A two-stage distributed task allocation method based on prediction

Abstract

The present application relates to the technical field of multi-agent cooperative control, and specifically provides a two-stage distributed task allocation method based on prediction, wherein in the first stage, an initial solution is generated through task removal and task inclusion, wherein the task inclusion stage uses a deadline-sensitive cost function to expand the task gap, and the task removal stage introduces bid prediction to reduce invalid iterations; in the second stage, the overall scheme is further optimized through local search. The present application exhibits significant advantages in terms of rapid task information propagation, reduced iteration number and improved task allocation efficiency, and is particularly suitable for application scenarios such as search and rescue, which have extremely high requirements for task response speed. Since task information can be efficiently transmitted to the entire cluster in fewer iterations, the present method can significantly speed up the collaborative decision-making process, improve task execution efficiency, and thus enhance the response speed and adaptability of the cluster in search and rescue tasks.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of multi-agent cooperative control, in particular to a dual-stage distributed task allocation method based on prediction. BACKGROUND

[0002] With the rapid development of robot technology, mobile robots have made significant progress in functionality and operational stability. However, individual robots still have limitations in perception, decision-making and execution capabilities, making it difficult for them to independently complete complex tasks. Through communication and cooperation, multi-robot systems have become an efficient solution for executing tasks in complex environments, and multi-robot task allocation is a core problem in deploying multi-robot systems in practice, with the goal of optimizing overall system performance through rational task allocation.

[0003] In search and rescue scenarios, task allocation faces many real-world constraints, such as fuel limitations, deadline requirements, etc., while the primary goal is to maximize the rescue of survivors. Such problems usually belong to the task allocation problem under time constraints (ST-SR-TA), which is characterized by strong time dependency between tasks and a vast solution space, greatly increasing the complexity of the problem.

[0004] According to the different system organization architecture, the multi-robot task allocation method in search and rescue scenarios can be divided into centralized and distributed two categories. The centralized method relies on the central node to collect global information and optimize the target, although it can generate better solutions, but its computational complexity is higher, and it is easy to be affected by single point failure, it is difficult to adapt to the demand of large-scale task allocation. In contrast, the distributed method significantly improves the system robustness through decentralized computing and task execution, and is more advantageous in practical applications. Therefore, designing an efficient and feasible distributed task allocation method algorithm to generate a reasonable task allocation scheme is more suitable for practical applications, attracting extensive research attention. SUMMARY

[0005] In view of the problems of unstable communication of multi-robot system in search and rescue environment, low task completion rate, etc., the present application provides a dual-stage distributed task allocation method based on prediction (Dual-stage distributed task allocation (PDTA)) to improve the task planning ability of multi-robot system.

[0006] To solve the above technical problems, the technical solution adopted by the present application is: a dual-stage distributed task allocation method based on prediction, the specific content includes:

[0007] Constructing a task allocation objective function based on task requirements;

[0008] Defining a time cost function of a robot performing a task, optimizing a task allocation strategy based on the time cost function;

[0009] With the goal of maximizing the number of allocated tasks, an improved performance impact algorithm is used to generate a task allocation scheme;

[0010] A local search mechanism is used to disturb the task allocation scheme, integrate unallocated tasks into the task allocation scheme, and generate a final task allocation scheme.

[0011] Further, the task allocation objective function is defined as follows:

[0012]

[0013] Wherein, α i represents the task list allocated to robot i, the number of tasks in task list |α i | does not exceed its maximum task capacity L i , robot i executes tasks in order according to the task list, where t iκ (α i ) represents the time to arrive at the κth task , f i represents the maximum travel time of the robot, represents its deadline, h i,j ∈[0,1] represents whether the robot can execute task T j , h i,j =1 means that the task can be executed, otherwise 0.

[0014] Further, the time cost function is defined as follows:

[0015]

[0016] Wherein, C i (α i ) represents the total cost of robot i executing all tasks allocated to it, represents the travel time cost of the κth task in the task list α i of robot number i, represents the task deadline cost of the κth task in the task list α i of robot number i.

[0017] Further, in the process of optimizing the task allocation strategy based on the time cost function, tasks with smaller deadline costs are preferentially allocated, and tasks are quantitatively evaluated by combining the travel time cost and deadline cost to optimize the task allocation strategy.

