A multi-robot cooperative control system

Abstract

The application relates to the technical field of intelligent control, in particular to a multi-robot cooperation control system which comprises an end point density analysis module, a weight updating module, a posture evaluation module, a speed correction module and a candidate path scheduling module. The application constructs a fixed grid area based on the spatial distribution of candidate path end points and quantifies the end point density, can comprehensively perceive the scheduling concentration trend of robots in a target area, can effectively improve the utilization rate of sparse end points in task allocation by expanding the boundaries of grids with low density and introducing weight distribution, can further optimize the balance of resource allocation, can refine the dynamic complexity evaluation in the path moving process by dynamically identifying the posture stability level of path segments in combination with the posture sequence change trend, can correct the basic moving speed parameters by stability level mapping, and further introduces the corrected speed information and spatial conflict judgment in path scheduling scoring, which is helpful for realizing the comprehensive optimal ordering of candidate paths.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control technology, and in particular to a multi-robot collaborative control system. Background Art

[0002] The field of intelligent control technology mainly involves the coordinated action of computational models, perception mechanisms and actuators to achieve autonomous perception, decision-making and operational control of complex dynamic systems.

[0003] Among them, the traditional multi-robot collaborative control system refers to the use of communication mechanisms and distributed control strategies to enable multiple robots to work together in the process of completing specific tasks, coordinating candidate path planning, resource allocation, task execution and other behaviors.

[0004] In the process of multi-robot collaborative control, existing technologies rely on distributed control and communication mechanisms to complete path coordination and task allocation, and lack detailed modeling of the spatial density of candidate path endpoints, which can easily lead to repeated scheduling in task-concentrated areas and idle resources in sparse areas, resulting in uneven scheduling load. When the robot's posture changes drastically in the candidate path, existing technologies have not established a linkage control between posture stability and movement speed, making it difficult to adapt to the safety requirements during dynamic operation. In addition, the combined effect of endpoint spatial conflict and path execution difficulty is not fully considered during path selection, which may lead to path scheduling score deviations, resulting in delays, conflicts, or even failure of the selected path during execution. For example, when multiple robots converge at the endpoint of the same area at high frequency, it is easy to cause spatial conflicts and task blockages, thereby affecting the overall operating efficiency and safety performance of the system. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a multi-robot collaborative control system.

[0006] In order to achieve the above objectives, the present invention adopts the following technical solution: a multi-robot collaborative control system, the system comprising:

[0007] The endpoint density analysis module obtains the coordinate set of all candidate path endpoints in the robot's current scheduling cycle, constructs the endpoint sampling boundary based on the endpoint coordinate set and divides the fixed grid area, calculates the local density of each grid area, and obtains the endpoint area density set;

[0008] A weight updating module is configured to expand the grid area boundary of the endpoint position in the endpoint area density set that is less than the local density threshold and assign corresponding weights to obtain an endpoint weight update result;

[0009] The posture evaluation module simulates a posture vector sequence in the process of moving the multi-robot along the candidate path, detects a posture change point in the posture vector sequence, and divides a posture stability level of each candidate path segment according to a change point trend.

[0010] The speed correction module corrects a robot basic moving speed parameter planned for each candidate path segment based on the posture stability level, and generates a moving speed correction result.

[0011] The application improves that the terminal point area density set comprises a grid density value and a grid division structure, the terminal point weight update result comprises a terminal point weight and a boundary expansion area, the posture stability level comprises a posture change point and a candidate path segment stability level, and the moving speed correction result comprises a corrected moving speed parameter and a posture level corresponding relationship.

[0012] The application improves that the terminal point density analysis module comprises:

[0013] The terminal point coordinate acquisition submodule acquires a terminal point coordinate set of all candidate paths in a current scheduling period of the robot, performs two-dimensional space mapping processing on a terminal point position in each candidate path, and generates a terminal point coordinate distribution data set.

[0014] The boundary grid construction submodule establishes a fixed boundary area around each terminal point according to the terminal point coordinate distribution data set, divides the boundary area into a regular grid structure with a consistent number, and generates terminal boundary grid structure information.

[0015] The local density extraction submodule calls the terminal boundary grid structure information, counts the number of terminal points in each grid area and calculates a local density according to an area, marks the density of all grids, and obtains a terminal point area density set.

[0016] The application improves that the weight update module comprises:

[0017] The density threshold determination submodule calls the terminal point area density set, compares the density corresponding to each terminal point in the candidate path with a local density threshold, extracts a terminal point position smaller than the local density threshold, and generates a terminal density threshold matching list.

[0018] The boundary area expansion submodule expands the original boundary area corresponding to each terminal point in multiple directions according to the terminal density threshold matching list, re-divides the expanded area into a standard grid structure, counts the number of terminal points in the expanded area, calculates an average density in combination with an area, and generates a terminal expanded area density data.

[0019] The weight value allocation submodule calls the density data of the terminal extension area, calculates the difference between the average density of the extension area and the average density of the original boundary area, performs scheduling weight value update on the terminal where the difference exceeds the density improvement judgment threshold, and obtains the terminal weight update result.

