A heterogeneous multi-robot cooperative scheduling method for dynamic task insertion

Abstract

The application discloses a kind of heterogeneous multi-robot cooperative scheduling methods for dynamic task insertion, belong to multi-robot cooperative scheduling technical field.It includes: responsibility area is divided based on heterogeneous state parameter to patrol inspection area;In responsibility area, last coverage timestamp and idleness of each node are updated in real time when executing coverage patrol;When there is discrete task, candidate robot is filtered based on hard constraint ability set, and optimal response node is determined by priority control multidimensional composite utility bidding;When response node is stripped from patrol network, its original responsibility area is defined as to-be-repaired hole area, and is divided into multiple hole sub-blocks according to connectivity, and is taken over by adjacent robot;After response node completes discrete task and re-joins patrol network, the idleness of each node is retained, so that robot restores coverage from the area with higher idleness.The application can cover patrol, respond to emergency task and repair the local area with low disturbance after response node is stripped.

Description

Technical Field

[0001] This invention relates to the field of multi-robot cooperative scheduling technology, and in particular to a heterogeneous multi-robot cooperative scheduling method for dynamic task insertion. Background Technology

[0002] With the widespread deployment of intelligent robots in areas such as park inspection, smart buildings, public services, and emergency response, collaborative task scheduling and coverage control of multi-robot systems have become key technologies for improving task efficiency and system reliability. In dynamic, congested environments with diverse task types, multi-robot systems not only need to undertake routine area coverage inspection tasks to maintain continuous awareness of environmental conditions, but also need to respond to sudden, discrete tasks such as abnormal alarms, temporary service requests, or emergency response at any time to ensure system operational safety and service timeliness.

[0003] Existing multi-robot collaborative scheduling methods typically handle routine coverage inspections and dynamic task responses separately. One type of method focuses on achieving regional coverage through spatial partitioning and path planning, but usually assumes that the set of inspection nodes and the robot fleet structure are relatively stable, making it difficult to handle coverage gaps created when response robots are removed from the inspection network. Another type of method focuses on the allocation of sudden tasks, usually selecting response robots based on distance, task capability, or a single utility metric, making it difficult to simultaneously consider heterogeneous capabilities, remaining power, current load, and task priority. Furthermore, when a response robot is removed from its original responsibility area, existing methods often require triggering a global responsibility area replanning, resulting in significant computational overhead and potentially disrupting the existing inspection rhythm. If historical coverage memory is not retained during the recovery phase, the system is prone to clearing historical coverage states and re-executing coverage planning, making it difficult to prioritize compensating areas with high idle rates or those that have not been inspected for a long time.

[0004] In summary, existing technologies still struggle to simultaneously address routine coverage inspections, rapid response to unexpected tasks, and low-disturbance local repairs after response nodes are removed, all under dynamic task insertion conditions. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a heterogeneous multi-robot cooperative scheduling method for dynamic task insertion, which can take into account normal coverage inspection, rapid response to sudden tasks, and low-disturbance local repair after the removal of response nodes under dynamic task insertion conditions.

[0006] To achieve the above objectives, the present invention is implemented using the following technical solution:

[0007] This invention provides a heterogeneous multi-robot cooperative scheduling method for dynamic task insertion, comprising:

[0008] Based on the heterogeneous state parameters of each robot in the multi-robot system, the comprehensive capability weight is calculated and the allocation cost is corrected. Based on the allocation cost, the responsibility area of ​​the inspectable nodes in the inspection network is divided to obtain the responsibility area of ​​each robot.

[0009] Within each robot's area of ​​responsibility, the node with the highest coverage score is selected as the next target node for coverage, and the following scheduling steps are executed cyclically:

[0010] The robot arrives at the next target node and completes a partial inspection, then updates the last coverage timestamp and idle time of the next target node.

[0011] When discrete tasks exist, the set of hard constraint capabilities, the set of soft constraint capabilities, and the task priority of the discrete tasks are extracted. Based on the set of hard constraint capabilities, a set of candidate robots is selected from the multi-robot system. Based on the set of soft constraint capabilities and the task priority, the multidimensional composite response utility is calculated. Each robot in the candidate robot set is comprehensively evaluated to obtain the optimal response node. The optimal response node is then removed from the inspection network.

[0012] The responsibility area corresponding to the optimal response node before stripping is taken as the void area to be repaired. The last coverage timestamp and idle degree of each node in the void area to be repaired are retained. Based on the spatial connectivity, the void area to be repaired is divided into multiple void sub-blocks. The robots that are in contact with the boundary of the responsibility area and the boundary of the void sub-block are taken as neighboring robots. A set of neighboring robots is constructed. Local secondary bidding is performed in the set of neighboring robots to determine the takeover robot of each void sub-block so as to take over the corresponding void sub-block.

[0013] After the optimal response node completes the discrete task and rejoins the inspection network, the recovery responsibility area is divided based on the latest heterogeneous state parameters of each robot. The last coverage timestamp and idle degree of each node are retained. The node with the largest idle degree in the recovery responsibility area is selected as the next coverage target node.

[0014] Optionally, the heterogeneous state parameters include current location, remaining battery power, average moving speed, effective sensing radius, and capability tag set.

[0015] Optionally, the comprehensive capability weights are represented as follows:

[0016] ;

[0017] In the formula, Represents robots The overall capability weight; All represent non-negative comprehensive ability weighting coefficients; They represent robots Normalized remaining battery power, normalized average moving speed, and normalized effective sensing radius;

[0018] The allocation cost is expressed as:

[0019] ;

[0020] In the formula, Represents robots To the node The cost of allocation; This represents the obstacle-constrained grid diagram corresponding to the inspection network. China Robot Current location To the node The shortest accessible path distance; Represents positive integers;

[0021] The responsibility area of ​​each robot is represented as follows:

[0022] ;

[0023] in, ;

[0024] ;

[0025] In the formula, They represent robots Responsibility area, robot The area of ​​responsibility; Represents the set of inspectable nodes; Represents a node The robot that belongs to you alone; Indicates the number of robots; Indicates from Selected The smallest robot number; This represents the empty set.

[0026] Optionally, the coverage score is expressed as:

[0027] ;

[0028] in, ;

[0029] In the formula, Represents robots For nodes At any moment Coverage score; All of these represent non-negative coverage score weighting coefficients; Represents a node At any moment Normalized idle time; This represents the obstacle-constrained grid diagram corresponding to the inspection network. China Robot At any moment Location To the node The normalized shortest path distance; Represents a node Normalized geometric coverage consistency; Represents a node At any moment The degree of free time; Represents a node The last overwrite timestamp;

[0030] The next target node for coverage is represented as:

[0031] ;

[0032] In the formula, Represents robots At any moment The next target node to be covered; Represents robots At any moment The set of candidate access nodes; Indicates from Selected The largest node.

