dynamic target capture adapted to infrared sensor data. Weights are used to dynamically balance predicted values and actual observed values.
[0054] The present invention The expression for the environmental state estimate after time-matter fusion is achieved by weighted fusion of preprocessed data from LiDAR, visual sensors, and infrared sensors, combined with the dynamic error correction mechanism of Kalman filtering. This integrates the performance advantages of multiple sensors and outputs an accurate environmental state estimate containing information such as the 3D coordinates and contours of static obstacles, the thermal position and motion state of dynamic targets, and environmental semantics with category labels. This provides robust environmental perception support for downstream tasks such as robot obstacle avoidance and intelligent monitoring decision-making, effectively improving the accuracy and scene adaptability of multimodal sensor data fusion.
[0055] Furthermore, ; in, For: X-dimensional data in the three-dimensional coordinates of a static obstacle, that is: the X-axis coordinate of the center point of the static obstacle (unit: m); For: Y-dimensional data in the three-dimensional coordinates of a static obstacle, that is: the Y-axis coordinate of the center point of the static obstacle (unit: m); Z-dimensional data in the three-dimensional coordinates of a static obstacle, i.e., the Z-axis coordinate of the center point of the static obstacle (unit: m, with 0 for the ground). For: the length dimension data in the contour data of static obstacles, that is: the length of static obstacles (along the X-axis direction, unit: m); For: the width dimension data in the contour data of static obstacles, that is: the width of static obstacles (along the Y-axis direction, unit: m); For: the height dimension data in the contour data of static obstacles, that is: the height of static obstacles (along the Z-axis direction, unit: m).
[0056] Furthermore, ; in, This refers to category information in environmental semantic data with category labels, i.e., object category encoding; The X-axis coordinates (in meters) of the center point of the semantic contour of an object in the spatial location of the contour in the environmental semantic data with category labels. Y-axis coordinates (in meters) of the center point of the semantic contour of an object in the spatial location of the contour in the environmental semantic data with category labels. Z-axis coordinates (in meters) of the center point of the semantic contour of an object in the spatial location of the contour in the environmental semantic data with category labels. The length of the semantic contour (along the X-axis, in meters) in the environmental semantic data with category labels. The width of the semantic contour (in meters) in the context semantic data with category labels.
[0057] Furthermore, ; in, The X-dimensional data in the thermal location data of the dynamic target is the X-axis coordinate (unit: m) of the thermal center point of the dynamic target. This refers to the Y-dimensional data in the thermal location data of a dynamic target, specifically the Y-axis coordinate (unit: m) of the thermal center point of the dynamic target. Z-dimensional data in the thermal location data of a dynamic target, namely: the Z-axis coordinate (unit: m) of the thermal center point of the dynamic target; The thermal value in the thermal location data of a dynamic target, i.e., the thermal intensity of the dynamic target (unit: °C); The velocity of the dynamic target along the X-axis (unit: m / s, positive value is the positive X-axis direction) is derived from the thermal position changes in the thermal position data of the dynamic target. The velocity of the dynamic target along the Y-axis (unit: m / s, positive value is the positive Y-axis direction) is derived from the thermal position changes in the thermal position data of the dynamic target.
[0058] Furthermore, ; in, The first distance value of a static obstacle measured by lidar, in meters; The second distance value for a static obstacle measured by lidar, in meters; The probability of a visual sensor recognizing a specified target; The value is the dynamic target thermal value measured by an infrared sensor, in °C.
[0059] Specifically, the specified target can be a static obstacle or a dynamic target, and the specific object of the specified target is set according to the monitoring requirements of the current scene.
[0060] Furthermore, Based on Time-fusion environmental state estimate Predicted by environmental patterns (e.g., uniform motion of a dynamic target). Current state vector dimension and The vector dimensions are consistent. As an initial reference for integration.
[0061] Furthermore, for The observation matrix, that is: the n-dimensional state vector (e.g.: The dimension of ) is mapped to m dimensions (and) (Matching), the elements in the observation matrix are mapping coefficients (e.g., ... middle , Transformed using the distance formula (Predicted value).
[0062] Furthermore, for The Kalman gain matrix, where the elements are error correction coefficients, is obtained from... and" The observation error is dynamically adjusted to adjust the error correction coefficient, ensuring that the error correction amount can accurately match the dimension of the n-dimensional state vector.