[0020] The present invention is improved in that the posture assessment module includes:

[0021] The posture sequence construction submodule simulates the dynamic posture behavior of multiple robots moving along the candidate path, collects the pitch angle, roll angle and center of gravity coordinate values ​​of each robot in continuous motion on each candidate path segment, and constructs the posture change vector in chronological order to generate a multi-robot posture vector sequence set;

[0022] A change point extraction submodule detects the change trend in the posture sequence of each candidate path segment using a cumulative sum control chart algorithm based on the multi-robot posture vector sequence set, calculates the cumulative offset of each group of posture values ​​relative to the mean, and identifies the position of the change points that appear in the continuous offset to obtain posture change point trend data;

[0023] The grade interval matching submodule calls the posture change point trend data, extracts the joint features of the posture change amplitude and duration in the candidate path segment, performs interval matching classification on the joint features, and obtains the posture stability grade.

[0024] The present invention is improved in that the speed correction module includes:

[0025] The basic speed extraction submodule obtains the basic movement speed parameters of the robot set in the planning stage for each candidate path segment, structures them according to the candidate path segment number, and generates a candidate path basic speed dataset;

[0026] a stability level identification submodule, which calls the posture stability level, maps the stability level of each candidate path segment to the corresponding candidate path segment number, and forms a candidate path segment stability level index table;

[0027] The speed value adjustment submodule combines the candidate path basic speed data set and the candidate path segment stability level index table to correct the speed parameters of the candidate path segments whose stability levels are within the set adjustment range to obtain a moving speed correction result.

[0028] The present invention is improved in that it further includes a candidate path scheduling module, which calculates a score for each candidate path based on the endpoint weight update result and the movement speed correction result in combination with multi-robot task allocation information, selects a target candidate path according to the score, and generates a candidate path task scheduling instruction;

[0029] The candidate path task scheduling instruction includes a candidate path scoring result and a target candidate path.

[0030] The present invention is improved in that the candidate path scheduling module includes:

[0031] The candidate path scoring construction submodule calls the endpoint weight update result and the moving speed correction result, combines the endpoint scheduling weight and the corresponding speed correction parameter in each candidate path, and generates a candidate path basic scoring data set;

[0032] The candidate path conflict identification submodule determines the execution space distance between the candidate path endpoint and the endpoints of other robots' assigned candidate paths based on the candidate path basic scoring dataset and the multi-robot task allocation information in the current scheduling cycle, and adjusts the scores of candidate paths that conflict with the endpoint positions to obtain candidate path conflict adjustment scoring results.

[0033] The scheduling instruction generation submodule extracts the priority candidate path number of the candidate path conflict adjustment score result, matches the corresponding candidate path task identifier with the robot execution sequence, and generates the candidate path task scheduling instruction.

[0034] Compared with the prior art, the advantages and positive effects of the present invention are:

[0035] In the present invention, a fixed grid area is constructed based on the spatial distribution of the candidate path endpoints and the endpoint density is quantified, which can fully perceive the scheduling concentration trend of robots in the target area. By expanding the grid boundaries with low density and introducing weight distribution, the utilization rate of sparse endpoints in task allocation can be effectively improved, thereby optimizing the balance of resource allocation. The posture stability level of the path segment is dynamically identified in combination with the posture sequence change trend, which can refine the dynamic complexity evaluation during the path movement process. The basic moving speed parameters are corrected by stability level mapping, which can enhance the safety adaptation to the posture change sensitive segments during the path execution process. The corrected speed information and spatial conflict judgment are further introduced into the path scheduling score, which helps to achieve comprehensive optimal sorting of candidate paths, improve the rationality and execution efficiency of the overall task allocation, and ensure the accuracy of collaborative path selection and the reliability of scheduling execution of multiple robots in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 It is a system module diagram of the present invention;

[0037] Figure 2 It is a system framework diagram of the present invention;

[0038] Figure 3 Schematic diagram of the endpoint density analysis module of the present invention;

[0039] Figure 4 Schematic diagram of the weight update module of the present invention;

[0040] Figure 5 Schematic diagram of the posture assessment module of the present invention;

[0041] Figure 6 Schematic diagram of the speed correction module of the present invention;

[0042] Figure 7 Schematic diagram of the candidate path scheduling module of the present invention. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0044] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings and are only for the convenience of describing the present invention and simplifying the description. They do not indicate or imply that the devices or elements referred to must have a specific direction, be constructed and operate in a specific direction, and therefore should not be understood as limiting the present invention. In addition, in the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0045] See also Figure 1 The present invention provides a technical solution: a multi-robot collaborative control system, the system comprising:

[0046] The endpoint density analysis module obtains the coordinate set of all candidate path endpoints in the robot's current scheduling cycle, constructs the endpoint sampling boundary based on the endpoint coordinate set and divides the fixed grid area, calculates the local density of each grid area, and obtains the endpoint area density set;