[0033] Optionally, the candidate robot set is represented as:

[0034] ;

[0035] In the formula, Represents the set of candidate robots; Indicates the number of robots; Represents the set of hard constraint capabilities; Represents robots A set of capability tags; Represents robots The current remaining battery power; This indicates the preset minimum safe power threshold; Represents robots The current working status; Represents a set of fault or offline states;

[0036] Based on the set of soft constraints and task priorities, the multidimensional composite response utility is calculated, including:

[0037] Calculate the soft matching score based on the soft constraint capability set;

[0038] Calculate the estimated arrival time score of the robot to the discrete task location based on the estimated arrival time and the maximum estimated arrival time of the robot.

[0039] Based on the robot's remaining inspection load and discrete task load, calculate the robot's current total load, and based on the robot's current total load, calculate the robot's remaining load capacity score.

[0040] The multidimensional composite response utility is calculated based on task priority, soft matching score, expected arrival time score of robot to discrete task location, robot power score, and remaining carrying capacity score.

[0041] Optionally, the soft-match score is represented as:

[0042] ;

[0043] In the formula, Represents robots Soft match score; Represents the set of soft constraint capabilities; Labels indicating soft constraint capabilities Importance weights; Represents positive integers;

[0044] The estimated arrival time of the robot at the discrete task location is expressed as:

[0045] ;

[0046] In the formula, Represents robots The estimated arrival time to the discrete task location; This represents the obstacle-constrained grid diagram corresponding to the inspection network. China Robot At any moment Location To discrete task location The shortest accessible path distance; Represents robots Average moving speed;

[0047] The maximum estimated arrival time is expressed as:

[0048] ;

[0049] In the formula, Indicates the maximum estimated arrival time; This indicates taking the maximum value;

[0050] The estimated arrival time score of the robot to the discrete task location is expressed as:

[0051] ;

[0052] In the formula, Represents robots The estimated arrival time score for reaching the discrete task location;

[0053] The remaining inspection load of the robot is represented as follows:

[0054] ;

[0055] In the formula, Represents robots The remaining load of the inspection; Represents robots Inspection status indicator variables; Represents robots The size of the remaining area to be inspected within the current area of ​​responsibility; Represents robots Environmental reduction factor; Represents robots Effective inspection width;

[0056] The discrete task load of the robot is represented as:

[0057] ;

[0058] In the formula, Represents robots Discrete task load; Represents robots The discrete tasks already undertaken The remaining execution time; Represents robots The number of discrete tasks currently undertaken;

[0059] The robot's current total load is expressed as:

[0060] ;

[0061] In the formula, Represents robots The current total load;

[0062] The remaining load capacity score of the robot is expressed as:

[0063] ;

[0064] In the formula, Represents robots The remaining carrying capacity score; Represents robots The maximum load that can be handled within the preset scheduling period;

[0065] The robot's battery score is represented as follows:

[0066] ;

[0067] In the formula, Represents robots Battery score; Represents robots Rated power capacity;

[0068] The multidimensional composite response utility is expressed as follows:

[0069] ;

[0070] In the formula, Represents robots Compared to the multidimensional composite response utility of discrete tasks; All indicate task priority. Relevant weighting functions;

[0071] The optimal response node is represented as:

[0072] ;

[0073] In the formula, Indicates the node with the best response; Indicates from Selected The largest robot number.

[0074] Optionally, the hollow sub-block is represented as:

[0075] , ;

[0076] In the formula, Indicates the area of ​​cavity to be repaired; Indicates the number of empty sub-blocks; Representing empty sub-blocks Hollow sub-blocks ; This represents the empty set.

[0077] Optionally, the set of neighboring robots is represented as:

[0078] ;

[0079] In the formula, Indicates a group of nearby robots; Indicates the number of robots; Indicates the node with the best response; Represents robots The current boundaries of the area of ​​responsibility; Represents a hollow sub-block The boundary; Indicates the empty set; Represents robots The current remaining battery power; This indicates the preset minimum safe power threshold; Represents robots The current working status; This represents a set of fault or offline states.

[0080] Optionally, a local secondary bidding process is performed within the neighboring robot set to determine the takeover robot for each empty sub-block, including:

[0081] Based on the maximum path distance from the neighboring robot to the empty sub-block, calculate the adjacency score of the neighboring robot to the empty sub-block;

[0082] The average free space of the empty sub-blocks is normalized to obtain the normalized average free space of the empty sub-blocks;

[0083] The filling effect of neighboring robots on empty sub-blocks is calculated based on the adjacency score of neighboring robots on empty sub-blocks and the normalized average vacancy degree of empty sub-blocks.

[0084] Based on the filling effect of neighboring robots on the hollow sub-blocks, the takeover robot for each hollow sub-block is determined;

[0085] The maximum path distance from the neighboring robot to the hollow sub-block is represented as:

[0086] ;

[0087] In the formula, Indicates the current nearest robot to the empty sub-block. The maximum path distance; This represents the obstacle-constrained grid diagram corresponding to the inspection network. China Robot At any moment Location To the hollow sub-block Representative node The shortest accessible path distance; Indicates a group of nearby robots; This indicates taking the maximum value;

[0088] The adjacency score of the neighboring robot for the empty sub-block is represented as follows:

[0089] ;

[0090] In the formula, Represents robots Hollow sub-blocks Adjacency score; Represents positive integers;

[0091] The average freeness of the hollow sub-block is expressed as:

[0092] ;

[0093] In the formula, Represents a hollow sub-block At any moment Average idle time; Represents a hollow sub-block The number of nodes within; Represents a hollow sub-block ; Represents a node At any moment The degree of free time;

[0094] The normalized average free space of the empty sub-block is expressed as:

[0095] ;

[0096] In the formula, Represents a hollow sub-block At any moment Normalized average idle time; Indicates the number of empty sub-blocks;

[0097] The effect of the neighboring robot in filling the void sub-block is expressed as:

[0098] ;

[0099] In the formula, Represents robots Hollow sub-blocks The complementary effect; All of these represent non-negative filler utility weight coefficients; Represents robots The remaining carrying capacity score; Represents robots Battery score;

[0100] The takeover robot for each hollow sub-block is represented as follows:

[0101] ;

[0102] In the formula, Represents a hollow sub-block The takeover robot; Indicates from Selected The largest robot number.

[0103] Optionally, the node with the highest idle degree within the recovery responsibility area is represented as:

[0104] ;

[0105] In the formula, Represents robots At any moment Restore the node with the highest idle rate within the responsibility area; Indicates from Selected from The largest node; Represents robots The area of ​​responsibility for restoration; Represents a node At any moment The degree of free time.

[0106] Compared with the prior art, the beneficial effects achieved by the present invention are as follows:

[0107] This invention constructs an online collaborative scheduling framework that couples routine coverage inspection, event response, and coverage restoration under dynamic task insertion conditions. It constructs comprehensive capability weights based on heterogeneous state parameters and divides the inspection environment into weighted responsibility zones, ensuring that the inspection workload matches the robot's capabilities, thus improving the overall rationality of task allocation and coverage efficiency. The multi-dimensional composite utility bidding based on priority control helps reduce the risk of high-priority tasks being assigned to robots with incompatible capabilities, insufficient timeliness, or unsatisfactory operating states, improving the rationality and timeliness of responses to sudden discrete tasks. Finally, it utilizes a mechanism based on the division of void areas to be repaired and a local secondary bidding mechanism for void sub-blocks. It can limit the replacement range caused by the event response to the range of neighboring robots in contact with the empty sub-block, reduce the disturbance caused by the immediate triggering of global responsibility area replanning during the response node stripping, and perform restorative responsibility area re-division after the response node returns; the recovery mechanism based on idle degree memory enables the robot to prioritize the compensation of areas with high idle degree during the event response after the response node returns, which is conducive to improving the coverage continuity and load balancing of the system under long-term operation, thereby realizing the unified coordination of heterogeneous multi-robot system in normal coverage inspection, emergency task response and low-disturbance local repair after response node stripping in dynamic task environment. Attached Figure Description

[0108] Figure 1 This is a flowchart illustrating a heterogeneous multi-robot cooperative scheduling method for dynamic task insertion provided in an embodiment of the present invention.

