A multi-robot path dynamic planning method and system

Abstract

This application relates to the field of control and regulation technology for non-electrical variables, specifically to a method and system for dynamic path planning of multiple robots. The method includes: obtaining the real-time avoidance coefficient of the target robot in the target area; determining the robot to be updated based on the real-time avoidance coefficient, and replanning the path of the robot to be updated to obtain candidate new paths; then obtaining the update priority of the target robot to be updated for the candidate new paths, and determining the path to be updated based on the update priority; finally, obtaining the efficiency reduction index of the target paths to be updated, and determining the target path to be updated with the smallest efficiency reduction index as the final update path. This achieves a balanced optimization of global system efficiency, effectively alleviating regional congestion and deadlock problems caused by cyclic avoidance in intensive operation scenarios, and improving the operational stability, collaborative efficiency, and path executability of multi-robot systems.

Description

Technical Field

[0001] This application belongs to the field of control and regulation technology for non-electrical variables, specifically relating to a path dynamic planning method and system for multiple robots. Background Technology

[0002] With the rapid development of intelligent manufacturing, smart warehousing, unmanned inspection, and flexible production, multi-robot collaborative operation systems have become core equipment in industrial automation, intelligent logistics, and other scenarios. Existing path planning technologies are mainly divided into two major approaches: centralized and distributed. Among them, distributed path planning technology achieves collaborative obstacle avoidance through independent perception and autonomous planning by each robot combined with communication interaction, and has strong environmental adaptability.

[0003] However, handling path coupling between multiple robots is challenging, easily leading to localized congestion and deadlock issues caused by multiple robots repeatedly avoiding obstacles in intensive work scenarios. Furthermore, the lack of comprehensive consideration of multi-dimensional engineering constraints, such as robot kinematics, energy consumption, and work sequence constraints, results in poor executability of planned paths, failing to meet the actual operational needs of industrial scenarios. Therefore, a dynamic path planning method and system for multiple robots is urgently needed. Summary of the Invention

[0004] To address the aforementioned issues, this application provides a method and system for dynamic path planning for multiple robots.

[0005] According to a first aspect of the embodiments of this application, a path dynamic planning method for multiple robots is provided, the method comprising: Obtain the real-time avoidance coefficient of the target robot in the target area; Based on the real-time avoidance coefficient, the robot to be updated is determined, and the path of the robot to be updated is replanned to obtain candidate new paths; Obtain the update priority of the target candidate new path among the candidate new paths by the target path to be updated robot, and determine the path to be updated based on the update priority; Obtain the efficiency reduction index of the target path to be updated in the path to be updated, and determine the target path to be updated with the smallest efficiency reduction index as the final update path.

[0006] In one implementation, obtaining the real-time avoidance coefficient of the target robot in the target area includes: Obtain the factors influencing congestion trends in the target area; Obtain the target gravity weight of the target robot; Obtain the comprehensive judgment factors for avoiding the target robot; Based on the congestion trend influencing factor, the target gravity weight, and the avoidance comprehensive judgment factor, the real-time avoidance coefficient of the target robot in the target area is obtained.

[0007] In one implementation, obtaining the congestion trend influencing factors of the target area includes: The heat index of the target area is obtained, which is based on the number of robots in the target area and the upper limit of the number of robots that the target area can accommodate; Based on the heat index, the variance of robot speed in the target area, and the variance of robot heading angle in the target area, the path chaos coefficient of the target area is obtained. Obtain the traffic capacity coefficient of the target area; Based on the path confusion coefficient and the traffic capacity coefficient, the road congestion level of the target area is obtained, and the congestion trend influencing factor of the target area is obtained based on the road congestion level.

[0008] In one implementation, obtaining the traffic capacity coefficient of the target area includes: The entire work scene area is rasterized to the same size to construct a global weighted raster map; Assign a unit traffic weight to the walkable road grid based on at least one of the following: road width, traffic attributes, distance to work station, and historical traffic efficiency. Obtain the access weight of the grid cell to which the target region belongs, wherein the access weight is proportional to the number of grid cells contained in the target region; Obtain the road width of the target area, the average width of all roads in the map, and the road traffic label value. The road traffic label value is used to distinguish between two-way traffic and one-way traffic, wherein the label value of two-way traffic is greater than the label value of one-way traffic. The traffic capacity coefficient of the target area is obtained based on the traffic weight of the grid to which the target area belongs, the road width, the average road width, and the road traffic label value.

[0009] In one implementation, obtaining the target gravity weight of the target robot includes: Obtain the remaining distance of the planned path of the target robot; Obtain the average speed of the target robot at the current moment, and the background preset priority of the target robot task; Obtain the minimum remaining distance of all robot planned paths in the target area; The target gravity weight of the target robot is obtained based on the remaining distance, the average speed, the preset priority, and the minimum value of the remaining distance.

[0010] In one implementation, obtaining the comprehensive avoidance judgment factor of the target robot includes: The motion constraint factor of the target robot is obtained, which is based on the degree of matching between the target robot's maximum allowable speed, maximum allowable acceleration, and minimum turning radius and the real-time speed, real-time acceleration, and real-time turning radius required for the avoidance path. The energy consumption constraint factor of the target robot is obtained, which is based on the remaining power of the target robot and the energy consumption required for the avoidance path. The operation timing constraint factor of the target robot is obtained, which is based on the remaining task time of the target robot, the maximum operation delay threshold, and the estimated travel time when the target area is congested. Based on the motion constraint factor, the energy consumption constraint factor, and the operation timing constraint factor, the comprehensive avoidance judgment factor of the target robot is obtained.