[0047] The weight update module expands the grid area boundary of the endpoint position in the endpoint area density set that is less than the local density threshold and assigns the corresponding weight to obtain the endpoint weight update result;

[0048] The posture evaluation module simulates the posture vector sequence of multiple robots moving along the candidate path, detects the posture change points in the posture vector sequence, and classifies the posture stability level of each candidate path segment according to the trend of the change points;

[0049] The speed correction module corrects the basic movement speed parameters of the robot planned for each candidate path segment based on the posture stability level and generates a movement speed correction result;

[0050] The end point area density set comprises a grid density value, a grid division structure, the end point weight update result comprises an end point weight, a boundary expansion area, the attitude stability level comprises an attitude change point, a candidate path segment stability level, and the movement speed correction result comprises a corrected movement speed parameter and an attitude level corresponding relationship.

[0051] Referring to Figure 2 and Figure 3 , the end point density analysis module comprises:

[0052] An end point coordinate acquisition submodule acquires a set of end point coordinates of all candidate paths in a current scheduling period of the robot, performs two-dimensional space mapping processing on the end point position in each candidate path, and generates an end point coordinate distribution dataset.

[0053] The end point coordinate acquisition submodule acquires a set of end point coordinates of all candidate paths in a current scheduling period of the robot, performs two-dimensional space mapping processing on the end point position in each candidate path, and generates an end point coordinate distribution dataset. The end point coordinate acquisition submodule acquires a set of end point coordinates of all candidate paths in a current scheduling period of the robot, performs two-dimensional space mapping processing on the end point position in each candidate path, and generates an end point coordinate distribution dataset. represents the horizontal position of the robot in the active area, represents the longitudinal position, if there are candidate paths in the current scheduling period, the system will form an end point coordinate sequence with a length of , each element is like , and the unit is millimeter, and then the set is transmitted to a coordinate mapping function. The coordinate mapping function maps the coordinates to a standard two-dimensional space based on the actual physical dimensions of the active area, such as the width , length , and maps to , and the mapping process calculation formula is: , and are the lengths of the active area, and are the original coordinates, and are the normalized coordinates, and all mapped coordinates are combined into an end point coordinate distribution dataset.

[0054] The boundary grid construction submodule establishes a fixed boundary area around each endpoint based on the endpoint coordinate distribution dataset, divides the boundary area into a consistent number of regular grid structures, and generates the endpoint boundary grid structure information;

[0055] A fixed boundary area is established around each endpoint according to the endpoint coordinate distribution data set. During the execution process, the two-dimensional coordinate information of each endpoint in the endpoint coordinate distribution data set is first called, and the boundary area is constructed with each coordinate as the center. The area shape can be set to a square, and the side length can be set to , forming a square bounding box in the robot's activity area, for example, the end point coordinates are When the boundary area is to Repeating this step for each endpoint will form multiple boundary regions that may overlap or be densely distributed. Subsequently, the system divides each boundary region into a consistent number of regular grid structures, such as using The grid division scheme is as follows: the side length of each small grid is , forming a total of 100 sub-areas. The system executes this grid construction process at each end point. The grid numbers are marked in row-first order, for example, starting from the upper left corner, to In the specific implementation, if the number of endpoints is 20, the system will generate 20 boundary areas and their corresponding grid structures, and store all the structures uniformly as endpoint boundary grid structure information to facilitate subsequent density extraction and analysis. Taking the path management of general household or service robots as an example, this process can identify potential endpoint concentration areas, provide a fine spatial unit division basis for subsequent density calculations, and finally output uniformly numbered and organized endpoint boundary grid structure information.

[0056] The local density extraction submodule calls the endpoint boundary grid structure information, counts the number of endpoints in each grid area, calculates the local density based on the area of ​​the area, and numbers the density of all grids to obtain the endpoint area density set;

[0057] Call the endpoint boundary grid structure information, count the number of endpoints in each grid area and calculate the local density based on the area. During the execution process, the system first reads the grid unit number and its corresponding coordinate range from all the constructed endpoint boundary grid structures, assuming that each grid has a side length of The grid area is unified as The system performs a spatial attribution judgment on each grid cell to determine whether the coordinates of the end point fall within the two-dimensional space boundary of the grid. If so, the Euclidean distance between the end point and the current grid center is counted and recorded. The count value is recorded as , where the subscript Indicates the currently being processed In order to describe the spatial density of the end point within the grid and take into account the spatial offset factor, a Gaussian weight function based on distance attenuation is introduced:

[0058] ;

[0059] in, : No. The weighted local density of each grid cell, in points per square meter (points / m²), reflects the weighted density of endpoints in the grid cell. : The area of ​​a single grid, in units of , set to a fixed value here ; :To The traversal and summation operation of all endpoints in the grid, subscript Indicates the In the grid One falls into the end point; :Fall into The number of endpoints of a grid; : Natural exponential function, used to generate distance weights; : No. The Euclidean distance between the end point and the current grid center, in units of , calculated as , is the end point coordinate, is the current grid center coordinate; : Smoothing control parameter, determines the speed of change of distance attenuation, unit is , the recommended value is the upper limit of the robot navigation path positioning error.