[0109] Figure 2 This is a schematic diagram of the responsibility area of ​​each robot provided in an embodiment of the present invention;

[0110] Figure 3 This is a schematic diagram of the coverage path within the responsibility area of ​​each robot provided in the embodiments of the present invention;

[0111] Figure 4 This is a schematic diagram illustrating the results of discrete task triggering and optimal response node selection provided in an embodiment of the present invention;

[0112] Figure 5 A schematic diagram of the void region to be repaired formed after the optimal response node is stripped, as provided in an embodiment of the present invention.

[0113] Figure 6 This is a schematic diagram of a hollow sub-block and a neighboring robot provided in an embodiment of the present invention;

[0114] Figure 7 This is a schematic diagram of the result of the takeover of the hollow sub-block by the takeover robot provided in an embodiment of the present invention;

[0115] Figure 8 This is a schematic diagram of the recovery responsibility area results provided in an embodiment of the present invention;

[0116] Figure 9 A schematic diagram of the priority coverage target node provided in an embodiment of the present invention;

[0117] Figure 10 A schematic diagram showing the comparison of system coverage between the method of the present invention and the nearest-principle task allocation combined with global replanning method, provided for embodiments of the present invention;

[0118] Figure 11 A schematic diagram showing the comparison curves of the method of the present invention and the global replanning method combined with the nearest-principle task allocation in the maximum region idleness, provided for embodiments of the present invention. Detailed Implementation

[0119] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations thereof. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0120] The term "and / or" simply describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. Additionally, the character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0121] Example 1

[0122] This embodiment introduces a heterogeneous multi-robot cooperative scheduling method for dynamic task insertion, including:

[0123] Based on the heterogeneous state parameters of each robot in the multi-robot system, the comprehensive capability weight is calculated and the allocation cost is corrected. Based on the allocation cost, the responsibility area of ​​the inspectable nodes in the inspection network is divided to obtain the responsibility area of ​​each robot.

[0124] Within each robot's area of ​​responsibility, the node with the highest coverage score is selected as the next target node for coverage, and the following scheduling steps are executed cyclically:

[0125] The robot arrives at the next target node and completes a partial inspection, then updates the last coverage timestamp and idle time of the next target node.

[0126] When discrete tasks exist, the set of hard constraint capabilities, the set of soft constraint capabilities, and the task priority of the discrete tasks are extracted. Based on the set of hard constraint capabilities, a set of candidate robots is selected from the multi-robot system. Based on the set of soft constraint capabilities and the task priority, the multidimensional composite response utility is calculated. Each robot in the candidate robot set is comprehensively evaluated to obtain the optimal response node. The optimal response node is then removed from the inspection network.

[0127] The responsibility area corresponding to the optimal response node before stripping is taken as the void area to be repaired. The last coverage timestamp and idle degree of each node in the void area to be repaired are retained. Based on the spatial connectivity, the void area to be repaired is divided into multiple void sub-blocks. The robots that are in contact with the boundary of the responsibility area and the boundary of the void sub-block are taken as neighboring robots. A set of neighboring robots is constructed. Local secondary bidding is performed in the set of neighboring robots to determine the takeover robot of each void sub-block so as to take over the corresponding void sub-block.

[0128] After the optimal response node completes the discrete task and rejoins the inspection network, the recovery responsibility area is divided based on the latest heterogeneous state parameters of each robot. The last coverage timestamp and idle degree of each node are retained. The node with the largest idle degree in the recovery responsibility area is selected as the next coverage target node.

[0129] In this embodiment, during the routine coverage inspection task performed by a heterogeneous multi-robot system, a collaborative scheduling method is proposed to address the problems of unreasonable selection of response robots caused by the dynamic insertion of sudden discrete tasks, the formation of coverage gaps in the original responsibility area after the response node is stripped, and the large disturbance of global responsibility area replanning. This method can uniformly consider heterogeneous capabilities, task priorities, remaining power, expected arrival time, and remaining carrying capacity. This method aims to improve the response time of sudden discrete tasks while maintaining the continuity of coverage inspection and to achieve low-disturbance local repair of the gaps to be repaired.

[0130] Example 2

[0131] Based on Example 1, such as Figure 1 As shown in the figure, this embodiment introduces a heterogeneous multi-robot cooperative scheduling method for dynamic task insertion, including the following steps:

[0132] Step 1: Obtain heterogeneous state parameters and perform weighted responsibility region division, specifically as follows:

[0133] By coupling heterogeneous state perception with spatial partitioning, robots with higher battery power, faster movement speed and larger perception radius can get closer to inspectable nodes in an equivalent sense, thereby taking on a relatively larger area of ​​responsibility and improving the overall rationality of division of labor.

[0134] Obtain heterogeneous state parameters for each robot in a multi-robot system. These heterogeneous state parameters include current position, remaining battery power, average moving speed, effective perception radius, and capability tag set.

[0135] Discretize the inspection network into a grid graph with obstacle constraints. ,in, Represents the set of all grid nodes. Let denote the set of walkable edges, and denot the set of inspectable nodes as . .robot At any moment The state vector is denoted as ,in, They represent robots At any moment Location, remaining battery power, and operating status; They represent robots Average movement speed, effective perception radius, and capability tag set.

[0136] The remaining battery power, average moving speed, and effective sensing radius of each robot are normalized to obtain the normalized remaining battery power of each robot. Normalized average moving speed With normalized effective sensing radius ,in, , This refers to the number of robots.

[0137] Based on normalized remaining power Normalized average moving speed With normalized effective sensing radius Calculate the overall capability weights for the robot. Comprehensive ability weight Represented as:

[0138] ;

[0139] In the formula, All represent non-negative comprehensive ability weight coefficients, and satisfy the following conditions: , These are used to characterize the relative importance of battery power, speed, and sensing radius in the division of the area of ​​responsibility, by setting... This makes the overall ability weighting This constitutes a convex combination of normalized heterogeneous parameters, thereby keeping the weights of the comprehensive capabilities of different robots within a unified dimension and a comparable range, which is convenient for subsequent use as a correction factor for allocation costs.

[0140] Using comprehensive capability weights as a correction factor for allocation costs, the robot adjusts the allocation costs. To the node Allocation cost Represented as:

[0141] ;

[0142] In the formula, This represents the obstacle-constrained grid diagram corresponding to the inspection network. China Robot Current location To the node The shortest accessible path distance; To represent positive numbers, prevent the denominator from being zero.