[0063] S13: Based on the preprocessed state data, and The environmental state estimate after time-mapping is used to construct a three-dimensional dynamic environment model containing three core types of information: static obstacles, dynamic targets, and temporary constraints, through dynamic target trajectory prediction and multi-dimensional environmental information integration.
[0064] Furthermore, based on the preprocessed state data, and The sub-steps for constructing a three-dimensional dynamic environment model containing three core types of information—static obstacles, dynamic targets, and temporary constraints—from the time-lapse fusion of environmental state estimates, through dynamic target trajectory prediction and multi-dimensional environmental information integration, are as follows: S131: Obtain from historical data The first two frames of the time-fusiond environmental state estimate are used, and the first two frames of data are compared with... The three frames of data, including the time-fusion environmental state estimate, are extracted to obtain the real-time position and velocity of the dynamic target in each frame. Based on the real-time position and velocity of the dynamic target in each frame, a predicted trajectory of the dynamic target within a preset time period is generated. The first two frames are: Time-fusion environmental state estimates and The estimated environmental state after time-mapping.
[0065] Furthermore, multi-sensor fusion data analysis technology is used to analyze the data from the first two frames and... The environmental state estimate after time-lapse fusion is extracted from these three frames of data to obtain the real-time position and velocity of the dynamic target in each frame.
[0066] Specifically, multi-sensor fusion data analysis technology is a mature existing technology, so it will not be elaborated further.
[0067] Furthermore, the sub-steps for generating the predicted trajectory of the dynamic target within a preset time period based on the real-time position and velocity of the dynamic target in each frame of data are as follows: S1311: Calculate the velocity of the dynamic target in adjacent frame data using the velocity fluctuation ratio calculation method to obtain four sets of fluctuation data. Use a preset fluctuation threshold to judge each set of fluctuation data. If each set of fluctuation data is less than or equal to the preset fluctuation threshold, it is judged as uniform motion, and S1312 is executed; if any set of fluctuation data is greater than the preset fluctuation threshold, it is judged as variable motion, and S1313 is executed.
[0068] The four sets of fluctuation data are as follows: The first set of fluctuation data is: ; The second set of fluctuation data is as follows: ; The third set of fluctuation data is: ; The fourth set of fluctuation data is as follows: ; in, From The velocity of the dynamic target extracted from the time-fused environmental state estimate, where, for The X-axis velocity of the dynamic target at any given time. for The Y-axis velocity of the dynamic target at any given moment; From The velocity of the dynamic target extracted from the time-fused environmental state estimate, where, for The X-axis velocity of the dynamic target at any given time. for The Y-axis velocity of the dynamic target at any given moment; From The velocity of the dynamic target extracted from the time-fused environmental state estimate, where, for The X-axis velocity of the dynamic target at any given time. for The Y-axis velocity of the dynamic target at any given time.
[0069] Specifically, the preset fluctuation threshold value is set according to actual needs, and the preferred value in this application is 5%.
[0070] S1312: Using a linear model to analyze the... The real-time position and velocity of the dynamic target extracted from the environmental state estimate after time-mapping are calculated to obtain the predicted position of the dynamic target at multiple prediction times, which is used as the predicted trajectory of the dynamic target within a preset time period.
[0071] Furthermore, the first Predicted position of the dynamic target at each prediction time The expression is: ; ; in, From The X-axis coordinates of the real-time position of the dynamic target extracted from the environmental state estimate after time-mapping; From The Y-axis coordinate value of the dynamic target in the real-time position extracted from the environmental state estimate after time-mapping; for The X-axis velocity of the dynamic target at any given moment; for The Y-axis velocity of the dynamic target at any given moment; For the first Predicted time and The time interval between moments.
[0072] Specifically, the duration of the preset time period is set according to actual needs. This application preferably specifies that the preset prediction time period is within the next 3 seconds. The number of prediction moments within the preset time period is set according to actual needs. This application preferably specifies 3. The interval between adjacent prediction moments is set according to actual needs. To ensure trajectory continuity and computational efficiency, this application preferably specifies that the interval between the 3 prediction moments is 1 second, corresponding to... Predicted time intervals of 1 second, 2 seconds, and 3 seconds.