[0011] In one implementation, determining the robot whose path needs updating includes: Obtain the avoidance coefficient threshold; The target robot whose real-time avoidance coefficient is greater than or equal to the avoidance coefficient threshold is identified as the path robot to be updated.

[0012] In one implementation, the step of obtaining the update priority of the target candidate new paths among the candidate new paths by the robot acquiring the target path to be updated includes: The efficiency reduction coefficient of the target path to be updated robot on the target candidate new path is obtained. The efficiency reduction coefficient is based on the predicted avoidance coefficient of the target path to be updated robot on the target candidate new path and the predicted road congestion degree of the target candidate new path. Based on the efficiency reduction coefficient, the update priority of the target candidate new path by the target path-to-update robot is obtained.

[0013] In one implementation, obtaining the efficiency reduction index of the target path to be updated in the path to be updated includes: The influence intensity of the target path robot to be updated on the existing target robots on the target path to be updated is obtained, and the influence intensity is obtained based on the average speed, maximum moving speed and heading angle of the target path robot to be updated, as well as the average speed and heading angle of the existing target robots. Based on the intensity of the influence, determine the overall intensity of the influence of the target path to be updated robot on the target path to be updated; Based on the combined influence intensity and the path disorder coefficient of the target path to be updated, the interference coefficient of the robot on the target path to be updated is obtained. Based on the interference coefficient and the road congestion level of the target path to be updated, the efficiency reduction index of the target path to be updated is obtained.

[0014] According to a second aspect of the embodiments of this application, a multi-robot path dynamic planning system is provided, the system including a planning platform, the planning platform comprising: A memory on which computer programs are stored; A processor for executing the computer program in the memory to implement the steps of the method of any one of the first aspects.

[0015] The embodiments of this application have the following beneficial effects: This application provides a method and system for dynamic path planning of multiple robots. The method includes: obtaining the real-time avoidance coefficient of the target robot in the target area; determining the robot to be updated based on the real-time avoidance coefficient, and replanning the path of the robot to be updated to obtain candidate new paths; then obtaining the update priority of the target robot to be updated to the target candidate new paths, and determining the path to be updated according to the update priority; finally obtaining the efficiency reduction index of the target path to be updated, and determining the target path to be updated with the smallest efficiency reduction index as the final update path. This achieves a balanced optimization of global system efficiency, effectively alleviating the deadlock problem caused by regional congestion and cyclic avoidance in intensive operation scenarios, and improving the operational stability, collaborative efficiency, and path executability of the multi-robot system. Attached Figure Description

[0016] Figure 1 This is a flowchart of a multi-robot path dynamic planning method; Figure 2 This is a flowchart illustrating a method for obtaining the real-time avoidance coefficient of a target robot in a target area; Figure 3 This is a flowchart of a method for obtaining the influencing factors of congestion trends in a target area; Figure 4 This is a flowchart of the method for obtaining the traffic capacity coefficient of a target area; Figure 5 This is a flowchart of the method for obtaining the target gravity weight of the target robot; Figure 6 This is a flowchart of a method for obtaining the comprehensive judgment factors for avoiding a target robot; Figure 7 This is a flowchart illustrating the method for determining the robot whose path needs to be updated; Figure 8This is a flowchart of a method for obtaining the update priority of the target candidate new path among the candidate new paths by the robot to obtain the target path to be updated; Figure 9 This is a flowchart of a method for obtaining the efficiency reduction index of the target path to be updated in the path to be updated. Detailed Implementation

[0017] Figure 1 This is a flowchart illustrating a multi-robot path dynamic planning method according to an exemplary embodiment. Figure 1 As shown in the figure, this application provides a method for dynamic path planning for multiple robots, which may include the following steps: In step S10, the real-time avoidance coefficient of the target robot in the target area is obtained.

[0018] In step S20, based on the real-time avoidance coefficient, a robot to be updated is determined, and the path of the robot to be updated is replanned to obtain a candidate new path.

[0019] In step S30, the update priority of the target candidate new path among the candidate new paths is obtained by the target path to be updated robot, and the path to be updated is determined according to the update priority.

[0020] In step S40, the efficiency reduction index of the target path to be updated in the path to be updated is obtained, and the target path to be updated with the smallest efficiency reduction index is determined as the final update path.

[0021] For example, the system can first obtain the real-time avoidance coefficient of the target robot in the target area. This coefficient comprehensively reflects the avoidance pressure faced by the robot in a specific area. Based on this real-time avoidance coefficient, the system determines which robots are currently in a state of high avoidance pressure and identifies them as robots whose paths need to be updated. Then, the system replans the paths of these robots to generate multiple candidate new paths.

[0022] It should be noted that this application considers each intersection in the work scenario as a target area, and the path connecting two intersections as a target area as well. "Target path to be updated" refers to any path to be updated, "target robot" refers to any robot, and "target area" refers to any area.