[0018] Further, the improved performance impact algorithm comprises:

[0019] On the basis of the traditional performance impact algorithm, a task prediction mechanism is introduced in the task removal stage to predict the bid of the competitor, thereby reducing the interference of invalid bids on the allocation process.

[0020] A deadline-aware cost function is constructed to avoid hindering the allocation of subsequent tasks by preferentially allocating tasks with earlier deadlines.

[0021] A local search mechanism is constructed to broaden the solution space based on the local search mechanism.

[0022] Further, the process of generating a task allocation scheme using the improved performance impact algorithm comprises:

[0023] Each robot adds tasks in the task set to the task list until there are no more tasks that meet the conditions or the task list reaches the capacity.

[0024] When there is a conflict between robots in allocating the same task, a task removal process is performed based on the task prediction mechanism to generate the task allocation scheme.

[0025] Further, the task list comprises a global importance list, a marginal importance list, and a winner list.

[0026] In the process of each robot adding tasks in the task set to the task list, the contribution of adding or deleting tasks in the robot task list to the total cost of the robot is measured based on the performance impact IPI and removal performance impact RPI, when the IPI of the candidate task is less than the global importance, the task with the largest importance difference is inserted into the task list and the winner list is updated, when the IPI of the candidate task is not less than the global importance, the importance list is directly updated.

[0027] Further, performing a task removal process based on the task prediction mechanism comprises:

[0028] Determining a conflict task set and the robots i corresponding to the tasks in the conflict task set, if robot i is not the corresponding winner in the winner list, then the corresponding tasks in the conflict task set are deleted from the task list of robot i;

[0029] Calculating the significance difference and using a greedy algorithm to identify the task that maximizes the significance difference If If the task exists in the task lists of robots i and j at the same time, it is predicted that deleting Will cause the total cost to decrease, if so, continue to delete the task Otherwise, keep the task

[0030] Further, the process of perturbing the task allocation scheme by using a local search mechanism comprises:

[0031] Obtaining a task allocation result with reduced cost by exchanging the entire task list among the robots, or locally exchanging individual tasks in their task lists respectively;

[0032] Calculating a marginal importance list for all tasks in the unallocated task set, and obtaining the highest marginal importance, if a task in the unallocated task set is already included in the task list or does not satisfy the allocation constraint, assigning the marginal importance of the task as M;

[0033] If the highest marginal importance is less than M, inserting a task in the unallocated task set into the task list of a robot;

[0034] If the highest marginal importance is not less than M, indicating that a new task cannot be added, at this time, removing the allocated task with the highest RPI value, and calculating the maximum importance difference value among the robots If Further optimizing the task allocation result, and then re-performing local search optimization until

[0035] Further, the further optimization of the task allocation result is implemented based on the following formula:

[0036]

[0037] Wherein, represents removing task T1 from the task list of robot R1, and inserting task T2 into the position of the task list of robot R2, represents the global importance of task T1 in the task list, represents the boundary importance of task T2 in the task list. Compared with the prior art, the present application has the beneficial effects that:

[0038] The present application aims at the multi-robot multi-task allocation problem in search and rescue scenarios, constructs an optimization model with deadline, fuel limit and matching constraints, and proposes a two-stage distributed task allocation method based on prediction to maximize the number of rescues.

[0039]

[0040] ​​​​​​​​​The PDTA algorithm proposed in the application has excellent advantages in the task negotiation process. Compared with PI and PIassMax which rely on multiple rounds of task negotiation and bid updating to optimize task allocation, the method reduces unnecessary task removal operations in the task removal stage by predicting the bids of competing robots, thereby effectively reducing the number of iterations and accelerating the consensus.

[0041] The application has significant advantages in fast task information dissemination, reducing the number of iterations and improving task allocation efficiency, and is particularly suitable for search and rescue applications that require extremely high task response speed. Since task information can be efficiently transmitted to the entire cluster in fewer iterations, the method can significantly speed up the collaborative decision-making process, improve task execution efficiency, and thus enhance the response speed and adaptability of the cluster in search and rescue tasks. BRIEF DESCRIPTION OF DRAWINGS

[0042] Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are for purposes of illustration only and are not considered a limitation of the present application. Moreover, like reference numerals are used to designate identical components throughout the specification. In the drawings:

[0043] Figure 1 PDTA algorithm flowchart of the application

[0044] Figure 2 PI algorithm flowchart of the embodiment of the application

[0045] Figure 3 Overall flowchart of the task inclusion stage in the embodiment of the application

[0046] Figure 4 Conflict resolution flowchart based on prediction in the embodiment of the application

[0047] Figure 5 Flowchart of local search and fine solution in the embodiment of the application

[0048] Figure 6 Communication network topology of the PDTA algorithm in the embodiment of the application, wherein (a) is a mesh communication topology; (b) is a linear communication topology; (c) is a ring communication topology; (d) is a star communication topology

[0049] Figure 7 Comparison chart of the average number of iterations of different algorithms under high robot task ratio

[0050] Figure 8 Comparison chart of the average number of task allocations of different algorithms under high robot task ratio

[0051] Figure 9 Box plot comparison of the number of iterations for the four topologies when the robot task ratio is 12:84;

[0052] Figure 10 Box plot comparison of the number of iterations for the four topologies when the robot task ratio is 12:84;

[0053] Figure 11 Box plot comparison of the number of iterations for the four topologies when the robot task ratio is 12:96;

[0054] Figure 12 Box plot comparison of the number of iterations for the four topologies when the robot task ratio is 12:96. DETAILED DESCRIPTION

[0055] Exemplary embodiments of the present disclosure will be described in detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it is to be understood that the present disclosure can be embodied in various forms without being limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the present disclosure to those skilled in the art. It should be noted that the embodiments in the present disclosure and the features in the embodiments can be combined with each other without conflict. The present disclosure will be described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.

[0056] The present embodiment proposes a two-stage distributed task allocation method based on prediction, and the algorithm flow is as shown in Figure 1 The specific content includes:

[0057] Suppose in a post-disaster rescue environment, consider a team of rescue robots, denoted as V = {u1, u2,..., u n}, and a group of m survivors, denoted as T = {T1, T2,..., T m}. Our task goal is to provide emergency support for the survivors, some of which are drug support tasks, and some of which are food support tasks. In order to successfully rescue the survivors, appropriate rescue robots must be sent to the survivor locations, each robot can only execute the assigned tasks in the order specified in its task list, and the entire rescue team does not need to return to the starting location. All tasks are classified by demand, and each task is only assigned to a robot that can execute it. The matching relationship is defined by the compatibility matrix H, where each element h i,j ∈ [0, 1] represents whether the robot u i can execute the task T j , h i,j = 1 means it can execute the task, otherwise 0. Each task T j has a deadline η j and a fixed execution time τj The robot must begin performing its task before the deadline. Additionally, the robot... i The activity time is limited by fuel, and the maximum driving time is denoted as f. i Due to limited transport capacity, each robot can be allocated a maximum of L. i One task.

[0058] In this invention, the objective of task allocation is to maximize the number of survivors successfully rescued while satisfying constraints related to robot capabilities, task deadlines, fuel limitations, and task matching conditions. Based on the above description, this problem can be written as the following optimization problem:

[0059]

[0060]

[0061] Where α i Let |α| represent the list of tasks assigned to robot i. i The number of tasks shall not exceed its maximum task capacity L. i Robot i executes tasks sequentially according to the task list, where t iκ (α i ) indicates reaching the κ-th task. Time, This indicates the deadline. Additionally, there is a task list α between different robots i and j. i and α j There should be no conflict.

[0062] When multiple robots perform tasks, the total activity time of each robot can be divided into two parts: task execution time and travel time. Task execution time refers to the cumulative operation time required to complete all tasks in its task list, while travel time covers the time spent by the robot moving from its initial position to each task location, as well as the time spent moving between tasks. Since task execution time is usually relatively fixed, shortening the travel time between adjacent tasks allows the robot to complete more tasks within a limited activity time. Based on this principle, this invention optimizes task scheduling by combining travel time with task deadline costs to quantitatively assess task-related costs.