[0073] S1313: Using a quadratic function model to analyze... The real-time position and velocity of dynamic targets extracted from the environmental state estimates after time-mapping, and based on... Time and The acceleration calculated from the velocity difference at different times is used to obtain the predicted position at multiple predicted times, which serves as the predicted trajectory of the dynamic target within the predicted time period.
[0074] Furthermore, the first Predicted position of the dynamic target at each prediction time The expression is: ; ; ; ; in, From The X-axis coordinates of the real-time position of the dynamic target extracted from the environmental state estimate after time-mapping; From The Y-axis coordinate value of the dynamic target in the real-time position extracted from the environmental state estimate after time-mapping; From The velocity of the dynamic target extracted from the time-fused environmental state estimate, where, for The X-axis velocity of the dynamic target at any given time. for The Y-axis velocity of the dynamic target at any given moment; From The velocity of the dynamic target extracted from the time-fused environmental state estimate, where, for The X-axis velocity of the dynamic target at any given time. for The Y-axis velocity of the dynamic target at any given moment; For the first Predicted time and The interval between moments; The acceleration along the X-axis of the dynamic target; The acceleration along the Y-axis of the dynamic target; for Time and The time interval between moments.
[0075] Specifically, the duration of the preset time period is set according to actual needs. This application preferably specifies that the preset prediction time period is within the next 3 seconds. The number of prediction moments within the preset time period is set according to actual needs. This application preferably specifies 3. The interval between adjacent prediction moments is set according to actual needs. To ensure trajectory continuity and computational efficiency, this application preferably specifies that the interval between the 3 prediction moments is 1 second, corresponding to... Predicted time intervals of 1 second, 2 seconds, and 3 seconds.
[0076] S132: Based on the three-dimensional coordinates of static obstacles in the preprocessed state data, the trajectory of dynamic targets within a preset time period is predicted, and... The core data from the time-series fusion-based environmental state estimates are used to generate a multi-source information integration package; among which... The core data in the time-fusion environmental state estimate include: fused size parameters of static obstacles, attribute labels of static obstacles, and attribute labels of dynamic targets.
[0077] Furthermore, based on the three-dimensional coordinates of static obstacles in the preprocessed state data, the predicted trajectory of dynamic targets within a preset time period, and... The sub-steps for generating a multi-source information integration package from the two types of core data in the time-fusion environmental state estimate are as follows: S1321: Retrieve the three-dimensional coordinates of static obstacles and the predicted trajectory of dynamic targets within a preset time period from the preprocessed state data, and The core data in the environmental state estimate after time-mapping.
[0078] S1322: Using the lidar coordinate system as a reference, a coordinate transformation matrix is used to map the three-dimensional coordinates of static obstacles in the preprocessed state data and the predicted trajectory of dynamic targets within a preset time period to the same coordinate system to obtain unified coordinates.
[0079] Specifically, the coordinate transformation matrix is a mature existing technology in the field of sensor data fusion, so it will not be elaborated further.
[0080] S1323: Unified coordinate and dimension parameter packaging based on spatial coordinates and attribute labels. The core data from the time-series fusion environmental state estimates are used to obtain a multi-source information integration package.
[0081] Specifically, step S1323 can be achieved using existing technology, so it will not be described in detail here.
[0082] S133: Using the multi-source information integration package as input, generate a temporary avoidance zone that changes with the dynamic target speed as a temporary constraint.
[0083] Furthermore, the sub-step of generating a temporary avoidance zone that varies with the dynamic target velocity as a temporary constraint, using the multi-source information integration package as input, is as follows: S1331: Retrieve from multi-source information integration package The dynamic target's X-axis velocity at time t and The Y-axis velocity of the dynamic target at time t, and the maximum value is used for calculation. The dynamic target's X-axis velocity at time t and The Y-axis velocity of the dynamic target at any given time is calculated to obtain the fastest velocity of the dynamic target.
[0084] Furthermore, the expression for the fastest speed of a dynamic target is: ; in, for The X-axis velocity of the dynamic target at any given time. for The Y-axis velocity of the dynamic target at any given time.
[0085] S1332: Analyze the fastest speed of a dynamic target using a preset speed scene speed threshold. If the fastest speed of the dynamic target is less than or equal to the speed scene speed threshold, the speed scene is determined to be a low-speed scene; if the fastest speed of the dynamic target is greater than the speed scene speed threshold, the speed scene is determined to be a high-speed scene.