[0143] Based on the allocation cost, the responsibility areas of the inspectable nodes in the inspection network are divided to obtain the responsibility areas of each robot:

[0144] The unique robot that owns a node is determined based on the allocation cost. The only robot Represented as:

[0145] ;

[0146] In the formula, Indicates from Selected The smallest robot number; when multiple robots have the same allocation cost, the node is determined according to a preset unique attribution rule. The robot that belongs to.

[0147] robot Responsibility Area Represented as:

[0148] ;

[0149] Therefore, we obtain The responsibility area division results for the robot are as follows, and the responsibility area division results satisfy:

[0150] ;

[0151] In the formula, Represents robots The area of ​​responsibility; This represents the empty set.

[0152] This division method retains the geometric intuition of path distance-based methods while introducing heterogeneous capability differences between robots, allowing robots with stronger capabilities to bear a lower cost in equivalent allocation and thus assume a relatively larger inspection responsibility area.

[0153] Step 2: Based on the coverage score area, perform coverage inspection and update the idle rate, specifically as follows:

[0154] Each robot performs area-based inspections within its designated responsibility zone.

[0155] For robots Responsibility Area Any candidate access node within Define nodes At any moment idle time For a moment With nodes Last overwrite timestamp The difference, i.e., the node At any moment idle time Represented as:

[0156] .

[0157] The higher the idle time, the longer the node has not been inspected, and the higher its priority for coverage updates.

[0158] Based on the normalized idleness, normalized shortest path distance, and normalized geometric coverage consistency of nodes, a coverage scoring function is constructed for the robot. For nodes At any moment Coverage rating Represented as:

[0159] ;

[0160] In the formula, All of these represent non-negative coverage score weighting coefficients; Represents a node At any moment The normalized idle rate is used to improve the coverage priority of long-unaccessed areas; This represents the obstacle-constrained grid diagram corresponding to the inspection network. China Robot At any moment Location To the node The normalized traversable shortest path distance is used to suppress the robot from frequently jumping to too far nodes, which would cause the coverage path to be overly discrete. Represents a node Normalized geometric coverage consistency is used to maintain the basic geometric coverage structure.

[0161] Within each robot's area of ​​responsibility, the node with the highest coverage score is selected as the next target node for coverage. At any moment The set of candidate access nodes In non-empty time, the robot At any moment The next target node to be covered Represented as:

[0162] ;

[0163] In the formula, Represents robots At any moment The set of candidate access nodes; Indicates from Selected The largest node.

[0164] When two or more candidate access nodes have the same coverage score, a unique next coverage target node is determined according to the preset target selection rules.

[0165] The robot reaches the next target node and completes a partial inspection, updating the coverage status and last coverage timestamp of the next target node. In addition, the idleness of the remaining unvisited nodes naturally increases over time, thus forming a time-accumulated inspection memory process within the region, providing fine-grained spatial state priors for void repair and restoration in subsequent steps.

[0166] Step 3: Detecting sudden tasks and determining the optimal response node, specifically:

[0167] The multi-robot system pre-defines a task capability mapping dictionary and maintains a set of capability labels for each robot; among them, the robot... The set of capability tags is .

[0168] When there is a sudden discrete task At that time, a sudden discrete task After parsing the task capability mapping dictionary, the set of hard constraint capabilities for discrete tasks is extracted. Soft constraint capability set and task priority Hard constraint capability set Used to determine whether a robot is capable of performing sudden, discrete tasks. Qualifications, set of soft constraint capabilities Used to characterize robots and sudden discrete tasks The degree of compatibility; task priority Used to adjust the weight of each component score in subsequent multidimensional composite utility bidding; operational constraints include minimum safe power constraints and non-faulty online state constraints.

[0169] Based on hard constraint capability set Based on operational constraints, a candidate robot set is selected from the multi-robot system. Represented as:

[0170] ;

[0171] In the formula, Represents robots The current remaining battery power; This indicates the preset minimum safe power threshold; Represents robots The current working status; This represents the set of fault states or offline states. This filtering process is used to eliminate robots that do not meet the hard constraints of the task or operational constraints, reducing the unnecessary comparison overhead of subsequent multidimensional composite utility calculations.

[0172] Based on soft constraint capability set and task priority The multidimensional composite response utility is calculated, taking into account the expected arrival time, soft constraint matching degree, remaining power, and remaining carrying capacity.

[0173] Calculate the soft matching score based on the soft constraint capability set;

[0174] Calculate the estimated arrival time score of the robot to the discrete task location based on the estimated arrival time and the maximum estimated arrival time of the robot.

[0175] Based on the robot's remaining inspection load and discrete task load, calculate the robot's current total load, and based on the robot's current total load, calculate the robot's remaining load capacity score.

[0176] The multidimensional composite response utility is calculated based on task priority, soft matching score, expected arrival time score of robot to discrete task location, robot power score, and remaining carrying capacity score.

[0177] soft constraint capability set The fit assessment is achieved through soft-match scoring, and the robot soft match score Represented as:

[0178] ;

[0179] In the formula, Labels indicating soft constraint capabilities Importance weighting; robots soft match score The larger the size, the more likely it is to be a robot. For sudden discrete tasks The higher the degree of compatibility.

[0180] robot Remaining load during inspection Represented as:

[0181] ;

[0182] In the formula, Represents robots The inspection status indicator variable, Represents robots Still within the routine inspection network, Represents robots It has been separated from the routine inspection network and is now responsible for handling incident response tasks related to discrete tasks. Represents robots The size of the remaining area to be inspected within the current area of ​​responsibility; Represents robots Environmental reduction factor; Represents robots Effective inspection width;

[0183] robot Discrete task load Represented as:

[0184] ;

[0185] In the formula, Represents robots The discrete tasks already undertaken The remaining execution time; Represents robots The number of discrete tasks currently undertaken;

[0186] For candidate robot set Any robot in , will the robot Current total load Defined as remaining load during inspection With discrete task load The sum is:

[0187] .

[0188] This leads to the robot. Remaining carrying capacity score , is represented as:

[0189] ;

[0190] In the formula, Represents robots Maximum load capacity within a preset scheduling period; robot Remaining carrying capacity score The larger the size, the more likely it is to be a robot. The more load margin you have when taking over newly added, sudden, discrete tasks or empty sub-blocks, the better.

[0191] The estimated arrival time score is constructed based on the length of the passable path and the average moving speed, while the battery score is constructed based on the relationship between the current remaining battery power and the minimum safe battery power threshold. Estimated arrival time score for reaching discrete task locations With robots Battery score They are respectively:

[0192] For candidate robot set Any robot in Define robot Reaching the location of a sudden discrete mission Estimated arrival time Represented as:

[0193] ;

[0194] Candidate robot set Maximum estimated arrival time Represented as:

[0195] ;

[0196] Then the robot Estimated arrival time score for reaching discrete task locations Represented as:

[0197] ;

[0198] In the formula, This represents the obstacle-constrained grid diagram corresponding to the inspection network. China Robot At any moment Location To discrete task location The shortest accessible path distance; Represents robots Average moving speed; This indicates taking the maximum value.