[0063] For task list α with robot number i i In the context, the travel time cost of the κ-th task The definition is as follows:

[0064]

[0065] Task deadline cost Defined as:

[0066]

[0067] Therefore, the total cost of the robot i performing all the tasks assigned to it can be defined as:

[0068]

[0069] The present application adopts a performance impact (PI) algorithm to solve the above problems. The algorithm is based on the concept of market transactions and is specially designed for task allocation problems in search and rescue environments. It belongs to the category of distributed task allocation algorithms. Although it originated from the CBBA algorithm, it is different from the CBBA algorithm which only focuses on local cost minimization. The PI algorithm introduces the concept of "importance" to optimize the overall goal, i.e. the sum of the arrival times of all robots. The implementation process of the traditional PI algorithm is shown in Figure 2 During the auction process, the algorithm continuously constructs and updates the task bundle of each robot, and reaches a consensus through the conflict resolution stage to generate a conflict-free task allocation scheme. However, the traditional PI algorithm has the problems of not being able to maximize the number of task allocations, being prone to local optimization, and having poor convergence. Therefore, the present application improves the traditional PI algorithm in the following aspects:

[0070] (1) Introducing a task prediction mechanism

[0071] Unlike existing methods, the present application introduces a task prediction mechanism in the task removal phase, aiming to predict the bids of competitors. This process significantly reduces the impact of invalid bids on the allocation process, thereby reducing the number of iterations required for the algorithm to converge. In addition, it also enhances the robustness of the algorithm in low-bandwidth communication environments.

[0072] (2) Designing a deadline-aware cost function

[0073] A deadline-aware cost function is developed to optimize task allocation by effectively integrating task deadlines. By giving priority to tasks with earlier deadlines, this function avoids inserting tasks with later deadlines too early, which can hinder the allocation of other critical tasks. This strategic priority ordering can maximize the number of allocated tasks, optimize resource usage, and improve overall system efficiency.

[0074] (3) Combining a local search mechanism

[0075] The present application adopts a local search mechanism to broaden the solution space and ensure that critical rescue tasks are not overlooked.

[0076] The specific steps of the present application mainly include two stages, the first stage is the initial solution generation based on prediction, and the second stage is the local search mechanism and refinement strategy.

[0077] I. Initial solution generation based on prediction

[0078] In this initial solution generation stage, the core process involves iterative optimization through two main steps: task inclusion and conflict resolution.

[0079] 1.1 Task inclusion

[0080] In the task inclusion stage, each robot successively adds tasks from the unassigned task set to its own task list until there are no more eligible tasks or the task list reaches its capacity. The present application uses inclusion performance impact (IPI) and removal performance impact (RPI) to measure the contribution of adding or removing a task from a robot's task list to the total cost of the robot, specifically, the IPI of a task t j inserted into the current task list α i of robot i, is calculated as follows:

[0081]

[0082] where denotes the task list α j after inserting task t i into position I. The RPI of a task t i removed from the task list α j of robot i is defined as follows:

[0083]

[0084] The overall flow of the task inclusion stage is described as follows: each robot i independently selects a candidate task set from the pool of tasks that need to be executed. This candidate set includes tasks that satisfy the following conditions: first, the task requirement type must match the robot type, indicated by h ij = 1. Second, no task in the candidate set should already be assigned to any other robot, i.e. Subsequently, the robot iteratively adds tasks to its task list within the limit of its capacity (|α i | ≤ L i ). During the task inclusion process, robot i maintains three lists: a global importance list a marginal importance list and a winner list δ iThe global importance list records the final bid of each robot for each task after the last round of task allocation negotiation, the marginal importance list records the local bid of each robot for all tasks in this round, and the winner list records the result of task allocation. The overall process is shown in Figure 3 .

[0085] 1.2 Conflict resolution

[0086] In the conflict resolution phase, there are two parts: consensus among robots and task removal based on prediction.

[0087] The consensus process among robots is described as follows: Since each robot independently builds its own task list in the task inclusion process, it may lead to conflicts among robots on the allocation of the same task. In order to solve these conflicts, the robot team must coordinate to reach a consensus. Specifically, in the consensus phase, robots communicate through the connected network, broadcast three types of lists among adjacent robots: the winner list δ i , which is used to record the result of task allocation; the global importance list ξ iκ , which is used to record the bidding state of robots to perform tasks; and the timestamp list, which is used to capture the latest communication time to facilitate information synchronization.

[0088] The consensus process follows the communication protocol in the CBBA algorithm. Each robot compares its global importance list with the list of adjacent robots. For each task, the task with a higher RPI can significantly reduce the total cost when removed, and thus is preferentially allocated to a robot with a lower RPI. In this process, the list is constantly updated and propagated throughout the network. After the consensus phase, all robots will have a consistent global importance list and winner list, indicating that a consensus has been reached. Then, each robot will enter the task removal process to eliminate tasks that are not designated as its own winner.