[0086] Specifically, the specific value of the speed threshold for the speed scenario is set according to the requirements, and the preferred value in this application is 0.5 m / s.
[0087] S1333: Determine the constraint radius based on the speed scenario. When the speed scenario is low speed, the constraint radius is: When the speed scenario is high-speed, the constraint radius is: ; in, This serves as the basic safety radius for low-speed scenarios. This refers to the radius coefficient in high-speed scenarios. The fastest speed of a dynamic target; for The X-axis velocity of the dynamic target at any given time. for The Y-axis velocity of the dynamic target at any given time.
[0088] Specifically, and For a real number greater than zero, the preferred value in this application is: , .
[0089] S1334: Retrieve the dynamic target prediction trajectory within the prediction time period from the multi-source information integration package. Using each prediction position on the dynamic target prediction trajectory within the prediction time period as the current center and the constraint radius as the current radius, generate a circular temporary avoidance area for each prediction position. All circular temporary avoidance areas together constitute a temporary constraint.
[0090] S134: Integrate the multi-source information package and the predicted trajectory of the dynamic target within a preset time period, and spatially associate and fuse them with the generated temporary constraints to construct a three-dimensional dynamic environment model containing three core types of information: static obstacles, dynamic targets, and temporary constraints.
[0091] Specifically, step S134 can be implemented using existing technologies, including: 3D modeling engines (such as scene building modules of Unity and Unreal Engine), spatial data fusion tools (such as Python's Open3D library and C++'s PCL point cloud library), and layer rendering technologies (such as WebGL's 3D scene rendering framework). No additional new solutions need to be designed, so they will not be elaborated further.
[0092] S2: Input the 3D dynamic environment model into the preset dynamic environment response algorithm, and the dynamic environment response algorithm outputs the global preliminary path; introduce a safety margin to optimize the local road segments of the global preliminary path to generate a feasible path.
[0093] Specifically, as an example, the preset dynamic environment response algorithm is the existing D algorithm or the existing RRT algorithm, but is not limited to the existing D (Dynamic Window Approach) algorithm or the existing RRT (Rapidly-exploring Random Trees) algorithm.
[0094] As another embodiment, the preset dynamic environment response algorithm is an improved D algorithm or an improved A* algorithm;
[0095] The expression for the improved D algorithm is: ;
[0096] in, For path points in the global preliminary path The overall cost; For path points in the global preliminary path The original basic cost; For dynamic target trajectory weighting coefficients; For path points in the global preliminary path arrive The distance; It is the predicted trajectory of dynamic targets in a three-dimensional dynamic environment model (such as the coordinate sequence of the future movement path of pedestrians and AGVs).
[0097] Specifically, Used to adjust the degree of impact of dynamic objectives on path costs. The specific value is set according to actual needs, and the preferred value in this application is 0.8~1.2.
[0098] The improved D algorithm in this application can enhance the adaptability to dynamic environments, reduce the cost of replanning and make the path smoother, and improve the ability to avoid dynamic targets in advance by 50%, making it more suitable for real-time path planning in dynamic scenarios.
[0099] The expression for the improved A* (Dynamic A*) algorithm is as follows: ; in, For path points in the global preliminary path The overall assessment cost; The path points in the global preliminary path are the preset initial position (i.e., the starting point) for the robot to perform path planning tasks. The actual cost of moving; For path points in the global preliminary path The actual cost of moving to the preset destination of the robot's path planning task; For dynamic target risk weight coefficients; For path points in the global preliminary path The dynamic collision risk value is the probability that a path point n in the global preliminary path will collide with the future trajectory of the dynamic target. The closer the target is to the trajectory, the higher the risk value. It is the predicted trajectory of dynamic targets in a three-dimensional dynamic environment model (such as the coordinate sequence of the future movement path of pedestrians and AGVs).
[0100] Specifically, Used to adjust the degree of influence of dynamic targets on the path. The specific value is set according to actual needs, and the preferred value in this application is 0.8~1.2.
[0101] The improved A* algorithm in this application incorporates a dynamic target prediction trajectory risk term, which enables more proactive dynamic obstacle avoidance, higher path planning efficiency, and flexible parameter adjustment. Compared with traditional algorithms, it reduces dynamic collision risk by more than 40% and significantly improves adaptability.