[0199] robot Battery score Represented as:

[0200] ;

[0201] In the formula, Represents robots Rated power capacity.

[0202] For sudden discrete tasks Candidate robot set China Robot Multidimensional composite response utility relative to discrete tasks Represented as:

[0203] ;

[0204] In the formula, All indicate task priority. The relevant weight function, and satisfies .

[0205] The weighting function dynamically adjusts with task priority, so that the higher the task priority, the greater the weight corresponding to the timeliness score and soft matching score; and the lower the task priority, the greater the weight corresponding to the remaining carrying capacity score and battery power score. This reduces the risk of high-priority tasks being assigned to robots with incompatible capabilities or insufficient timeliness, and reduces the risk of low-priority tasks crowding out the resources of robots with high loads or low battery power.

[0206] A comprehensive evaluation is performed on each robot in the candidate robot set to obtain the optimal response node. When not empty, the optimal response node Represented as:

[0207] ;

[0208] In the formula, Indicates from Selected The largest robot number.

[0209] When two or more candidate robots have the same multidimensional composite response utility, a unique optimal response node is determined according to the preset response node selection rules.

[0210] The optimal response node Switch to responsive state and disconnect from the inspection network.

[0211] Step 4: Perform partial takeover and restorative repartitioning of the empty character block, specifically as follows:

[0212] When the optimal response node After being separated from the inspection network, the optimal response node will be... The corresponding responsibility area before stripping As a cavity area to be repaired ,Right now And preserve the areas with cavities to be repaired. The last coverage timestamp and idle information of each node ensure that the coverage memory in the hole area to be repaired is not cleared during the event response execution.

[0213] According to the area of ​​the cavity to be repaired The spatial connectivity of the internal nodes divides the area of ​​the cavity to be repaired into... A number of non-overlapping empty sub-blocks, namely:

[0214] , ;

[0215] In the formula, Indicates the number of empty sub-blocks; Representing empty sub-blocks Hollow sub-blocks Each hollow sub-block contains nodes that satisfy spatial connectivity. The hollow sub-block serves as the smallest allocation unit for subsequent local secondary bidding. Through sub-blocking, the hollow areas to be repaired that are complex in shape or large in scale are divided into multiple local task units, reducing the load fluctuations caused by a single neighboring robot taking over the entire task.

[0216] For each empty sub-block, robots whose responsibility area boundary contacts the empty sub-block boundary are considered neighboring robots, and a set of neighboring robots is constructed. Neighboring robots satisfy operational constraints and are non-responsive robots. Represented as:

[0217] ;

[0218] In the formula, Represents robots The current boundaries of the area of ​​responsibility; Represents a hollow sub-block The boundary; Represents robots The current remaining battery power; Represents robots The current working state. This construction method restricts the robots participating in the replacement to the range of robots adjacent to the empty sub-block and satisfying the operational constraints, thereby reducing the disturbance of the remote responsibility area to the event response.

[0219] A local secondary bidding process is performed within the set of neighboring robots to determine the takeover robot for each void sub-block. One or more neighboring robots then take over the corresponding void sub-block. During the execution of discrete tasks at the optimal response node, the system does not trigger global responsibility region replanning. Instead, neighboring robots perform local takeover of the void area to be repaired.

[0220] Based on the maximum path distance from the neighboring robot to the empty sub-block, calculate the adjacency score of the neighboring robot to the empty sub-block;

[0221] The average free space of the empty sub-blocks is normalized to obtain the normalized average free space of the empty sub-blocks;

[0222] The filling effect of neighboring robots on empty sub-blocks is calculated based on the adjacency score of neighboring robots on empty sub-blocks and the normalized average vacancy degree of empty sub-blocks.

[0223] Based on the filling effect of neighboring robots on the empty sub-blocks, the takeover robot for each empty sub-block is determined.

[0224] robot Hollow sub-blocks Adjacency score Hollow sub-blocks At any moment Average idleness and hollow sub-blocks At any moment Normalized average idle time They are respectively:

[0225] When neighboring robots gather When not empty, the current nearest robot to the empty sub-block Maximum path distance Represented as:

[0226] ;

[0227] In the formula, This represents the obstacle-constrained grid diagram corresponding to the inspection network. China Robot At any moment Location To the hollow sub-block Representative node The shortest accessible path distance;

[0228] robot Hollow sub-blocks Adjacency score Represented as:

[0229] ;

[0230] Hollow sub-block At any moment Average idleness Represented as:

[0231] ;

[0232] In the formula, Represents a hollow sub-block The number of nodes within;

[0233] Hollow sub-block At any moment Normalized average idle time Represented as:

[0234] ;

[0235] robot Hollow sub-blocks Adjacency score The larger the size, the more likely it is to be a robot. The closer to the hollow sub-block, the more hollow sub-blocks... At any moment Average idleness The larger the value, the longer the entire empty sub-block has not been inspected.

[0236] For any neighboring robot Hollow sub-blocks The complementary effect Represented as:

[0237] ;

[0238] In the formula, All represent non-negative complement utility weight coefficients, and satisfy the following conditions: ; Represents robots The remaining carrying capacity score; Represents robots The battery score.

[0239] Hollow sub-block takeover robot Represented as:

[0240] ;

[0241] In the formula, Indicates from Selected The largest robot number.

[0242] When two or more neighboring robots have the same filling effect, a unique takeover robot is determined according to the preset filling selection rules; different void sub-blocks are subject to local secondary bidding. After any neighboring robot wins the bid, the system will temporarily add the corresponding void sub-block to the robot's takeover task and update the robot's remaining load capacity score, so that one or more neighboring robots can take over the void area to be repaired in blocks.

[0243] When the optimal response node After completing a sudden discrete task and rejoining the inspection network, the system switches its working state from response state back to normal inspection state. Instead of clearing the last coverage timestamp and idle time of each node, the system performs restorative responsibility area repartitioning based on the latest heterogeneous state parameters of each robot at the moment the optimal response node completed the discrete task, thus re-obtaining the restorative responsibility area for each robot. During this repartitioning process, the system does not clear the last coverage timestamp and idle time of each node; it only updates the responsibility area structure.

[0244] Within the recovery responsibility area, select the node with the highest idle rate as the next target node for coverage:

[0245] During the recovery phase, when the robot Restoration responsibility area In non-empty time, the robot At any moment Restore the node with the highest idle rate within the responsibility area. Represented as:

[0246] ;

[0247] In the formula, Indicates from Selected from The largest node.

[0248] When two or more nodes have the same level of idleness, a unique next coverage target node is determined according to the preset recovery target selection rules. This allows each robot to prioritize accessing the node with the highest level of idleness in its new responsibility area during the recovery phase, and ensures that areas with higher idleness during the stripping of the optimal response node are compensated first, avoiding the need to restart the inspection after clearing the historical coverage memory during the recovery phase.

[0249] After determining the next target node for coverage, return to step two. The robot will then reach the next target node for coverage, complete a local inspection, and update the last coverage timestamp and idle time of the next target node.