[0023] Next, for each robot with a path to be updated, the system evaluates its update priority for each candidate new path. This priority represents the potential impact on the robot's efficiency after selecting the path, and suitable update paths are selected based on the update priority. For example, candidate new paths with an update priority greater than a preset update threshold (e.g., 0.6, an empirical value) can be selected as update paths. This reduces unnecessary path updates. Finally, the comprehensive impact of each update path on the overall system efficiency, i.e., the efficiency reduction index, can be calculated, and the target update path with the smallest efficiency reduction index can be selected as the final update path.

[0024] Through this complete path dynamic planning process, the embodiments of this application can proactively identify and alleviate potential congestion and deadlock problems during multi-robot collaborative operations, prioritize the path optimization needs of robots with high avoidance pressure, and minimize the negative impact on the overall system efficiency, thereby significantly improving the operational stability and task execution efficiency of multi-robot systems in intensive operation scenarios.

[0025] Figure 2 This is a flowchart illustrating a method for obtaining the real-time avoidance coefficient of a target robot in a target area, according to an exemplary embodiment. Figure 2 As shown, obtaining the real-time avoidance coefficient of the target robot in the target area may include the following steps: In step S101, the congestion trend influencing factor of the target area is obtained.

[0026] In step S102, the target gravity weight of the target robot is obtained.

[0027] In step S103, the comprehensive judgment factor for avoiding the target robot is obtained.

[0028] In step S104, the real-time avoidance coefficient of the target robot in the target area is obtained based on the congestion trend influence factor, the target gravity weight, and the avoidance comprehensive judgment factor.

[0029] For example, a congestion trend influencing factor for the target area can be obtained first, which quantifies the probability of congestion occurring in the target area within a future period. Simultaneously, the target robot's target gravity weight can be obtained, reflecting the robot's passage demand intensity due to factors such as its task objective, remaining distance, and task priority. Furthermore, the target robot's comprehensive avoidance judgment factor can be obtained, which integrates the robot's avoidance feasibility under multiple constraints such as mobility, energy consumption, and operational timing. Finally, the congestion trend influencing factor, target gravity weight, and comprehensive avoidance judgment factor can be jointly calculated to obtain the real-time avoidance coefficient.

[0030] For example, the avoidance coefficient of the j-th target robot at time t in the i-th target region. It can be obtained from the following formula: Formula 1 in, Let be the congestion trend influencing factor for the i-th target area at time t. Let the target gravity weight be the gravitational weight of the j-th target robot at time t. Let cj be the comprehensive avoidance judgment factor for the j-th target robot at time t; c0 is a preset minimum positive constant used to prevent the denominator from being 0, which is taken here. Typically much lower It has little impact on the denominator.

[0031] The calculation process in this embodiment of the application ensures that the avoidance coefficient not only reflects the environmental pressure of the area but also incorporates the robot's own task urgency and physical limitations, achieving precise quantification of the robot's avoidance requirements. This enables the system to dynamically and adaptively identify robots that face high-risk congestion areas, have strong traffic rigidity, and are difficult to avoid, providing a scientific and reasonable decision-making basis for subsequent path replanning.

[0032] Figure 3 This is a flowchart illustrating a method for obtaining congestion trend influencing factors of a target area according to an exemplary embodiment. Figure 3 As shown, obtaining the congestion trend influencing factors of the target area may include the following steps: In step S1011, the heat coefficient of the target area is obtained, which is based on the number of robots in the target area and the upper limit of the number of robots that the target area can accommodate.

[0033] In step S1012, the path disorder coefficient of the target area is obtained based on the heat coefficient, the variance of the robot speed in the target area, and the variance of the robot heading angle in the target area.

[0034] In step S1013, the traffic capacity coefficient of the target area is obtained.

[0035] In step S1014, the road congestion level of the target area is obtained based on the path confusion coefficient and the traffic capacity coefficient, and the congestion trend influence factor of the target area is obtained based on the road congestion level.

[0036] For example, the heat index of the target area can be obtained first. This index is calculated based on the number of robots in the target area at the current moment and the upper limit of the number of robots that the physical space of the area can accommodate, reflecting the density of robots in the area. On this basis, the heat index, the variance of robot speed in the area, and the variance of robot heading angle can be combined to calculate the path disorder index. This index comprehensively reflects the degree of disorder of robots in terms of spatial distribution, travel direction, and speed consistency.

[0037] Simultaneously, the traffic capacity coefficient of the target area can be obtained, which represents the physical traffic conditions of the road itself. Subsequently, based on the ratio of the path disorder coefficient to the traffic capacity coefficient, and combined with the congestion status of the previous moment, a time-series recursive calculation can be performed to obtain the road congestion level at the current moment. Finally, the congestion level is normalized and used as the congestion trend influencing factor.

[0038] For example, the heat coefficient of the i-th target region at time t. It can be obtained from the following formula: Formula 2 in, Let be the number of robots in the i-th target region at time t. This represents the maximum number of robots that can be accommodated in the physical space of the i-th target area. It is not zero. It should be understood that when the number of robots in the i-th target area at time t is 0, its heat coefficient is assigned a value of 0.

[0039] Path confusion coefficient of the i-th target region at time t It can be obtained from the following formula: Formula 3 in, Let be the variance of the robot's velocity in the i-th target region at time t. The larger the variance, the greater the dispersion of the velocity and the larger the disorder coefficient. Let Variance be the robot heading angle variance in the i-th target region at time t. Preset weighting coefficients for the velocity variance term, Preset weighting coefficients for the heading angle variance term, These are preset weighting coefficients for the heat coefficient; each weighting coefficient is used to adjust the proportion of the corresponding physical quantity. This is for normalization purposes.