[0102] Furthermore, as an example, the safety margin is an existing safety margin, which is a preset fixed value parameter.
[0103] Furthermore, as another embodiment, the safety margin is a dynamic safety margin; wherein, the expression for the dynamic safety margin is: ; in, For dynamic safety margin; This is the speed influence coefficient; The real-time speed of the dynamic target; The basic safety distance is the minimum safe distance that the robot path must maintain between the robot and the dynamic target when the target is stationary.
[0104] Specifically, These parameters are preset based on the types of dynamic targets (such as pedestrians, AGVs, etc.) in the 3D dynamic environment model, and are used to reflect the influence weight of the speed of different types of dynamic targets on the safe distance. It comes directly from the perception of the dynamic target state by the three-dimensional dynamic environment model.
[0105] The dynamic safety margin of this application can achieve dynamic adaptation. Compared with the existing fixed safety margin parameters, it can accurately adjust the safety distance according to the type and real-time speed of the dynamic target, improve the collision avoidance accuracy of dynamic targets by more than 40%, and avoid over-obstacle avoidance, thus significantly optimizing the path planning efficiency and making it more suitable for the real-time path planning needs of complex dynamic scenarios.
[0106] S3: Extract key nodes in the feasible path, call the robot joint motion database to match and adapt the action, and generate a coherent action sequence through the inverse kinematics algorithm.
[0107] Specifically, in step S3, existing methods such as curvature analysis and equal-interval sampling can be used to extract key nodes of feasible paths to accurately identify turning points and feature points of the path; when calling the robot joint motion database, the key nodes are matched and adapted based on the pre-stored joint motion templates; the inverse kinematics algorithm adopts the existing mature algorithm of the Newton-Raphson method, but is not limited to the existing mature algorithm of the Newton-Raphson method, to calculate the angles, velocities and other parameters that each joint needs to execute, and finally generate a coherent motion sequence.
[0108] S4: Compare the real-time sensor data collected by the robot during the execution of a continuous sequence of actions with the feasible path. If the deviation between the real-time sensor data and the feasible path exceeds a preset deviation threshold, path fine-tuning and motion parameter correction are triggered to achieve closed-loop coordination. If the deviation between the real-time sensor data and the feasible path does not exceed the preset deviation threshold, the robot will execute the continuous sequence of actions normally until the end.
[0109] Specifically, when the robot executes a continuous sequence of actions, it uses existing sensors such as LiDAR, vision sensors, and joint encoders to collect data on its own pose and motion state in real time, and compares this data with feasible paths. If the deviation exceeds a preset deviation threshold, path fine-tuning uses the existing local path planning technique of Dynamic Window (DWA) to generate a locally corrected path. Motion parameter correction is achieved by recalculating parameters such as joint angles using forward and inverse kinematics algorithms (such as numerical iteration methods), while combining cubic spline interpolation techniques to smooth the motion sequence. If the deviation does not exceed the preset deviation threshold, the robot executes the continuous sequence of actions normally until the end.
[0110] The beneficial effects achieved by this application are as follows: (1) The intelligent path planning and action coordination method and system for robots in this application can significantly improve the dynamic environment adaptability and obstacle avoidance safety of robot paths. Specifically, by constructing a three-dimensional dynamic environment model that includes static obstacles, dynamic targets and temporary constraints, the key state parameters such as the type and real-time speed of dynamic targets are accurately captured. Then, combined with dynamic safety margin, the global preliminary path is optimized, so that path planning changes from passively responding to environmental changes to actively adapting to the state of dynamic targets, which greatly reduces the collision risk in multi-target interaction scenarios.
[0111] (2) The intelligent path planning and action coordination method and system for robots in this application can significantly optimize the coordination efficiency of robot path and action. Specifically, it can accurately extract key nodes from feasible paths, call the robot joint motion database to match and adapt actions, and then generate a continuous action sequence through the inverse kinematics algorithm to achieve deep coordination between path planning and joint action, improve the smoothness and energy efficiency of robot movement, and optimize task execution efficiency.