[0060] The formula is used to calculate the weighted spatial density of endpoints in each grid cell, reflecting the degree of concentration of endpoint distribution in the region. The system first divides the entire space into two sections with a side length of The square area is uniformly set to For each grid cell, the system determines whether it contains the end point and calculates the Euclidean distance of all the end points falling into the grid relative to the current grid center. . Then the exponential decay function is introduced , the grid density contribution is calculated by distance weighting, and the closer the endpoint is, the higher the weight is. The sum of the weighted contributions of each endpoint is normalized to the density per unit area, which is the weighted density of the grid. The density measures the local aggregation of the endpoints in space and is used to reflect the distribution characteristics of the path endpoints in the local area. Controls the distance attenuation speed in meters. The set value is usually based on the upper limit of the robot's navigation positioning error and is used to constrain the impact of distant endpoints on density calculation.

[0061] A grid area For example, its center coordinates are , in millimeters, there are 3 end points in the grid, and their coordinates are 、 、 , converted to meters are 、 、 , the grid center is converted to , calculate the Euclidean distance as follows:

[0062] ;

[0063] ;

[0064] .

[0065] Bring in the weight function:

[0066] ;

[0067] ;

[0068] .

[0069] The sum of the weights is: .

[0070] The final density is: .

[0071] The system repeats the above process to process all grids, and organizes the results into triple sets according to the numbers. ,in 、 For the The center coordinates of the grid, Its density value.

[0072] See also Figure 2 and Figure 4 , the weight update module includes:

[0073] The density threshold determination submodule calls the endpoint area density set, compares the density corresponding to each endpoint in the candidate path with the local density threshold, extracts the endpoint positions that are less than the local density threshold, and generates an endpoint density threshold matching list;

[0074] The calling endpoint area density set is matched and analyzed for each endpoint in the candidate path. During the execution process, the system first traverses the endpoint numbers of all candidate paths, finds the weighted local density value of each endpoint for the grid to which it belongs, and obtains the reference density threshold from the overall endpoint area density set. The threshold is set based on the distribution of all grid densities, and the quantile value setting method can be used, such as taking the 30% quantile point of all grid density values as the local density threshold. The system sorts all density values in ascending order and determines the quantile density value as the standard threshold. For example, when the weighted density values of all grids in the system range from about 300 to 4000 points per square meter, the density corresponding to the 30% quantile may be about 1500 points per square meter. The system sequentially compares the local density of each endpoint with the threshold value. If the density value of an endpoint is less than the set threshold, the endpoint is marked as a low-density location and added to the endpoint density threshold matching list. Finally, the sub-module outputs the endpoint numbers and their corresponding coordinates of all endpoints determined to be low-density locations, forming a unified endpoint density threshold matching list.

[0075] The boundary area expansion sub-module expands the original boundary area of each endpoint in multiple directions according to the endpoint density threshold matching list, re-divides the expanded area into a standard grid structure, calculates the average density of the expanded area by counting the number of endpoints in the area, and generates endpoint expanded area density data.

[0076] According to the endpoint density threshold matching list, the area of each endpoint determined to be low-density is expanded. During the execution process, the system first reads the original boundary area definition of each endpoint, and expands the boundary length outward in eight directions based on the original square boundary. The expansion ratio can be set to 100% of the original side length, i.e., when the original side length is 30 cm, the expanded boundary length becomes 60 cm. The system re-divides the standard grid within the new expanded boundary, such as using a 10x10 grid structure, which changes the side length of each small grid to 6 cm and increases the total area to 0.36 square meters. The system repeats the endpoint drop detection and counting process for all small grids in the new expanded area, obtains the average endpoint density of the area by dividing the number of endpoints by the total area of the expanded area, and forms the endpoint expanded area density data, such as an expanded area with 12 endpoints and an average density of 33.3 points per square meter. The system records and compares the average density with the average density of the original boundary area for subsequent comparison, and finally forms the expanded area average density data structure corresponding to each low-density endpoint.

[0077] The weight value allocation sub-module calls the endpoint expanded area density data, calculates the difference between the average density of the expanded area and the average density of the original boundary area, updates the scheduling weight value for endpoints with a difference exceeding the density improvement judgment threshold, and obtains the endpoint weight update result.

[0078] The density data of the terminal extension area is called. During the execution process, the system reads the original boundary area density of each terminal and the density value of its extension area one by one, and calculates the numerical difference between the two. If the average density of the extension area is significantly higher than that of the original area, the system compares the difference with the density improvement judgment threshold. The density improvement judgment threshold can be set based on the statistical results of the overall density difference. For example, the median of the density difference of all terminal points in the sample is taken as the reference benchmark for judgment. If the density improvement difference of a terminal exceeds the reference value, the terminal is deemed to have a density improvement effect, and the system adjusts its scheduling weight value by increasing the weight value by a certain proportion. For example, if the basic weight is 1.0, if the density of the extension area is about 30 points per square meter higher than that of the original area, the system will increase the weight value to 1.3. All terminal points that have completed the weight value update are re-marked as priority scheduling path points. The system finally outputs the updated terminal number and its corresponding new weight value to form a terminal weight update result list.