[0040] In this example, , , In dynamic planning scenarios involving intensive multi-robot operations, velocity variance and heading angle variance are explicit indicators that directly reflect the frequent dynamic conflict behaviors of robot groups in a local area, such as deceleration, avoidance, and turning. They contribute more to the degree of path chaos and have a more real-time response, so they are given higher weights. On the other hand, the regional heat coefficient is a robot distribution density statistically analyzed from the macro-business level. Although it can reflect potential congestion risks, it cannot directly reflect the intensity of current kinematic conflict. Its influence on the instantaneous path chaos at the current moment is relatively low, so it is given the lowest weight.

[0041] It should be noted that, unless otherwise specified in this application, All normalization processes use max-min normalization. The comparison set is set to the normalization target, encompassing all historical and real-time data. If the current real-time value is the maximum value, the normalization result is 1.

[0042] The road congestion level of the i-th target area at time t It can be obtained from the following formula: Formula 4 in, Let represent the road congestion level of the i-th target area at time t. Let i be the road congestion level of the i-th target area at time t-1. This is the time correlation coefficient between adjacent time windows, for example, it can be 0.7. This value is an empirical value. Congestion has physical inertia, that is, it forms slowly, dissipates slowly, and does not change abruptly. Therefore, historical state should have the dominant weight, and current compensation only makes corrections. At the same time, in the field of industrial control, the standard compromise coefficient for first-order inertial filtering is 0.6~0.8. Here, we use the middle value of this theory to achieve a theoretical balance between filtering effect and response speed. Let be the road capacity coefficient for the i-th target area. Not zero; Indicates the first The road congestion compensation value for the i-th target area at time i, that is, the weaker the road capacity and the... The greater the moment-to-moment chaos coefficient, the greater the road congestion.

[0043] Congestion trend influencing factors of the i-th target area at time t It can be obtained from the following formula: norm ( ) Formula 5 Here, norm represents the normalization process.

[0044] This embodiment of the application not only considers the static congestion situation at the current moment, but also transmits the historical impact of congestion to the future through time correlation coefficients, thereby realizing the advance prediction of congestion trends in any region and at any time, and providing forward-looking environmental situation information for route planning.

[0045] Figure 4 This is a flowchart illustrating a method for obtaining the traffic capacity coefficient of a target area according to an exemplary embodiment. Figure 4 As shown, obtaining the traffic capacity coefficient of the target area may include the following steps: In step S10131, the entire working scene area is rasterized to the same size to construct a global weighted raster map.

[0046] In step S10132, a unit traffic weight is assigned to the walkable road grid based on at least one of the following: road width, traffic attributes, distance to the work station, and historical traffic efficiency.

[0047] In step S10133, the access weight of the grid to which the target area belongs is obtained, and the access weight is proportional to the number of grids contained in the target area.

[0048] In step S10134, the road width of the target area, the average width of all roads in the map, and the road traffic label value are obtained. The road traffic label value is used to distinguish between two-way traffic and one-way traffic, wherein the label value of two-way traffic is greater than the label value of one-way traffic.

[0049] In step S10135, the traffic capacity coefficient of the target area is obtained based on the traffic weight of the grid to which the target area belongs, the road width, the average value of the road width, and the road traffic label value.

[0050] For example, the entire work scene can be first processed into a grid of equal size to build a global weighted grid map for subsequent processing: for impassable obstacles or restricted areas, their passage weight can be set to zero; for walkable road grids, a unit passage weight can be assigned based on various factors such as road width, passage attributes, distance from workstations, and historical passage efficiency, and this weight is positively correlated with the actual passage efficiency.

[0051] Assuming the work scenario is a 100m x 100m smart warehouse floor plan, perform uniform-size gridding. A uniform grid size of 5m x 5m can be set (all grids are exactly the same size), dividing the space into 20 columns x 20 rows, for a total of 400 uniform-size grids. Grid weights are set according to the content attributes within each grid.

[0052] Since the work scenarios (workshops / warehouses) are mostly irregularly shaped, the uniform-sized grid transforms the physical space into regular grid cells, eliminating the computational errors caused by irregular areas. The uniform-sized grid data structure is simple, and the computational load for robot localization, path replanning, and obstacle avoidance path retrieval is small. It avoids complex geometric calculations across the entire scene, improves computational efficiency, and facilitates simpler path planning.

[0053] Next, the traffic weight of the grid cells belonging to the target area can be obtained, which is proportional to the number of grid cells contained in that area. Simultaneously, the road width of the target area, the average road width across the entire map, and the road traffic label values ​​used to distinguish between two-way and one-way traffic can be obtained, with two-way traffic having a higher label value than one-way traffic. Finally, these parameters can be substituted into a specific formula to calculate the traffic capacity coefficient, which comprehensively reflects the physical space capacity and traffic convenience of the target area.

[0054] For example, the traffic capacity coefficient of the i-th target area. It can be obtained from the following formula: Formula 6 in, For the first The passage weight of the raster to which each target region belongs. Let the width of the road in the i-th target area be . This is the average road width across all target areas; this value is not zero. For the road access label value of the i-th target area, when two-way traffic is allowed... One-way traffic , This is for normalization purposes.