[0112] (3) The intelligent path planning and action coordination method and system for robots in this application can significantly enhance the robustness of robots in complex dynamic scenarios. Specifically, during the execution of the robot's action sequence, the real-time data of the sensors is compared with the feasible path. If the deviation exceeds the limit, the path fine-tuning and action parameter correction are triggered. Relying on the closed-loop feedback and deviation correction mechanism, dynamic working condition interference can be effectively dealt with, and the stability of task execution and environmental adaptability can be improved.
[0113] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the scope of protection of this application is intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application. Obviously, those skilled in the art can make various alterations and variations to this application without departing from the spirit and scope of this application. Thus, if these modifications and variations of this application fall within the scope of protection of this application and its equivalents, this application also intends to include these modifications and variations.
Claims
1. A method for intelligent path planning and motion coordination for robots, characterized in that, Includes the following steps: S1: Receive multimodal sensor data and construct a three-dimensional dynamic environment model based on the multimodal sensor data; wherein, the multimodal sensor data includes at least: lidar data, vision sensor data and infrared sensor data; the three-dimensional dynamic environment model contains three core types of information: static obstacles, dynamic targets and temporary constraints; S2: Input the 3D dynamic environment model into the preset dynamic environment response algorithm, and the dynamic environment response algorithm outputs the global preliminary path; introduce a safety margin to optimize the local road segments of the global preliminary path to generate a feasible path; S3: Extract key nodes in feasible paths, call the robot joint motion database to match and adapt actions, and generate a coherent action sequence through inverse kinematics algorithm; S4: Compare the real-time sensor data collected by the robot during the execution of a continuous sequence of actions with the feasible path. If the deviation between the real-time sensor data and the feasible path exceeds a preset deviation threshold, path fine-tuning and motion parameter correction are triggered to achieve closed-loop coordination. If the deviation between the real-time sensor data and the feasible path does not exceed the preset deviation threshold, the robot will execute the continuous sequence of actions normally until the end.
2. The intelligent path planning and motion coordination method for robots according to claim 1, characterized in that, The sub-steps for constructing a 3D dynamic environment model based on multimodal sensor data are as follows: S11: Preprocess the multimodal sensor data to obtain preprocessed state data, which includes: the three-dimensional coordinates of static obstacles, the contour data of static obstacles, environmental semantic data with category labels, and the thermal location data of dynamic targets. S12: Perform weighted Kalman filtering fusion on the preprocessed state data to obtain... The estimated environmental state after time-mapping; S13: Based on the preprocessed state data, and The environmental state estimate after time-mapping is used to construct a three-dimensional dynamic environment model containing three core types of information: static obstacles, dynamic targets, and temporary constraints, through dynamic target trajectory prediction and multi-dimensional environmental information integration.
3. The intelligent path planning and motion coordination method for robots according to claim 2, characterized in that, The sub-steps for preprocessing multimodal sensor data to obtain preprocessed state data are as follows: S111: Perform noise filtering on the lidar data in the multimodal sensor data to obtain the three-dimensional coordinates and contour data of the static obstacle; S112: Perform edge smoothing processing on the visual sensor data in the multimodal sensor data to obtain environmental semantic data with category labels; S113: Perform temperature threshold calibration on the infrared sensor data in the multimodal sensor data, filter out the environmental background thermal signals that are lower than the preset dynamic target thermal threshold, and obtain the thermal position data of the dynamic target; S114: Use the three-dimensional coordinates of static obstacles, the contour data of static obstacles, the environmental semantic data with category labels, and the thermal location data of dynamic targets as preprocessed state data.
4. The intelligent path planning and motion coordination method for robots according to claim 3, characterized in that, The sub-steps for performing noise filtering on the lidar data in the multimodal sensor data to obtain the three-dimensional coordinates and contour data of the static obstacle are as follows: S1111: Gaussian filtering is used to remove impulse noise from the original distance data in the lidar data to obtain clean distance data; S1112: Combine the distance data of the clean area with the laser beam emission angle in the lidar data, generate three-dimensional point cloud data through spatial coordinate conversion, and use the three-dimensional point cloud data as the three-dimensional coordinates of the static obstacle; S1113: Based on 3D point cloud data, reconstruct the 3D outline of static obstacles in the environment, clarify the spatial dimensions and location of static obstacles, and use the spatial dimensions and location of static obstacles as the outline data of static obstacles.