[0089] In this invention, unlike the PI algorithm that relies on the outdated global importance list in the last round of consensus for task removal, the method proposed in this invention adopts a task prediction mechanism. The traditional PI method ignores the fact that other robots are simultaneously performing task removal operations, which can lead to the erroneous removal of tasks that should be retained. To solve this problem, the PDTA method in this invention uses the exchange of task lists in the communication process to predict the subsequent bids of other robots, providing a more up-to-date and reliable reference for the task removal phase.

[0090] As shown in Figure 4 , the task removal process based on the prediction mechanism is described as follows: First, determine the conflict task set d i , which is composed of tasks in the task list α i , and the winner list δ iThe corresponding winner is not robot i. These tasks will be removed from robot i's task list. Similar to the task inclusion process, the decision to continue removing tasks is based on calculating significant differences. A greedy method is used to identify tasks that maximize this significant difference. From task list α i and conflict task set d i Delete this task. Finally, update α. i List of remaining tasks and their global importance ξ i .

[0091] The next stage is prediction. If the task... It is also included in the task list α of another robot j. j In the next step, it's necessary to predict the impact of removing this task from robot j to determine whether the task truly needs to be deleted. Using the same principles outlined in the previous task deletion phase, a new list of predicted importance is established. For the task If the conditions are met This indicates that removing the task from robot i will not reduce the total cost. Therefore, robot i retains the task and waits for the result of the subsequent consensus round. Conversely, if The task removal process continues iteratively, removing tasks until the conflict set d is reached. i Exhausted or This indicates that removing further tasks will not reduce the overall cost.

[0092] Subsequently, the task inclusion phase is performed for each robot, and the above procedure is repeated until the consensus, task deletion, and task inclusion phases no longer change, marking the end of the first phase of the algorithm.

[0093] II. Local Search Mechanism and Refinement Strategy

[0094] Although the algorithm generated an initial solution that satisfies the deadline constraints in the previous step, some tasks may still remain unassigned. In emergency rescue scenarios, the highest priority is to maximize the recovery of survivors. Therefore, this embodiment proposes a rescheduling method aimed at optimizing the initial task allocation, enabling the robot team to undertake additional tasks while ensuring that all constraints are met.

[0095] 2.1 Local Search Mechanism

[0096] like Figure 5As shown, the present invention explores whether a lower total cost can be achieved by integrating unassigned tasks based on the initial solution through a local search mechanism, thus more tasks can be assigned. The present invention employs two ways to conduct the exploration. First consider two robots of the same performance, i and j, and denote their respective task lists as i = [α i1 , α i2 ,..., α iκ ] and j = [α j1 , α j2 ,..., α jλ ], where α iκ and α jλ represent the κth task of robot i and j, respectively. The first method is to exchange the entire task lists of the two robots, exchanging α i with α j , to explore the task assignment between robots. The second method requires a local exchange within the task list by exchanging task α i in α ik with task α jλ in α j . This method prevents the algorithm from converging to a local optimum and can further explore the solution space.

[0097] 2.2 Refinement Strategy

[0098] After the local search mechanism perturbs the initial solution, a task assignment result D with reduced cost is obtained. Subsequently, similar to the task inclusion process outlined in the initial solution, an attempt is made to integrate tasks from the unassigned task set U into the current solution.

[0099] The refinement strategy is described as follows: Compute a marginal importance list ξ * (α υ , T) for all tasks in the unassigned task set U, where the highest marginal importance is denoted as This list represents how the insertion of each candidate task will affect the cost of the current robot. Used as a criterion to decide whether to include the task.

[0100] When attempting to insert task T into the task list of robot , if the insertion of task T into the task list α υ of robot , if the insertion of task T into the task list α υ of robot T is infeasible (the task is already included in the task list or does not satisfy the assignment constraints), the marginal importance of the task T is assigned an infinite value M, which identifies the task as unselectable at the current stage and effectively excludes it from further optimization calculations.

[0101] If it means that the assignment result can add the possibility of other unassigned tasks. The need to add a task, the added task is determined using the following formula:

[0102]

[0103] where the triplet (υ * , T * , p * ) represents the combination of marginal importance minimum is in the task list a * of the robot υ υ , in the position p * insert the task T * .