[0250] The scheduling is executed cyclically. Through the above process, the division of responsibility areas, coverage within areas, event response scheduling, and hole repair and recovery form a closed-loop collaborative scheduling process.

[0251] Example 3

[0252] Based on Example 2, this example presents an experimental example of a heterogeneous multi-robot cooperative scheduling method for dynamic task insertion:

[0253] Step 1: Obtain the heterogeneous state parameters of each robot in the multi-robot system. Construct a comprehensive capability weight based on these parameters, and use this weight as a correction factor for the allocation cost. Divide the responsibility areas of the inspectable nodes in the inspection network to obtain the responsibility areas for each robot. This step couples heterogeneous state perception with spatial partitioning, allowing robots with higher battery power, faster movement speed, and larger perception radius to be closer to the inspectable nodes in an equivalent sense, thus assuming a relatively larger responsibility area and improving the overall rationality of task allocation. Specifically, this includes:

[0254] First, the heterogeneous state parameters of the robot are obtained, and the inspection network is discretized into a grid diagram with obstacle constraints. ,robot At any moment The state vector is denoted as .

[0255] Then, the heterogeneous state parameters of the robots are normalized by normalizing the remaining battery power, average moving speed, and effective sensing radius of each robot to the maximum and minimum values ​​of the current robot group. This eliminates the dimensional differences between different physical quantities.

[0256] Secondly, a comprehensive capability weighting is constructed, and based on the normalized results, a robot is built through weighted summation. Comprehensive ability weight .

[0257] Then, the weighted responsibility area is divided for any inspectable node. In order to allocate costs For robots To the node The equivalent distance, and the nodes The robot is assigned to the robot that minimizes its allocation cost, thus obtaining the robot. Responsibility Area Once all inspectable nodes have been assigned, the inspection environment is divided into: Each area of ​​responsibility is non-overlapping, and .

[0258] Step Two: Perform coverage inspection within the responsibility area of ​​each robot. Based on the normalized idleness, normalized shortest path distance, and normalized geometric coverage consistency of nodes, construct a coverage scoring function, select the node with the highest coverage score as the next coverage target, and update the coverage status, last coverage timestamp, and idleness of each node in real time. Specifically, this includes:

[0259] First, extract the coverable nodes within the responsibility area for each robot. In its area of ​​responsibility Inside, based on the current location Accessibility constraints and maximum single-step access range are used to determine the timing. The set of candidate access nodes .

[0260] Secondly, generate coverage paths within the region based on the coverage scoring function. The node with the highest coverage score in the candidate visit node set is selected as the robot. The next coverage target And proceed to the target using the shortest possible path.

[0261] Finally, update the coverage status, last coverage timestamp, and idle time when the robot... Reach the target node After completing the partial inspection, update the most recent access timestamp of the node to... idle time The idle time of unvisited nodes naturally increases over time. Meanwhile, the system at any given moment... System coverage is defined as ,in, Indicates at time The number of nodes that have been effectively inspected. (Time) System coverage With time node Maximum regional idleness It is also used as an indicator to evaluate the quality of routine inspections.

[0262] Step 3: When the system detects a sudden discrete task, it determines the optimal response node by resolving task constraints, filtering candidate sets, and performing multi-dimensional composite utility bidding. Specifically, this includes:

[0263] First, task analysis: sudden discrete tasks. Modeled as a quadruple ,in The location where the mission occurred. As a task priority, and These are the hard constraint capability set and the soft constraint capability set, respectively. After receiving a sudden task, the system first extracts all the information contained in the above four-tuples according to the preset task capability mapping dictionary.

[0264] Then, candidate set selection is performed, based on the set of hard constraint capabilities, the minimum safe power threshold, and basic online state constraints, to obtain a set of candidate robots that can participate in the task τ competition. This screening process prioritizes excluding robots that are clearly not qualified to perform the task, reducing the overhead of subsequent invalid comparisons.

[0265] Secondly, a two-layer capability matching process involving both hard and soft constraints is employed. Building upon the hard constraint screening, a weighted hit rate evaluation is performed on the soft constraint capability set to obtain a soft matching score. The higher the soft-match score, the better the robot. For sudden discrete tasks The higher the degree of capability compatibility.

[0266] Then, the scores for each sub-item are calculated along with the current load, and the expected arrival time score is calculated for each robot in the candidate robot set. Battery Score and remaining carrying capacity score Among them, the remaining load capacity score is determined by the robot's current total load. With maximum load capacity The comparison shows that the remaining load of the inspection and the discrete task load are respectively... and Give the inspection status indicator variable. Used to switch the load source between routine inspection and incident response states, so that the two types of loads can be accumulated under a unified time dimension.

[0267] Finally, response nodes are selected based on the overall utility, and a multidimensional composite response utility is constructed. And through a weight function related to task priority. to The relative importance of each sub-item is dynamically adjusted, so that high-priority tasks place greater emphasis on estimated arrival time and capacity matching, while medium- and low-priority tasks appropriately increase the weight given to load balancing and power security; ultimately, through Define the task The optimal response node is identified, and its working state is switched from normal inspection state to discrete task response state.

[0268] Step 4: After the optimal response node is removed from the inspection network, the hole repair and recovery process is executed, which specifically includes:

[0269] First, define the void region and place the optimal response node. The original area of ​​responsibility before stripping The whole area is defined as the cavity to be repaired. Instead of using only the local area around its current location as a hole, this avoids missing edge nodes of the responsibility area during event response.

[0270] Then, the idleness is calculated and retained. During the event response execution, the system does not clear the last cover timestamp and idleness of each grid in the area to be repaired, but allows its idleness to continue to accumulate over time, thereby retaining fine-grained spatial state priors for subsequent filling and restoration processes.

[0271] Secondly, the hollow sub-blocks are divided according to the hollow areas to be repaired. The spatial connectivity of the inner grid is decomposed into A number of non-overlapping connected sub-blocks There is no common grid between different sub-blocks; through sub-blocking, the hollow areas with complex shapes and large scale spans are reduced to multiple local task units with controllable scales, which are easy for different neighboring robots to take over in parallel.

[0272] Then, construct a neighboring robot set for each empty sub-block. Only within the range of robots whose responsibility area boundary contacts the boundary of the void sub-block, a neighboring robot set is constructed. This strictly limits the robots participating in the replacement to the adjacent topology range, effectively controlling the disturbance radius caused by the event response.

[0273] Next, a local secondary bidding process is conducted at the sub-block level, where the fill utility is independently calculated for each empty sub-block within the neighboring robot set. And take the one with the greatest effect. As the takeover robot for this sub-block.

[0274] Then, the multi-robot block takeover allows different neighboring robots to win bids for different void sub-blocks. The winning neighboring robot temporarily incorporates the void sub-block into its current responsibility area and participates in the coverage inspection. Its working state is switched from normal inspection state to void repair state, thereby realizing the multi-robot block takeover of the void area to be repaired.

[0275] Next, the robot returns to its previous state, and the optimal response node... After completing the sudden discrete task, its working status is switched from discrete task response state back to normal inspection state, and a return signal is sent to the system.