5. The intelligent path planning and motion coordination method for robots according to claim 2, characterized in that, The expression for the estimated environmental state after time-mapping is: ； in, for The estimated environmental state after time-mapping; These are the weighting coefficients for the lidar data; This was obtained based on the three-dimensional coordinates and contour data of the static obstacle. The estimated state of the lidar after time-preprocessing; These are the weighting coefficients for the visual sensor data; Obtained from environmental semantic data with category labels Visual semantic state estimation value after time-mapping; These are the weighting coefficients for the infrared sensor data; To obtain from the thermal location data of dynamic targets Preprocessed infrared thermal state estimate at time step; for Kalman gain matrix at time step; for Actual observation data from the time sensor; for The time-observation matrix is used to map the state estimates to the observation space; for Prior state estimate at time step.
6. The intelligent path planning and motion coordination method for robots according to claim 2, characterized in that, Based on the preprocessed state data, and The sub-steps for constructing a three-dimensional dynamic environment model containing three core types of information—static obstacles, dynamic targets, and temporary constraints—from the time-lapse fusion of environmental state estimates, through dynamic target trajectory prediction and multi-dimensional environmental information integration, are as follows: S131: Obtain from historical data The first two frames of the time-fusiond environmental state estimate are used, and the first two frames of data are compared with... The three frames of data, including the time-fusion environmental state estimate, are extracted to obtain the real-time position and velocity of the dynamic target in each frame. Based on the real-time position and velocity of the dynamic target in each frame, a predicted trajectory of the dynamic target within a preset time period is generated. The first two frames are: Time-fusion environmental state estimates and The estimated environmental state after time-mapping; S132: Based on the three-dimensional coordinates of static obstacles in the preprocessed state data, the trajectory of dynamic targets within a preset time period is predicted, and... The core data from the time-series fusion-based environmental state estimates are used to generate a multi-source information integration package; among which... The core data in the time-fusion environmental state estimate include: fused size parameters of static obstacles, attribute labels of static obstacles, and attribute labels of dynamic targets; S133: Using the multi-source information integration package as input, generate a temporary avoidance zone that changes with the dynamic target speed as a temporary constraint; S134: Integrate the multi-source information package and the predicted trajectory of the dynamic target within a preset time period, and spatially associate and fuse them with the generated temporary constraints to construct a three-dimensional dynamic environment model containing three core types of information: static obstacles, dynamic targets, and temporary constraints.
7. The intelligent path planning and motion coordination method for robots according to claim 6, characterized in that, Multi-sensor fusion data analysis technology is used to analyze the data from the first two frames and The environmental state estimate after time-lapse fusion is extracted from these three frames of data to obtain the real-time position and velocity of the dynamic target in each frame.
8. The intelligent path planning and motion coordination method for robots according to claim 1, characterized in that, The preset dynamic environment response algorithm is the existing D algorithm, the existing RRT algorithm, or the improved A* algorithm; The expression for the improved A* algorithm is as follows: ； in, For path points in the global preliminary path The overall assessment cost; The path points from the preset initial position to the global preliminary path for the robot to perform path planning tasks. The actual cost of moving; For path points in the global preliminary path The actual cost of moving to the preset destination of the robot's path planning task; For dynamic target risk weight coefficients; For path points in the global preliminary path Dynamic collision risk value; This refers to the predicted trajectory of a dynamic target in a three-dimensional dynamic environment model.
9. The intelligent path planning and motion coordination method for robots according to claim 1, characterized in that, The safety margin can be either the existing safety margin or a dynamic safety margin; The expression for dynamic safety margin is: ； in, For dynamic safety margin; This is the speed influence coefficient; The real-time speed of the dynamic target; The basic safety distance is the minimum safe distance that the robot path must maintain between the robot and the dynamic target when the target is stationary.
10. An intelligent path planning and motion coordination system for robots, characterized in that, include: The robot body, the sensor subsystem mounted on the robot body, and the intelligent path planning and motion coordination subsystem mounted on the robot body; Among them, the sensor subsystem is used to collect multimodal sensor data and send the multimodal sensor data to the intelligent path planning and action coordination subsystem; Intelligent path planning and motion coordination subsystem: used to execute the intelligent path planning and motion coordination method for robots as described in any one of claims 1-9.