[0055] This application embodiment quantifies the differentiated traffic capacity of different roads, enabling congestion trend analysis to closely match the road characteristics of actual work scenarios, avoiding misjudgments caused by ignoring the differences in road conditions, thereby improving the effectiveness and feasibility of route planning.

[0056] Figure 5 This is a flowchart illustrating a method for obtaining the target gravity weight of a target robot according to an exemplary embodiment. Figure 5 As shown, obtaining the target gravity weight of the target robot may include the following steps: In step S1021, the remaining distance of the target robot's planned path is obtained.

[0057] In step S1022, the average speed of the target robot at the current moment and the background preset priority of the target robot task are obtained.

[0058] In step S1023, the minimum remaining distance of all robot planned paths in the target area is obtained.

[0059] In step S1024, the target gravity weight of the target robot is obtained based on the remaining distance, the average speed, the preset priority, and the minimum value of the remaining distance.

[0060] For example, the remaining distance of the target robot's planned path can be obtained first, along with the robot's current average speed and its task's preset priority in the background. Simultaneously, the minimum remaining distance among all robot planned paths within the target area can be obtained. Based on this, the robot's remaining distance can be divided by its average speed to obtain a component reflecting the remaining expected travel time. This component is then multiplied by the reciprocal of the minimum remaining distance to obtain a task urgency index representing the target robot's urgency relative to the nearest task target within the area. Finally, a preset task priority (automatically assigned by the system background, with a preset value range of [0, 1], where a larger value indicates a higher priority) is added to form the target gravity weight.

[0061] For example, the target gravity weight of the j-th target robot at time t. It can be obtained from the following formula: Formula 7 in, Let be the remaining distance of the planned path of the j-th target robot at time t. Let be the average velocity of the j-th target robot up to time t. Let be the minimum remaining distance among all robot-planned paths within the i-th target region at time t. The task background is preset with priority for the j-th target robot; This represents the distance weighting coefficient, used to adjust the influence of distance on gravity. Adding 0.01 to the denominator is to prevent the denominator from being zero. It is usually much lower than the physical quantity corresponding to the denominator and has little impact on the denominator.

[0062] The embodiments of this application enable the target gravity weight to dynamically reflect the current task urgency of the robot: the greater the remaining distance, the slower the average speed, and the higher the task priority of the robot, the greater its target gravity weight, indicating that it needs to pass first. This effectively solves the problem in traditional methods where low-priority robots continuously avoid obstacles and high-priority tasks are delayed due to fixed or simply set target gravity weights, ensuring the timely completion of critical tasks.

[0063] Figure 6 This is a flowchart illustrating a method for obtaining a comprehensive judgment factor for avoiding a target robot, according to an exemplary embodiment. Figure 6As shown, obtaining the comprehensive judgment factor for avoiding the target robot may include the following steps: In step S1031, the motion constraint factor of the target robot is obtained. The motion constraint factor is obtained based on the degree of matching between the target robot's maximum allowable speed, maximum allowable acceleration, and minimum turning radius and the real-time speed, real-time acceleration, and real-time turning radius required for the avoidance path.

[0064] In step S1032, the energy consumption constraint factor of the target robot is obtained, which is based on the remaining power of the target robot and the energy consumption required for the avoidance path.

[0065] In step S1033, the operation timing constraint factor of the target robot is obtained. The operation timing constraint factor is obtained based on the remaining task time of the target robot, the maximum operation delay threshold, and the estimated travel time when the target area is congested.

[0066] In step S1034, the comprehensive avoidance judgment factor of the target robot is obtained based on the motion constraint factor, the energy consumption constraint factor, and the operation timing constraint factor.

[0067] For example, constraints can be evaluated from three dimensions: motion, energy consumption, and timing. First, a motion constraint factor can be obtained. This factor is calculated based on the degree of matching between the target robot's maximum permissible speed, maximum permissible acceleration, and minimum turning radius, and the real-time speed, acceleration, and turning radius required for the avoidance path. A higher degree of matching indicates a stronger physical feasibility for the target robot to execute the avoidance path, and a more stringent constraint. The avoidance path is an alternative travel path replanned for robots under high avoidance pressure, used to replace the robot's original planned path to complete the avoidance behavior.

[0068] Secondly, an energy consumption constraint factor can be obtained. This factor is based on the ratio of the target robot's remaining battery power to the energy required for the avoidance path; the less power available to support avoidance, the stronger the energy consumption constraint. Thirdly, a task timing constraint factor can be obtained. This factor is calculated based on the robot's remaining task time, the maximum operational delay threshold, and the estimated travel time when the target area is congested; the tighter the remaining time, the stronger the timing constraint. Finally, these three constraint factors can be added together to obtain the comprehensive avoidance judgment factor.

[0069] For example, the motion constraint factor of the j-th target robot at time t. It can be obtained from the following formula: Formula 8 in, To avoid the path Required speed Let j be the maximum permissible speed of the target robot. To avoid the path Required acceleration Let j be the maximum allowable acceleration of the target robot. To avoid the path Required turning radius Let be the minimum turning radius of the j-th target robot, and have In other words, the turning radius required for the obstacle avoidance path must be greater than or equal to the robot's physical minimum radius. To obtain the minimum value.