[0276] Finally, the system globally restores and retains the idle time memory. After receiving the regression signal, based on the latest heterogeneous state parameters of each robot, it re-executes the global responsibility area division according to the responsibility area division method in step one to obtain the responsibility area for the recovery phase, i.e., the recovery responsibility area. During the re-division process, the system does not clear the last cover timestamp and idle time of each grid cell, but only updates the responsibility area structure. During the recovery phase, the robot... At the moment The node with the highest idle time within the recovery area is defined as... This means that the system prioritizes accessing the grid cells with the highest free space within its new responsibility area. Thus, the system naturally transitions from the hole repair phase back to the routine inspection phase. The four parts—responsibility area division, regional coverage, event response scheduling, and hole repair and recovery—form a closed-loop structure that connects seamlessly.

[0277] The effectiveness of the method of the present invention is further illustrated below through simulation experimental data. A multi-robot collaborative inspection simulation model containing four heterogeneous mobile robots was built. The simulation environment is a typical indoor inspection environment with room partitions, corridors and obstacle areas, which is discretized into a grid map with obstacle constraints; Figures 2 to 9 The division of responsibility areas, event response, hole takeover, and coverage restoration processes at each stage are illustrated using grid coordinates. Each robot differs in its capability tag set, remaining battery power, average movement speed, and effective perception radius. The capability tag set is pre-configured based on task types such as inspection, handling, voice interaction, anomaly detection, or emergency response. Sudden discrete tasks can be inserted during simulation at preset times or through random arrival processes. Task priorities include high, medium, and low levels. Both hard-constraint and soft-constraint capability sets are given according to a preset task capability mapping dictionary.

[0278] like Figure 2 As shown, during the routine inspection phase, the method of this invention divides the inspection environment into weighted responsibility areas based on comprehensive capability weights, allowing robots with stronger capabilities to assume relatively larger responsibility areas, and as... Figure 3As shown, each robot generates a coverage path within its corresponding responsibility area according to the coverage scoring function.

[0279] like Figure 4 As shown, when a sudden discrete task arrives, the method of this invention determines the optimal response node based on the candidate robot set and multi-dimensional composite utility bidding, enabling the optimal response node to travel along the response path to the location of the sudden task; as... Figure 5 As shown, after the response node is removed from the inspection network, its original responsibility area before removal is retained as a void area to be repaired.

[0280] like Figure 6 As shown, during the optimal response node stripping process, the void region to be repaired is divided into void sub-blocks according to spatial connectivity, such as... Figure 7 As shown, each empty sub-block is taken over by neighboring robots within the adjacent topology range through local secondary bidding, thereby avoiding immediate triggering of global responsibility area replanning during event response.

[0281] like Figure 8 As shown, after the response node completes the event task and returns, the system re-obtains the recovery phase responsibility area based on the latest heterogeneous state parameters of each robot, and as follows... Figure 9 As shown, the memory of the vacancy level of each grid is retained, so that the robot prioritizes covering the grids with higher vacancy levels in its new responsibility area.

[0282] like Figure 10 and Figure 11 As shown, compared to the nearest-principle task allocation + global replanning method, the method of this invention can maintain a higher system coverage after task insertion. and the maximum area idleness The system is kept at a low level; after the response node returns, the method of the present invention can recover to the normal inspection level more quickly, thus demonstrating that it can reduce the impact of dynamic task insertion on coverage continuity.

[0283] exist Figures 2 to 9 middle, , , , These represent the four heterogeneous robots participating in the collaborative inspection; Indicates the weight of the robot's overall capabilities; Represents robots The area of ​​responsibility; Indicates the location of a sudden, discrete task; This represents the set of candidate robots corresponding to sudden discrete tasks. This represents the void area to be repaired after the optimal response robot is removed from the inspection network; This represents the partition obtained based on spatial connectivity. A hollow sub-block.

[0284] The dashed arrows in the diagram represent the event response path or the return path after the response robot completes its task, while the solid arrows represent the coverage inspection direction, the partial takeover direction, or the coverage restoration direction. Figure 9 The shade of color in the grid indicates the grid's free space. The relative size of the grid is indicated by its color; darker colors represent longer periods since the grid was last covered. During the coverage restoration phase, the robot prioritizes accessing nodes with higher idle rates within its responsibility area to reduce the risk of long-term uncovering of empty areas.

[0285] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0286] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0287] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0288] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0289] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A heterogeneous multi-robot cooperative scheduling method for dynamic task insertion, characterized in that, include: Based on the heterogeneous state parameters of each robot in the multi-robot system, the comprehensive capability weight is calculated and the allocation cost is corrected. Based on the allocation cost, the responsibility area of ​​the inspectable nodes in the inspection network is divided to obtain the responsibility area of ​​each robot. Within each robot's area of ​​responsibility, the node with the highest coverage score is selected as the next target node for coverage, and the following scheduling steps are executed cyclically: The robot arrives at the next target node and completes a partial inspection, then updates the last coverage timestamp and idle time of the next target node. When discrete tasks exist, the set of hard constraint capabilities, the set of soft constraint capabilities, and the task priority of the discrete tasks are extracted. Based on the set of hard constraint capabilities, a set of candidate robots is selected from the multi-robot system. Based on the set of soft constraint capabilities and the task priority, the multidimensional composite response utility is calculated. Each robot in the candidate robot set is comprehensively evaluated to obtain the optimal response node. The optimal response node is then removed from the inspection network. The responsibility area corresponding to the optimal response node before stripping is taken as the void area to be repaired. The last coverage timestamp and idle degree of each node in the void area to be repaired are retained. Based on the spatial connectivity, the void area to be repaired is divided into multiple void sub-blocks. The robots that are in contact with the boundary of the responsibility area and the boundary of the void sub-block are taken as neighboring robots. A set of neighboring robots is constructed. Local secondary bidding is performed in the set of neighboring robots to determine the takeover robot of each void sub-block so as to take over the corresponding void sub-block. After the optimal response node completes the discrete task and rejoins the inspection network, the recovery responsibility area is divided based on the latest heterogeneous state parameters of each robot. The last coverage timestamp and idle degree of each node are retained. The node with the largest idle degree in the recovery responsibility area is selected as the next coverage target node.

2. The heterogeneous multi-robot cooperative scheduling method for dynamic task insertion according to claim 1, characterized in that, The heterogeneous state parameters include current location, remaining battery power, average moving speed, effective sensing radius, and capability tag set.

3. The heterogeneous multi-robot cooperative scheduling method for dynamic task insertion according to claim 1, characterized in that, The comprehensive capability weight is expressed as follows: ； In the formula, Represents robots The overall capability weight; All represent non-negative comprehensive ability weighting coefficients; They represent robots Normalized remaining battery power, normalized average moving speed, and normalized effective sensing radius; The allocation cost is expressed as: ； In the formula, Represents robots To the node The cost of allocation; This represents the obstacle-constrained grid diagram corresponding to the inspection network. China Robot Current location To the node The shortest accessible path distance; Represents positive integers; The responsibility area of ​​each robot is represented as follows: ； in, ; ； In the formula, They represent robots Responsibility area, robot The area of ​​responsibility; Represents the set of inspectable nodes; Represents a node The robot that belongs to you alone; Indicates the number of robots; Indicates from Selected The smallest robot number; This represents the empty set.