[0079] See also Figure 2 and Figure 5 , the posture assessment module includes:

[0080] The posture sequence construction submodule simulates the dynamic posture behavior of multiple robots moving along the candidate path, collects the pitch angle, roll angle and center of gravity coordinate values ​​of each robot in continuous motion on each candidate path segment, and constructs the posture change vector in chronological order to generate a multi-robot posture vector sequence set;

[0081] Simulate the dynamic posture behavior of multiple robots moving along candidate paths. During the execution process, the system first reads the geometric data of the candidate path. Each path is discretized into several continuous path segments, and the adjacent connection relationship is maintained between the path segments. The system assigns a fixed step size to each path segment, such as each segment is 0.1 meters in length, and sets the initial speed and starting posture state for each robot. The simulation process adopts a time-driven mode based on trajectory tracking. The system advances the robot state with a set time step size. For example, if the time step is set to 0.1 seconds, a position state and posture state are simulated at each time point. The system constructs a rigid body dynamics model in three-dimensional space, simplifies the robot's body structure into a mixed model of mass point and rigid body, and drives the robot to complete the posture change process by introducing motion control input (such as linear velocity, angular velocity). The system sets external interference factors according to the curvature, slope and surface changes of the path segment. For example, when a small robot passes through an inclined ramp path segment, When the slope increases, the system simulates the gradual change in pitch angle as the slope increases, gradually transitioning from a horizontal posture to a pitched state. Roll jitter is also introduced to simulate speed fluctuations. If the path segment has horizontal surface unevenness, the system periodically perturbs the center of gravity coordinates up and down within the corresponding time step, simulating the vibration trajectory of the center of gravity when the wheeled chassis traverses uneven terrain. In addition, the system can set structural parameters (such as height and center of gravity distribution) for different robot types to adjust the posture change response curve. For example, if the center of gravity height of a robot is set to 0.3 meters, the system will output an increase in the instantaneous pitch angle when the slope suddenly changes during the path transition section. These changes are synchronously recorded in a time series, generating continuous time posture change data for each path segment. Each posture sample includes three parameters: pitch angle, roll angle, and center of gravity coordinates. After the system completes the complete movement simulation of multiple robots on all candidate paths, all data are organized into a multi-robot posture vector sequence set.

[0082] The change point extraction submodule uses the cumulative sum control chart algorithm to detect the change trend in the posture sequence of each candidate path segment based on the multi-robot posture vector sequence set, calculates the cumulative offset of each group of posture values ​​relative to the mean, and identifies the position of the change points that appear in the continuous offset to obtain the posture change point trend data;

[0083] According to the posture vector sequence set, analysis processing is performed, and in the execution process, the system reads the posture sequence generated by the robot on the path segment by piece, and adopts the cumulative sum control chart algorithm for trend detection analysis. The system first calculates the historical average value of the current posture value as a reference, and then gradually traverses each time data point in the vector sequence, sequentially calculates the offset difference value of the posture value at this time relative to the historical average value, and forms a cumulative offset sequence by time accumulation summation of these difference values. When the system detects that the cumulative offset continuously increases or decreases in a certain period of time and exceeds the set control range, the segment is automatically marked as a trend change interval. The position where the trend change first exceeds the control range is identified as a change point. The system synchronously stores the change point time of each path segment and the corresponding posture value to form trend change point data. For example, if the posture offset of a robot in a path segment continuously accumulates to the critical level at 4.2 seconds, the time is marked as a change point. Finally, the module outputs the trend change point data set including the path segment number, robot number, change point timestamp, and change direction identifier.

[0084] The grade interval matching sub-module calls the posture change point trend data, extracts the joint features of the posture change amplitude and duration time in the candidate path segment, performs interval matching classification on the joint features, and obtains the posture stability grade.

[0085] The system first performs time window interception on the change point position identified in each path segment during the execution process, selects the posture sequence in the fixed length time period before and after the change point, extracts the posture change amplitude and change duration time in the segment data as joint features, the change amplitude is calculated by the maximum and minimum value difference, and the duration time is the time length from the offset start to the recovery to the stable range. The system establishes a unified joint feature interval division rule, maps all change features to three grade intervals, and sets the interval division according to sample statistical data or experience rules, such as a change amplitude greater than 20 degrees and a duration time more than 2 seconds for unstable grade, a change amplitude between 10 to 20 degrees and a duration time between 1 to 2 seconds for medium grade, and a change amplitude less than 10 degrees or a duration time less than 1 second for stable grade. The system matches and compares the joint features of each path segment according to the rule, completes the grade identification process, and finally outputs the posture stability grade of each candidate path segment.