[0070] Energy consumption constraint factor of the j-th target robot at time t It can be obtained from the following formula: Formula 9 in, Let be the remaining battery power of the j-th target robot at time t. To avoid the path Required energy consumption, this value is not zero. This is for normalization purposes.

[0071] The timing constraint factor for the j-th target robot at time t. It can be obtained from the following formula: Formula 10 in, Let be the remaining task time for the j-th target robot at time t. Let be the estimated travel time for the i-th region at time t when congestion occurs. Let be the maximum task delay threshold for the j-th target robot; this value is not zero. This is for normalization purposes.

[0072] The comprehensive judgment factor for avoiding the j-th target robot at time t. It can be obtained from the following formula: Formula 11 The embodiments of this application comprehensively evaluate the overall cost and feasibility of robots performing obstacle avoidance behavior, avoiding the planning of obstacle avoidance paths that cannot be actually executed due to insufficient power, inadequate acceleration and deceleration capabilities, or insufficient time, thereby significantly improving the engineering feasibility and robustness of the planned path.

[0073] Figure 7 This is a flowchart illustrating a method for determining a robot's path to be updated according to an exemplary embodiment. Figure 7As shown, determining the robot whose path needs updating may include the following steps: In step S201, the avoidance coefficient threshold is obtained.

[0074] In step S202, the target robot whose real-time avoidance coefficient is greater than or equal to the avoidance coefficient threshold is determined as the path robot to be updated.

[0075] For example, the calculated real-time avoidance coefficient of each target robot in its target area can be compared with an avoidance coefficient threshold (e.g., 0.7, an empirical value). All robots with real-time avoidance coefficients greater than or equal to this threshold are identified as robots whose paths need updating. The avoidance coefficient threshold of 0.7 is an engineering experience-based balance point between system stability and response efficiency in intensive work scenarios. This threshold corresponds to the high-pressure zone starting line (70th percentile) of the normalized avoidance coefficient index, accurately identifying "robots whose paths need updating" that face significant congestion trends, have high task urgency, and incur high avoidance costs.

[0076] If the threshold is too low (e.g., 0.5), it will cause a large number of robots that have not yet formed a real congestion to frequently trigger path replanning, resulting in a waste of computing resources and path oscillation, which will exacerbate the loop avoidance and system deadlock. If the threshold is too high (e.g., 0.9), the robots are often trapped in a severe congestion or deadlocked state of waiting for each other, missing the optimal adjustment window and making it difficult to achieve effective traffic flow. Therefore, a middle value is chosen as the judgment threshold to achieve the desired result.

[0077] This allows for priority handling of robots in a state of "high avoidance pressure," improving the overall efficiency of the system.

[0078] This application's embodiments can accurately identify objects currently under "high avoidance pressure" from a large number of online robots. If these robots continue to travel along their original paths, they are highly likely to get caught in a dilemma of repeated avoidance, continuous decline in efficiency, or even deadlock. By prioritizing path replanning for these high-avoidance-pressure robots, computational resources are concentrated on resolving the most urgent path conflicts, while avoiding the huge computational overhead of replanning for all robots, a good balance is achieved between system efficiency and computational resource consumption.

[0079] Figure 8 This is a flowchart illustrating a method for obtaining the update priority of a target new path among candidate new paths by a robot seeking to update a target new path, according to an exemplary embodiment. Figure 8 As shown, the step of obtaining the update priority of the target candidate new path among the candidate new paths by the robot to obtain the target path to be updated may include the following steps: In step S301, the efficiency reduction coefficient of the target path to be updated robot on the target candidate new path is obtained. The efficiency reduction coefficient is obtained based on the predicted avoidance coefficient of the target path to be updated robot on the target candidate new path and the predicted road congestion level of the target candidate new path.

[0080] In step S302, based on the efficiency reduction coefficient, the update priority of the target path to be updated robot for the target candidate new path is obtained.

[0081] For example, the efficiency reduction coefficient of the target robot on a certain candidate new path can be calculated first. This efficiency reduction coefficient is based on two core prediction values: the predicted avoidance coefficient of the target robot on the new path, and the predicted road congestion level of the new path itself. Specifically, it is possible to predict the avoidance pressure and congestion status of the target robot at future moments after switching to the new path, and thereby quantify the degree of loss of the path to the robot's own operating efficiency.

[0082] Subsequently, based on this efficiency reduction coefficient, and through mathematical transformations such as taking the reciprocal, the robot's update priority for the candidate new path to the target can be obtained. The advantage of this process is that it makes the determination of the update priority depend not only on the distance or time of the path, but also on the congestion trend of the path in future actual operation and its impact on the robot's avoidance behavior.

[0083] For example, the efficiency reduction coefficient of robot k on the target candidate new path v. It can be obtained from the following formula: Formula 12 in, Let V be the road congestion level at time t+1 for the v-th candidate new path, i.e., the predicted road congestion level. For the first The avoidance coefficient of the robot on the candidate new path v at time t+1, i.e., the predicted avoidance coefficient. To obtain the maximum value, This is for normalization purposes.

[0084] The update priority of the target path to be updated by robot k to the target candidate new path v. It can be obtained from the following formula: Formula 13 The 1 in the denominator is used to ensure that the update priority decreases smoothly as the efficiency reduction coefficient increases, and to avoid the denominator being zero.

[0085] The embodiments of this application can automatically filter out candidate new paths that, although they may be slightly longer, can significantly reduce the robot's future avoidance pressure and thus improve its overall operating efficiency, thereby realizing decision optimization from "shortest path" to "most efficient path".