[0104] Subsequently, the task list and the unassigned task set U are updated according to the following formula: where represents the insertion of the task T * into the task list a of the robot υ in the p * position, and U = U{T *} represents the removal of the task T * from the unassigned task set U.

[0105] If the highest marginal importance it means that it is not possible to add new tasks through the local search mechanism. In this case, the assigned task with the highest RPI value is removed and an unassigned task is added to further refine the solution space. This process aims to reduce the overall cost, after which the local search optimization is reapplied.

[0106] The maximum importance difference value between robots is as follows:

[0107]

[0108] If it can be inferred that there is further optimization space for the task assignment result. Then, for the existing assignment result of the robot group V, the removal and reallocation operation of the task is performed, first based on the maximum difference between the marginal importance and the global importance, determine the task-robot pair that needs to be adjusted, remove the task from the task list of the robot υ . removing, and inserting inserting robots task list in The position can maximize the gain between marginal importance and global importance, and the two steps can vacate space for the insertion of subsequent tasks. The robot-task pair is calculated using the following formula.

[0109]

[0110] wherein, removing, and inserting from the task list of the robot and inserting the task into the task list of the robot at position, denotes the global importance of task T1 in the task list denotes the marginal importance of task T2 in the task list denotes the global importance of task T1 in the task list denotes the marginal importance of task T2 in the task list

[0111] This process aims to reduce the overall cost, and then reapply local search optimization. Until θ'≤0, there is no room for improvement in the overall allocation scheme, and the algorithm ends.

[0112] Further, the embodiment constructs a post-disaster rescue scene simulation model, aiming to simulate the complexity of the actual rescue environment. Under the condition of limited communication, quickly deploying heterogeneous unmanned rescue robots to provide emergency support for survivors becomes a key task. In the system setting: the survivor demand is divided into two categories of medical supplies and food, corresponding to the formation of two types of professional rescue teams of medicine transportation and food supply; the spatial distribution of survivors is randomly generated in a 10,000m×10,000m×1,000m (X×Y×Z) three-dimensional space using the Monte Carlo method, wherein the Z-axis coordinate simulates the distribution of different height layers; all tasks are set to have equal priority, wherein the medicine transportation task has a service time window of 300 seconds, and the food supply task extends the time window to 350 seconds; the earliest start time of each task is uniformly set to 0 seconds, and the actual deadline is randomly generated in the interval of 0 to 4,000 seconds; the rescue robots are randomly initialized at the starting position in a 10,000m×10,000m two-dimensional ground space, and are respectively configured with fixed cruising speeds of 30m / s and 50m / s; the fuel endurance time of each robot is limited in the interval of 3,500 to 4,000 seconds.

[0113] The proposed algorithm is compared with three well-known distributed task allocation algorithms, CBBA, PI and PImaxAss. Simulations are performed under four topologies. To eliminate the influence of randomness, each algorithm is run 50 times with randomly generated task scenarios and the same random seed is used to ensure fairness. The relevant parameter settings in this example are shown in Table 1.

[0114] Table 1

[0115]

[0116] The proposed algorithm is compared with three well-known distributed task allocation algorithms, CBBA, PI and PImaxAss. Simulations are performed under four topologies. To eliminate the influence of randomness, each algorithm is run 50 times with randomly generated task scenarios and the same random seed is used to ensure fairness. The relevant parameter settings in this example are shown in Table 1.

[0117] From Table 2 and Table 3, it can be seen that the test results under the four robot task ratios show that the proposed PDTA algorithm in the present application is superior to the other three algorithms, achieving a higher average task allocation number and requiring fewer iteration numbers. It is worth noting that when the number of tasks is small, although some tasks are not completed due to the early deadline (some robots cannot arrive before the deadline), the proposed algorithm, PI and PImaxAss algorithm can complete the remaining tasks with a relatively high completion rate, while the completion rate of the CBBA algorithm is relatively low. This shows that the present application can effectively reduce the communication amount between robots and fully exert the capacity of the multi-robot system to maximize the number of rescues.