[0086] Please refer to Figure 2 and Figure 6 , the speed correction module includes:

[0087] The basic speed extraction sub-module obtains the robot basic movement speed parameter set in the planning stage for each candidate path segment, and structures and organizes the candidate path basic speed data set according to the candidate path segment number.

[0088] Obtain the robot's basic moving speed parameters set for each candidate path segment in the planning stage. During the execution process, the system first retrieves the attribute information of all candidate path segments from the path planning module and reads the basic speed setting value. The basic speed is usually preset by the planning algorithm based on path length, environmental factors, turning radius and other conditions when the path is generated. The unit is meters per second. The system traverses all path segments and extracts the corresponding speed values. Then, the speed parameters are uniformly organized and sorted according to the path segment number to ensure that the number corresponds to the speed value, forming a structured candidate path basic speed data set. For example, if there are 20 candidate path segments, the system will generate a data set of length 20, and each record contains the path segment number and its set basic moving speed.

[0089] The stability level identification submodule calls the posture stability level, maps the stability level of each candidate path segment with the corresponding candidate path segment number, and forms a candidate path segment stability level index table;

[0090] The posture stability level result is called. During the execution process, the system reads the stability level label of each candidate path segment one by one. The label is generated by the previous module through joint feature interval matching, including classification identifications such as "high stability", "medium stability", and "low stability". The system maps and binds each stability level label with its corresponding candidate path segment number, and establishes a unified index table. The index table structure is sorted by the path segment number. Each record contains the stability level identification information corresponding to the number. For example, the stability level of the path segment numbered 12 is "low stability". The system records this association in the index table. After the module completes the aggregation of the stability levels of all candidate path segments, a complete candidate path segment stability level index table is formed.

[0091] The speed value adjustment submodule combines the candidate path basic speed data set and the candidate path segment stability level index table to correct the speed parameters of the candidate path segments whose stability levels are within the set adjustment range to obtain the moving speed correction result;

[0092] Combining the basic speed dataset of candidate paths with the stability level index table of candidate path segments, speed parameter correction is performed on candidate path segments whose stability levels are within the set adjustment range. During the execution process, the system reads the candidate path segment numbers one by one, obtains the original set speed value from the basic speed dataset, and obtains the corresponding stability level from the stability level index table. The system sets different speed correction strategies based on the stability level classification, introduces the stability adjustment coefficient and upper and lower limit control mechanism by constructing a multi-segment linear correction function, and the corrected speed value is given by the following formula:

[0093] ;

[0094] in, :Number is The final corrected moving speed of the candidate path segment is in meters per second (m / s), which indicates the robot's movement speed in the path segment. The running speed after adjustment based on stability; :Number is The basic moving speed of the candidate path segment set in the planning stage, in meters per second (m / s), usually comes from the path planning module; : Stability correction coefficient, the value range is generally between 0.1 and 0.3, indicating the speed proportional reduction caused by each level of stability reduction; :Number is The stability level numerical mapping corresponding to the path segment is 0 for high stability, 1 for medium stability, and 2 for low stability, which are discrete classification values; : The minimum operating speed limit allowed by the robot under the current task or hardware constraints, used to prevent excessive deceleration from causing path execution failure, in meters per second; : The maximum allowable operating speed of the robot to prevent abnormal acceleration in the high stability section, in meters per second.

[0095] The formula adjusts the basic moving speed of the path segment by introducing the stability level, aiming to improve the safety and stability of the path execution. The system first looks up the table according to the stability level of the candidate path segment to obtain its corresponding value mapping (high stability is 0, medium stability is 1, and low stability is 2), and then combines it with the stability correction coefficient Calculate the correction factor , which is compared to the path segment's base velocity To prevent the speed from exceeding the runnable boundary, the final corrected speed is limited to the minimum speed through the upper and lower limit clipping functions. (i.e. the minimum operating speed allowed by the current task and hardware) and the maximum speed (i.e., the maximum speed allowed for the robot). This numerical mapping converts the abstract stability level into a quantifiable adjustment factor, so that path stability has a linear influence on speed control, thereby achieving the function of dynamically adjusting the operating speed based on path risk.

[0096] During the calculation process, the system first determines the correction amplitude based on the stability level of the path segment, and then limits the correction speed to a reasonable speed range to ensure that the adjustment process is controlled. The following is a specific explanation using two path segments as an example:

[0097] Example 1: The path segment number is 14, the base speed is 0.8 meters per second, and the stability level is medium stability. The mapping level value is , correction factor , the upper and lower speed limits are set to 0.1 meters per second and 1.2 meters per second, and the formula is:

[0098] ;

[0099] Example 2: Path segment number is 6, base speed is 0.6 m / s, stability level is low stability, corresponding mapping value , and its correction speed is:

[0100] .

[0101] The system calculates the corrected speeds for all path segments one by one in this manner. Upon completion, it generates a moving speed correction result table containing the path segment numbers and corrected speed values. This table will be used by the subsequent path execution module to complete dynamic speed control based on stability level feedback, effectively implementing differentiated driving strategies for path segments.