[0086] Figure 9 This is a flowchart illustrating a method for obtaining the efficiency reduction index of a target path to be updated in an exemplary embodiment. Figure 9 As shown, obtaining the efficiency reduction index of the target path to be updated in the path to be updated may include the following steps: In step S401, the influence intensity of the target path robot to be updated on the existing target robot on the target path to be updated is obtained. The influence intensity is obtained based on the average speed, maximum moving speed and heading angle of the target path robot to be updated, as well as the average speed and heading angle of the existing target robot.

[0087] In step S402, the comprehensive influence intensity of the target path to be updated robot on the target path to be updated is determined based on the influence intensity.

[0088] In step S403, the interference coefficient of the robot on the target path to be updated is obtained based on the comprehensive influence intensity and the path confusion coefficient of the target path to be updated.

[0089] In step S404, the efficiency reduction index of the target path to be updated is obtained based on the interference coefficient and the road congestion level of the target path to be updated.

[0090] For example, the embodiments of this application fully consider the impact of the robot to be updated on existing robots on the path. First, the influence intensity of the target robot on the target path to be updated on the existing target robots on the target path can be calculated. This influence intensity is obtained based on parameters such as the average speed difference between the two, the heading angle difference, and the maximum moving speed of the target robot, reflecting the degree of speed interference and direction interference.

[0091] Next, based on this impact strength, and considering all affected existing robots, the overall impact strength of the robot to be updated on the entire path to be updated can be determined. Then, this overall impact strength is multiplied by the path chaos coefficient of the target path to be updated itself to obtain the interference coefficient, which represents the additional chaos that the robot to be updated may cause after being inserted into the path. Finally, by combining the interference coefficient and the current road congestion level of the path, and adjusting through weighted summation and correction coefficients, the final efficiency reduction index can be obtained.

[0092] For example, the influence strength of target robot k on the target path to be updated on the existing target robot q on the target path to be updated u. It can be obtained from the following formula: Formula 14 in, Let k be the average moving speed of the robot whose path needs to be updated. Let q be the average moving speed of the existing target robot q on the path u to be updated. Let k be the maximum moving speed of the robot whose path needs to be updated. Let the heading angle be the target path of robot k. Let be the heading angle of the existing target robot q on the path u to be updated. The heading angle is in degrees. For normalization processing; This represents interference with the robot's movement direction; a larger value indicates that the k-th robot is affecting the movement direction of the th robot. The greater the impact of each robot, the stronger its influence.

[0093] The combined influence strength of robot k on the target path to be updated u. It can be obtained from the following formula: Formula 15 in, The target is the number of existing robots on the path u to be updated. Not zero.

[0094] If the number of existing robots on the target path u to be updated is 0, then the path repulsion coefficient is directly assigned to 0, indicating that when the target robot k switches to this path, it is not subject to any spatial repulsion interference from any vehicles traveling in the same or opposite direction.

[0095] The interference coefficient of robot k on the target path to be updated u It can be obtained from the following formula: Formula 16 in, Let u be the path disorder coefficient at time t, which is the target path to be updated. This is for normalization purposes.

[0096] when The time is without intervention, indicating the first The path to be updated is acceptable. A robot with a path to be updated; when At this stage, a mild intervention is needed, prioritizing other paths that are yet to be updated; when At this time, the intervention was severe, and updates to the previous level were prohibited. There are several paths to be updated. The thresholds of 0.4 and 0.7 are empirical values ​​and can be adjusted based on actual operating conditions.

[0097] The efficiency reduction exponential of robot k on target path u to be updated. It can be obtained from the following formula: Formula 17 in, The factor representing the impact of the intervention on efficiency, for example, 0.8, is used to adjust the interference coefficient. Efficiency reduction index Influence weight, Let u be the target path to be updated, and let u be the road congestion level at time t. This is a correction factor for the intervention type; when there is no intervention... When mild intervention Severe intervention E is an empirical value that can be adjusted based on actual operating conditions. This is for normalization purposes.

[0098] Regarding the first For each robot with a path to be updated, calculate the average update priority of its path, denoted as the i-th. The final path update priority of each robot to be updated is determined, and the robots are sorted in descending order of their final path update priority to obtain a sequence of robots to be updated. Then, based on this sequence, the efficiency reduction index of each path to be updated is selected. The shortest path is used as the update target to achieve dynamic programming of the path.

[0099] This application embodiment quantifies the comprehensive loss of path update behavior on the operating efficiency of local areas or even the entire system from the perspective of the global system, thereby avoiding the "robbing Peter to pay Paul" style of local optimization and ensuring that the finally selected update path has the lowest negative efficiency impact at the global level.

[0100] This application also provides a multi-robot path dynamic planning system, including a planning platform. The planning platform includes: a memory storing a computer program; and a processor for executing the computer program in the memory to implement the above-described multi-robot path dynamic planning method.

[0101] This application also provides a computer-readable storage medium storing computer program instructions that, when executed by a processor, implement the steps of the multi-robot path dynamic planning method provided in this application.