[0118] Table 2

[0119]

[0120] Table 3

[0121]

[0122]

[0123] From Table 2, it can be seen that as the number of tasks increases, the task allocation capability of some algorithms will decrease significantly due to the constraint of the robot fuel limit. The proposed PDTA algorithm shows a significant improvement in task allocation performance, especially in high task scenarios. Further, we further tested with 16 and 18 robots, and gradually increased the number of tasks from 32 to 144. Figure 7 andFigure 8 It is clear that the PDTA algorithm has an advantage in the performance of task allocation under high robot-to-task ratio.

[0124] In addition, in practical applications, if the distance between robots exceeds the communication range, direct information exchange is not possible. Therefore, in each iteration, the embodiment adopts four communication network topologies, namely mesh, line, ring and star, as the communication framework to simulate the actual network communication of the robot cluster within the limited perception range and evaluate the performance of the algorithm under different topologies. The four topologies are shown in Figure 6 As shown in the figure, and are randomly regenerated in each iteration, bidirectional communication is used between nodes, and the sequence number is only for reference.

[0125] The embodiment compares the performance of the PDTA, CBBA, PI and PIassMax algorithms under different topologies, Figures 9-10 and Figures 11-12 The number of iterations and task allocation when the robot-to-task ratio is 12:84 and 12:96 are shown respectively. The results show that in the mesh topology, task information can quickly spread to all robots, achieving the highest negotiation efficiency and significantly reducing the number of iterations. In contrast, the information propagation range of star and ring topologies is limited, resulting in an increase in the number of iterations, while the line topology has the largest network diameter, limiting the speed of information transmission and resulting in the highest number of iterations.

[0126] The PDTA algorithm proposed in the present application has excellent advantages in the task negotiation process. Compared with PI and PIassMax, which rely on multiple rounds of task negotiation and bid updates to optimize task allocation, the present method reduces unnecessary task withdrawal operations by predicting the bids of competing robots during the task removal phase, thereby effectively reducing the number of iterations and accelerating consensus. In contrast, CBBA has the lowest number of task allocations, reflecting its limitations in performance; PI and PIassMax perform better in task allocation, but at the cost of higher iteration numbers, affecting overall efficiency.

[0127] The application is based on the traditional performance impact (PI) algorithm, and a task prediction mechanism is introduced in the task removal stage to predict the bid of the competitor, reduce the interference of invalid bids on the allocation process, thereby reducing the number of iterations required for algorithm convergence and enhancing the robustness in low-bandwidth communication environment. At the same time, a deadline-aware cost function is designed, which allocates tasks with earlier deadlines first to avoid subsequent tasks being blocked, thereby optimizing resource utilization, improving overall system efficiency, and maximizing the number of allocable tasks. In addition, the method combines a local search mechanism to search within the solution space in one step to optimize the task allocation strategy, ensuring that critical tasks are not ignored, so that the robot can maximize the number of survivors rescued in the search and rescue environment, thereby improving the rationality and stability of the algorithm.

[0128] Unlike the defects of existing methods that may remove tasks that should be retained, the application introduces a task prediction mechanism in the task removal stage. This mechanism uses the exchange of task lists in the communication process to predict the possible subsequent bids of other robots, thereby providing more timely and reliable references for task removal decisions. By reducing the occurrence of invalid task removal, this method effectively speeds up the convergence speed of the algorithm and improves the efficiency and stability of task allocation.

[0129] The application introduces a local search mechanism to expand the solution space and ensure that critical rescue tasks are not ignored. Although the initial scheme meets the deadline constraint, some tasks may still be unallocated, especially in emergency rescue, where maximizing survivor rescue is crucial. To this end, the method proposes a rescheduling strategy to optimize task allocation through local search, so that the robot takes on more tasks. Specifically, based on the cost function, the total cost is reduced and the available time interval of the task list is increased to insert additional tasks. The task exchange strategy includes two ways: one is to exchange robot task lists, and the other is to adjust locally within the task list to optimize the task order. Without changing the total number of tasks, this method improves the efficiency of task allocation, avoids local optimality, and more comprehensively explores the solution space.

[0130] In summary, the application has significant advantages in fast propagation of task information, reduction of iteration number and improvement of task allocation efficiency, and is particularly suitable for search and rescue and other application scenarios that require extremely high response speed. Since task information can be efficiently transmitted to the entire cluster in fewer iterations, the embodiment can significantly speed up the collaborative decision-making process, improve task execution efficiency, and thereby enhance the response speed and adaptability of the cluster in search and rescue tasks.