4. The heterogeneous multi-robot cooperative scheduling method for dynamic task insertion according to claim 1, characterized in that, The coverage score is expressed as: ； in, ; In the formula, Represents robots For nodes At any moment Coverage score; All of these represent non-negative coverage score weighting coefficients; Represents a node At any moment Normalized idle time; This represents the obstacle-constrained grid diagram corresponding to the inspection network. China Robot At any moment Location To the node The normalized shortest path distance; Represents a node Normalized geometric coverage consistency; Represents a node At any moment The degree of free time; Represents a node The last overwrite timestamp; The next target node for coverage is represented as: ； In the formula, Represents robots At any moment The next target node to be covered; Represents robots At any moment The set of candidate access nodes; Indicates from Selected The largest node.

5. The heterogeneous multi-robot cooperative scheduling method for dynamic task insertion according to claim 1, characterized in that, The set of candidate robots is represented as follows: ； In the formula, Represents the set of candidate robots; Indicates the number of robots; Represents the set of hard constraint capabilities; Represents robots A set of capability tags; Represents robots The current remaining battery power; This indicates the preset minimum safe power threshold; Represents robots The current working status; Represents a set of fault or offline states; Based on the set of soft constraints and task priorities, the multidimensional composite response utility is calculated, including: Calculate the soft matching score based on the soft constraint capability set; Calculate the estimated arrival time score of the robot to the discrete task location based on the estimated arrival time and the maximum estimated arrival time of the robot. Based on the robot's remaining inspection load and discrete task load, calculate the robot's current total load, and based on the robot's current total load, calculate the robot's remaining load capacity score. The multidimensional composite response utility is calculated based on task priority, soft matching score, expected arrival time score of robot to discrete task location, robot power score, and remaining carrying capacity score.

6. The heterogeneous multi-robot cooperative scheduling method for dynamic task insertion according to claim 5, characterized in that, The soft-match score is represented as: ； In the formula, Represents robots Soft match score; Represents the set of soft constraint capabilities; Labels indicating soft constraint capabilities Importance weights; Represents positive integers; The estimated arrival time of the robot at the discrete task location is expressed as: ； In the formula, Represents robots The estimated arrival time to the discrete task location; This represents the obstacle-constrained grid diagram corresponding to the inspection network. China Robot At any moment Location To discrete task location The shortest accessible path distance; Represents robots Average moving speed; The maximum estimated arrival time is expressed as: ； In the formula, Indicates the maximum estimated arrival time; This indicates taking the maximum value; The estimated arrival time score of the robot to the discrete task location is expressed as: ； In the formula, Represents robots The estimated arrival time score for reaching the discrete task location; The remaining inspection load of the robot is represented as follows: ； In the formula, Represents robots The remaining load of the inspection; Represents robots Inspection status indicator variables; Represents robots The size of the remaining area to be inspected within the current area of ​​responsibility; Represents robots Environmental reduction factor; Represents robots Effective inspection width; The discrete task load of the robot is represented as: ； In the formula, Represents robots Discrete task load; Represents robots The discrete tasks already undertaken The remaining execution time; Represents robots The number of discrete tasks currently undertaken; The robot's current total load is expressed as: ； In the formula, Represents robots The current total load; The remaining load capacity score of the robot is expressed as: ； In the formula, Represents robots The remaining carrying capacity score; Represents robots The maximum load that can be handled within the preset scheduling period; The robot's battery score is represented as follows: ； In the formula, Represents robots Battery score; Represents robots Rated power capacity; The multidimensional composite response utility is expressed as follows: ； In the formula, Represents robots Compared to the multidimensional composite response utility of discrete tasks; All indicate task priority. Relevant weighting functions; The optimal response node is represented as: ； In the formula, Indicates the node with the best response; Indicates from Selected The largest robot number.

7. The heterogeneous multi-robot cooperative scheduling method for dynamic task insertion according to claim 1, characterized in that, The hollow sub-block is represented as: ， ； In the formula, Indicates the area of ​​cavity to be repaired; Indicates the number of empty sub-blocks; Representing empty sub-blocks Hollow sub-blocks ; This represents the empty set.

8. The heterogeneous multi-robot cooperative scheduling method for dynamic task insertion according to claim 1, characterized in that, The set of neighboring robots is represented as: ； In the formula, Indicates a group of nearby robots; Indicates the number of robots; Indicates the node with the best response; Represents robots The current boundaries of the area of ​​responsibility; Represents a hollow sub-block The boundary; Indicates the empty set; Represents robots The current remaining battery power; This indicates the preset minimum safe power threshold; Represents robots The current working status; This represents a set of fault or offline states.

9. The heterogeneous multi-robot cooperative scheduling method for dynamic task insertion according to claim 1, characterized in that, Perform local secondary bidding within the neighboring robot set to determine the takeover robot for each empty sub-block, including: Based on the maximum path distance from the neighboring robot to the empty sub-block, calculate the adjacency score of the neighboring robot to the empty sub-block; The average free space of the empty sub-blocks is normalized to obtain the normalized average free space of the empty sub-blocks; The filling effect of neighboring robots on empty sub-blocks is calculated based on the adjacency score of neighboring robots on empty sub-blocks and the normalized average vacancy degree of empty sub-blocks. Based on the filling effect of neighboring robots on the hollow sub-blocks, the takeover robot for each hollow sub-block is determined; The maximum path distance from the neighboring robot to the hollow sub-block is represented as: ； In the formula, Indicates the current nearest robot to the empty sub-block. The maximum path distance; This represents the obstacle-constrained grid diagram corresponding to the inspection network. China Robot At any moment Location To the hollow sub-block Representative node The shortest accessible path distance; Indicates a group of nearby robots; This indicates taking the maximum value; The adjacency score of the neighboring robot for the empty sub-block is represented as follows: ； In the formula, Represents robots Hollow sub-blocks Adjacency score; Represents positive integers; The average freeness of the hollow sub-block is expressed as: ； In the formula, Represents a hollow sub-block At any moment Average idle time; Represents a hollow sub-block The number of nodes within; Represents a hollow sub-block ; Represents a node At any moment The degree of free time; The normalized average free space of the empty sub-block is expressed as: ； In the formula, Represents a hollow sub-block At any moment Normalized average idle time; Indicates the number of empty sub-blocks; The effect of the neighboring robot in filling the void sub-block is expressed as: ； In the formula, Represents robots Hollow sub-blocks The complementary effect; All of these represent non-negative filler utility weight coefficients; Represents robots The remaining carrying capacity score; Represents robots Battery score; The takeover robot for each hollow sub-block is represented as follows: ； In the formula, Represents a hollow sub-block The takeover robot; Indicates from Selected The largest robot number.

10. The heterogeneous multi-robot cooperative scheduling method for dynamic task insertion according to claim 1, characterized in that, The node with the highest idle time within the recovery responsibility area is represented as: ； In the formula, Represents robots At any moment Restore the node with the highest idle rate within the responsibility area; Indicates from Selected from The largest node; Represents robots The area of ​​responsibility for restoration; Represents a node At any moment The degree of free time.

Images