[0102] See also Figure 2 and Figure 7 ,It also includes a candidate path scheduling module, which calculates the score of each candidate path based on the end point weight update result and the moving speed correction result, combined with the multi-robot task allocation information, selects the target candidate path according to the score, and generates the candidate path task scheduling instruction;

[0103] The candidate path task scheduling instruction includes the candidate path scoring result and the target candidate path;

[0104] The candidate path scheduling module includes:

[0105] The candidate path scoring submodule calls the endpoint weight update results and the moving speed correction results, combines the endpoint scheduling weight in each candidate path with the corresponding speed correction parameter, and generates a candidate path basic scoring dataset;

[0106] The end point weight update result and the moving speed correction result are called to score and combine each candidate path. During the execution process, the system first indexes and matches the end point weight value and the speed correction value according to the path number to ensure that each candidate path can obtain these two parameters at the same time. Then the system performs combined calculation processing on a path segment basis, and fuses the end point weight value and the corresponding speed correction value using weighted aggregation or normalized scoring. During the combination process, the system can set different parameter contribution ratios to make the end point position play a stronger or weaker role in the scheduling priority. For example, in scenarios where posture stability is important, the weight of the speed factor in the score can be increased. Each candidate path will eventually get a comprehensive score value. The larger the value, the better the comprehensive scheduling conditions of the path. After all the scoring results are summarized and sorted by path number, a basic scoring data set for the candidate path is generated.

[0107] The candidate path conflict identification submodule determines the execution space distance between the candidate path endpoint and the endpoints of other robots' assigned candidate paths based on the candidate path basic score dataset and the multi-robot task allocation information in the current scheduling cycle. It adjusts the scores of candidate paths that conflict with the endpoint positions and obtains the candidate path conflict adjustment score results.

[0108] Conflict analysis is performed based on the basic scoring data set of candidate paths. During the execution process, the system calls the multi-robot task information assigned in the current scheduling cycle, extracts the end point positions of all robots' selected paths, and compares them with the end point positions of all current candidate paths. The system uses a distance threshold method to make conflict judgments. If the spatial distance between the two end points is less than the set safety distance, it is considered that there is a scheduling conflict. The system marks the candidate path as a conflict path and adjusts its score value. For example, a conflict penalty coefficient can be set to reduce the original score value by a certain proportion or directly set the minimum score to ensure that such paths are avoided first in subsequent scheduling decisions. At the same time, the system records the adjustment process in the candidate path scoring structure and outputs it as the candidate path conflict adjustment score result.

[0109] The scheduling instruction generation submodule extracts the priority candidate path number from the candidate path conflict adjustment score result, matches the corresponding candidate path task identifier with the robot execution sequence, and generates the candidate path task scheduling instruction;

[0110] Based on the candidate path conflict adjustment scoring results, the final instruction construction for the scheduling task is completed. During the execution process, the system first sorts all the scoring results in descending order by value, and filters out the path number list with the highest priority. The system matches each path number with its corresponding robot task identifier and scheduling sequence number, and reads the robot idle status and task load information during the scheduling cycle to generate scheduling instruction format data. Each instruction contains fields such as robot number, task number, path number and execution order. In the multi-robot parallel task scheduling scenario, the system assigns multiple priority paths to the idle robot queue in order of scores to avoid scheduling conflicts and resource waste. Finally, all instructions are uniformly output as candidate path task scheduling instructions to complete path assignment and task execution startup.

[0111] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A multi-robot collaborative control system, characterized by: The system comprises: The endpoint density analysis module obtains the coordinate set of all candidate path endpoints in the robot's current scheduling cycle, constructs the endpoint sampling boundary based on the endpoint coordinate set and divides the fixed grid area, calculates the local density of each grid area, and obtains the endpoint area density set; A weight updating module is configured to expand the grid area boundary of the endpoint position in the endpoint area density set that is less than the local density threshold and assign corresponding weights to obtain an endpoint weight update result; The posture evaluation module simulates the posture vector sequence of multiple robots moving along the candidate path, detects the posture change points in the posture vector sequence, and classifies the posture stability level of each candidate path segment according to the trend of the change points; a speed correction module, which corrects a basic movement speed parameter of the robot planned for each candidate path segment based on the posture stability level and generates a movement speed correction result; The endpoint area density set includes a grid density value and a grid division structure; the endpoint weight update result includes an endpoint weight and a boundary extension area; the posture stability level includes a posture change point and a candidate path segment stability level; and the moving speed correction result includes a corrected moving speed parameter and a corresponding relationship between the posture levels.

2. The multi-robot collaborative control system according to claim 1, characterized in that: The endpoint density analysis module includes: The end point coordinate acquisition submodule obtains the end point coordinate set of all candidate paths in the current scheduling cycle of the robot, performs two-dimensional spatial mapping processing on the end point position in each candidate path, and generates an end point coordinate distribution data set; The boundary grid construction submodule establishes a fixed boundary area around each endpoint according to the endpoint coordinate distribution data set, divides the boundary area into a consistent number of regular grid structures, and generates endpoint boundary grid structure information; The local density extraction submodule calls the endpoint boundary grid structure information, counts the number of endpoints in each grid area and calculates the local density based on the area of ​​the area, and numbers and marks the density of all grids to obtain the endpoint area density set.