Claims

1. A path dynamic planning method for multiple robots, characterized in that, The method includes: Obtain the real-time avoidance coefficient of the target robot in the target area; Based on the real-time avoidance coefficient, the robot to be updated is determined, and the path of the robot to be updated is replanned to obtain candidate new paths; Obtain the update priority of the target candidate new path among the candidate new paths by the target path to be updated robot, and determine the path to be updated based on the update priority; Obtain the efficiency reduction index of the target path to be updated in the path to be updated, and determine the target path to be updated with the smallest efficiency reduction index as the final update path.

2. The multi-robot path dynamic planning method according to claim 1, characterized in that, The process of obtaining the real-time avoidance coefficient of the target robot in the target area includes: Obtain the factors influencing congestion trends in the target area; Obtain the target gravity weight of the target robot; Obtain the comprehensive judgment factors for avoiding the target robot; Based on the congestion trend influencing factor, the target gravity weight, and the avoidance comprehensive judgment factor, the real-time avoidance coefficient of the target robot in the target area is obtained.

3. The multi-robot path dynamic planning method according to claim 2, characterized in that, The factors influencing the congestion trend in the target area include: The heat index of the target area is obtained, which is based on the number of robots in the target area and the upper limit of the number of robots that the target area can accommodate; Based on the heat index, the variance of robot speed in the target area, and the variance of robot heading angle in the target area, the path chaos coefficient of the target area is obtained. Obtain the traffic capacity coefficient of the target area; Based on the path confusion coefficient and the traffic capacity coefficient, the road congestion level of the target area is obtained, and the congestion trend influencing factor of the target area is obtained based on the road congestion level.

4. The multi-robot path dynamic planning method according to claim 3, characterized in that, The process of obtaining the traffic capacity coefficient of the target area includes: The entire work scene area is rasterized to the same size to construct a global weighted raster map; Assign a unit traffic weight to the walkable road grid based on at least one of the following: road width, traffic attributes, distance to work station, and historical traffic efficiency. Obtain the access weight of the grid cell to which the target region belongs, wherein the access weight is proportional to the number of grid cells contained in the target region; Obtain the road width of the target area, the average width of all roads in the map, and the road traffic label value. The road traffic label value is used to distinguish between two-way traffic and one-way traffic, wherein the label value of two-way traffic is greater than the label value of one-way traffic. The traffic capacity coefficient of the target area is obtained based on the traffic weight of the grid to which the target area belongs, the road width, the average road width, and the road traffic label value.

5. The multi-robot path dynamic planning method according to claim 2, characterized in that, The acquisition of the target gravity weight of the target robot includes: Obtain the remaining distance of the planned path of the target robot; Obtain the average speed of the target robot at the current moment, and the background preset priority of the target robot task; Obtain the minimum remaining distance of all robot planned paths in the target area; The target gravity weight of the target robot is obtained based on the remaining distance, the average speed, the preset priority, and the minimum value of the remaining distance.

6. The multi-robot path dynamic planning method according to claim 2, characterized in that, The comprehensive judgment factors for obtaining the avoidance of the target robot include: The motion constraint factor of the target robot is obtained, which is based on the degree of matching between the target robot's maximum allowable speed, maximum allowable acceleration, and minimum turning radius and the real-time speed, real-time acceleration, and real-time turning radius required for the avoidance path. The energy consumption constraint factor of the target robot is obtained, which is based on the remaining power of the target robot and the energy consumption required for the avoidance path. The operation timing constraint factor of the target robot is obtained, which is based on the remaining task time of the target robot, the maximum operation delay threshold, and the estimated travel time when the target area is congested. Based on the motion constraint factor, the energy consumption constraint factor, and the operation timing constraint factor, the comprehensive avoidance judgment factor of the target robot is obtained.

7. The multi-robot path dynamic planning method according to claim 1, characterized in that, The process of determining the robot whose path needs to be updated includes: Obtain the avoidance coefficient threshold; The target robot whose real-time avoidance coefficient is greater than or equal to the avoidance coefficient threshold is identified as the path robot to be updated.

8. The multi-robot path dynamic planning method according to claim 3, characterized in that, The update priority of the target candidate new path among the candidate new paths by the robot that obtains the target path to be updated includes: The efficiency reduction coefficient of the target path to be updated robot on the target candidate new path is obtained. The efficiency reduction coefficient is based on the predicted avoidance coefficient of the target path to be updated robot on the target candidate new path and the predicted road congestion degree of the target candidate new path. Based on the efficiency reduction coefficient, the update priority of the target candidate new path by the target path-to-update robot is obtained.

9. The multi-robot path dynamic planning method according to claim 3, characterized in that, The efficiency reduction index for obtaining the target path to be updated in the path to be updated includes: The influence intensity of the target path robot to be updated on the existing target robots on the target path to be updated is obtained, and the influence intensity is obtained based on the average speed, maximum moving speed and heading angle of the target path robot to be updated, as well as the average speed and heading angle of the existing target robots. Based on the intensity of the influence, determine the overall intensity of the influence of the target path to be updated robot on the target path to be updated; Based on the combined influence intensity and the path disorder coefficient of the target path to be updated, the interference coefficient of the robot on the target path to be updated is obtained. Based on the interference coefficient and the road congestion level of the target path to be updated, the efficiency reduction index of the target path to be updated is obtained.

10. A multi-robot path dynamic planning system, characterized in that, The system includes a planning platform, which includes: A memory on which computer programs are stored; A processor for executing the computer program in the memory to implement the steps of the method according to any one of claims 1-9.

Images