[0131] It should be pointed out finally that the above embodiments are only used to illustrate the technical solutions of the present application but not to limit it. Although the present application has been described in detail with reference to the above embodiments, it should be understood by those skilled in the art that the specific embodiments of the present application can be modified or replaced equivalently without departing from the spirit and scope of the present application, and any modification or equivalent replacement without departing from the spirit and scope of the present application should be covered in the protection scope of the claims of the present application.

Claims

1. A two-stage distributed task allocation method based on prediction, characterized in that, The application relates to a task allocation method and device. A task allocation target function is constructed based on task requirements; A time cost function of robot task execution is defined, and a task allocation strategy is optimized based on the time cost function; An improved performance impact algorithm is used to generate a task allocation scheme with the goal of maximizing the number of allocated tasks; A local search mechanism is used to disturb the task allocation scheme, integrate unallocated tasks into the task allocation scheme, and generate a final task allocation scheme; The time cost function is defined as shown in the following formula: wherein, denotes the total cost for robot i to perform all its assigned tasks, denotes the list of tasks for robot number i i in which the travel time cost of the kth task, denotes the list of tasks for robot number i i in which the task deadline cost of the kth task; The process of generating a task allocation scheme using an improved performance impact algorithm includes: Each robot adds tasks in the allocated task set to the task list until there are no more tasks meeting the conditions or the task list reaches the capacity; When conflicts occur among robots in allocating the same task, a task removal process is performed based on a task prediction mechanism to generate the task allocation scheme; The process of disturbing the task allocation scheme using a local search mechanism includes: The task allocation result with reduced cost is obtained by exchanging the entire task list among robots or locally exchanging individual tasks in the task list of each robot; The marginal importance list of all tasks in the unallocated task set is calculated, and the highest marginal importance is obtained, if a task in the unallocated task set is already included in the task list or does not meet the allocation constraint, the marginal importance of the task is assigned as M; If the highest marginal importance is less than M, the task in the unallocated task set is inserted into the task list of the robot; If the highest marginal importance is not less than M, it indicates that a new task cannot be added, at which time the assigned task with the highest RPI value is removed, and the maximum importance difference value between robots is calculated , if , the task assignment result is further optimized, and then the local search optimization is re-performed until ; The process of performing a task removal process based on a task prediction mechanism includes: The conflict task set and the robot i corresponding to the tasks in the conflict task set are determined, if the robot i is not the corresponding winner in the winner list, the tasks in the conflict task set are deleted from the task list of the robot i; computing a significance difference and using a greedy algorithm to identify a task that maximizes the significance difference if, if the task is present in both the task lists of robots i and j, then predict that removing will result in a decrease in the total cost, if so, then continue to remove the task otherwise, keep the task ; Further optimization of the task allocation result is realized based on the following formula: in, This represents the maximum difference in importance among the robots. Indicates the task From robots Task List Remove from, and will the task Insert robot Task List In Location, Represents the task list Medium task Its global importance Represents the task list Medium task The importance of the boundary, where V represents the assembly of rescue robots. .

2. The method of claim 1, wherein, The task allocation target function is defined as shown in the following formula: where J denotes the task assignment objective, α i denotes the task list assigned to robot i, α j denotes the task list assigned to robot j, task list |α i | has no more than its maximum task capacity L i , robot i executes tasks in the order of the task list, where t iκ (α i ) denotes the time to reach the κth task , f i denotes the maximum travel time of the robot, denotes its deadline, h i,j ∈ [0,1] denotes whether the robot can execute task T j , h i,j =1 means it can execute the task, otherwise 0.

3. The method of claim 1, wherein, In the process of optimizing the task allocation strategy based on the time cost function, tasks with smaller deadline costs are preferentially allocated, and the tasks are quantitatively evaluated in combination with the travel time cost and the deadline cost to optimize the task allocation strategy.

4. The method of claim 1, wherein, The task list includes a global importance list, a marginal importance list and a winner list; In the process of adding tasks in the allocated task set to the task list of each robot, the contribution of adding or deleting tasks in the task list of the robot to the total cost of the robot is measured based on the included performance impact IPI and the removed performance impact RPI, when the IPI of the candidate task is less than the global importance, the task with the largest importance difference is inserted into the task list and the winner list is updated, and when the IPI of the candidate task is not less than the global importance, the importance list is directly updated.

Images