3. The multi-robot collaborative control system according to claim 2, characterized in that: The local density is calculated based on the area of ​​the region using the formula: ； in, It is The local density of each grid cell reflects the density of the endpoints in the grid. is the area of ​​a single grid, It is the The sum of all endpoints in a grid, subscript Indicates the In the grid One falls into the end point, Is in the The number of end points of a grid, is a natural exponential function used to generate distance weights, It is The Euclidean distance between the end point and the current grid center, It is a smoothing control parameter that determines the changing speed of distance attenuation and its value is the upper limit of the robot navigation path positioning error.

4. The multi-robot collaborative control system according to claim 1, characterized in that: The weight updating module includes: The density threshold determination submodule calls the endpoint area density set, compares the density corresponding to each endpoint in the candidate path with the local density threshold, extracts the endpoint positions that are less than the local density threshold, and generates an endpoint density threshold matching list; The boundary area expansion submodule performs multi-directional equidistant expansion of the original boundary area corresponding to each endpoint according to the endpoint density threshold matching list, and redivides the expanded area into a standard grid structure. The number of endpoints in the expanded area is counted and the average density is calculated based on the area of ​​the area to generate endpoint expansion area density data. The weight value allocation submodule calls the density data of the terminal extension area, calculates the difference between the average density of the extension area and the average density of the original boundary area, performs scheduling weight value update on the terminal where the difference exceeds the density improvement judgment threshold, and obtains the terminal weight update result.

5. The multi-robot collaborative control system according to claim 1, characterized in that: The posture assessment module includes: The posture sequence construction submodule simulates the dynamic posture behavior of multiple robots moving along the candidate path, collects the pitch angle, roll angle and center of gravity coordinate values ​​of each robot in continuous motion on each candidate path segment, and constructs the posture change vector in chronological order to generate a multi-robot posture vector sequence set; A change point extraction submodule detects the change trend in the posture sequence of each candidate path segment using a cumulative sum control chart algorithm based on the multi-robot posture vector sequence set, calculates the cumulative offset of each group of posture values ​​relative to the mean, and identifies the position of the change points that appear in the continuous offset to obtain posture change point trend data; The grade interval matching submodule calls the posture change point trend data, extracts the joint features of the posture change amplitude and duration in the candidate path segment, performs interval matching classification on the joint features, and obtains the posture stability grade.

6. The multi-robot collaborative control system according to claim 1, characterized in that: The speed correction module includes: The basic speed extraction submodule obtains the basic movement speed parameters of the robot set in the planning stage for each candidate path segment, structures them according to the candidate path segment number, and generates a candidate path basic speed dataset; a stability level identification submodule, which calls the posture stability level, maps the stability level of each candidate path segment to the corresponding candidate path segment number, and forms a candidate path segment stability level index table; The speed value adjustment submodule combines the candidate path basic speed data set and the candidate path segment stability level index table to correct the speed parameters of the candidate path segments whose stability levels are within the set adjustment range to obtain a moving speed correction result.

7. The multi-robot collaborative control system according to claim 6, characterized in that: For speed parameter correction, the formula is used: ； in, It is numbered The final corrected moving speed of the candidate path segment, It is numbered The basic moving speed of the candidate path segment is set in the planning stage, is the stability correction factor, It is numbered The stability level numerical mapping corresponding to the path segment is a discrete classification value. It is the minimum operating speed limit allowed by the robot under the current task or hardware constraints. It is the upper limit of the robot's maximum allowed operating speed.

8. The multi-robot collaborative control system according to claim 1, characterized in that: The system also includes a candidate path scheduling module, which calculates the score of each candidate path based on the endpoint weight update result and the movement speed correction result, combined with the multi-robot task allocation information, selects the target candidate path according to the score, and generates the candidate path task scheduling instruction; The candidate path task scheduling instruction includes a candidate path scoring result and a target candidate path.

9. The multi-robot collaborative control system according to claim 8, characterized in that: The candidate path scheduling module includes: The candidate path scoring construction submodule calls the endpoint weight update result and the moving speed correction result, combines the endpoint scheduling weight and the corresponding speed correction parameter in each candidate path, and generates a candidate path basic scoring data set; The candidate path conflict identification submodule determines the execution space distance between the candidate path endpoint and the endpoints of other robots' assigned candidate paths based on the candidate path basic scoring dataset and the multi-robot task allocation information in the current scheduling cycle, and adjusts the scores of candidate paths that conflict with the endpoint positions to obtain candidate path conflict adjustment scoring results. The scheduling instruction generation submodule extracts the priority candidate path number of the candidate path conflict adjustment score result, matches the corresponding candidate path task identifier with the robot execution sequence, and generates the candidate path task scheduling instruction.

Images