Heterogeneous mobile robot task allocation method for complex inspection environment

Abstract

The invention belongs to the field of multi-robot task allocation, and discloses a heterogeneous mobile robot task allocation method for a complex inspection environment. According to the method, the matching relation between the robot and a task is constructed by fully utilizing the characteristics of heterogeneous robots (wheel type and four-foot robots) and combining the known target inspection environment topographic features; secondly, a revenue function in an auction algorithm is improved, a task dynamic reward value and a penalty term are added, task idling or excessive competition is avoided by judging the difference between actual auction of the task and the number of the needed robots, the execution efficiency of the overall task is improved, the utilization rate of the robots is balanced, and a certain robot is prevented from being used excessively. Through the method, the limitation of an isomorphic robot is overcome, the isomorphic robot fully utilizes the capability to distribute a reasonable inspection task, and the reliability of completing the task through the terrain is ensured, so that the whole task distribution system efficiently and stably runs, and the method has important application value for executing the inspection task in industrial and field environments.

Description

Technical Field

[0001] The present invention relates to a task allocation method for heterogeneous mobile robots in complex inspection environments, specifically including a task allocation method that utilizes the difference in the passing capabilities of quadruped robots and wheeled robots in different terrains to match the target task terrain. Background Art

[0002] In recent years, with the rapid development of related technologies such as artificial intelligence, people's research on multi-agent cooperation has become increasingly in-depth and achieved certain progress. Compared with single robots, multi-robot systems can provide higher efficiency, flexibility, and robustness. Especially in dangerous environments, they can achieve efficient task allocation and collaborative execution, making them widely used in fields such as warehousing logistics, intelligent inspection, and post-disaster rescue.

[0003] In the field of inspection, mobile robots have gradually become an important means to replace traditional manual inspections due to their efficient, accurate, and continuous working characteristics. The scale of equipment in industries such as factories, power grids, and petrochemicals is constantly expanding, and the complexity is constantly increasing, resulting in higher requirements for equipment inspection and maintenance. Traditional inspection methods often rely on manual labor, which is not only inefficient but also poses safety hazards. Therefore, using mobile robots for collaborative inspection has become an inevitable trend in the industry's development. However, when dealing with inspection tasks in some potentially dangerous and complex terrains, if multiple robots with the same structure are used for collaboration, it will result in limitations and single-mindedness in task execution. Therefore, considering the characteristics of each robot, combining multiple robots with different capabilities to collaboratively complete inspection tasks with diverse terrains has become the current research focus. For example, wheeled robots have a high driving speed and load capacity and are suitable for performing tasks in flat or less obstacle-ridden terrains. However, when there are obstacles at the entrance of a region, they cannot enter the target task area to complete the inspection task, or when there are rough terrains in the target area and it is inconvenient for the robot to complete all inspection tasks, then a quadruped robot is considered. With its strong obstacle-crossing ability, it can perform tasks in complex and rough terrains, but its own load capacity is low and its running speed is slower than that of wheeled robots. Therefore, how to allocate the target inspection area by utilizing the different performance structures of wheeled robots and quadruped robots has become a problem to be solved.

[0004] Task allocation is usually classified into 8 different cases from three characteristic perspectives of different tasks, robot capabilities, and allocation time. The inspection task to be implemented in this paper belongs to the ST–SR–IA category (that is, at the same moment, a robot can only execute the same task, one task only requires one robot to complete, and the task is allocated to the robot at each moment without planning future allocations), and usually, an auction algorithm based on the market mechanism is used to solve it. However, most current research focuses on homogeneous robots, and usually does not consider the degree of competition of robots for tasks and the utilization rate of each robot during allocation, resulting in the problems of overuse of a certain robot and too much competition. Therefore, this paper proposes a task allocation method for multi-heterogeneous mobile robots to solve this problem. Summary of the Invention

[0005] The purpose of the present invention is to design an overall solution for the task allocation problem of heterogeneous multi-mobile robots for inspection tasks applied in complex environments, and effectively select the most suitable robot among heterogeneous robots to complete the corresponding inspection tasks according to the terrain environment.

[0006] To achieve the above purpose and improve the deficiencies of existing methods, the overall solution provided by the present invention is as follows: construct a robot and task model based on known capabilities and environmental information, and then use the obstacle height, slope size, and gully width in the task model to match the limit capabilities of each robot in the robot model to judge the matching degree; secondly, improve the revenue function during the auction of task allocation, combine the matching degree between the robot and the task, add a task dynamic reward value to reduce the competitiveness of the task, add a penalty term to balance the utilization rate of the robot to prevent overuse of a certain robot, and first calculate and select the task with the largest revenue value to submit for auction to ensure the efficiency of the robot.

[0007] The implementation process is as follows:

[0008] S1: Construct a robot and task model, and establish a matching relationship between the target task terrain and robot capabilities;

[0009] S2: Construct a dynamic task allocation model with optimal revenue, where the revenue function consists of the matching degree p ij , task priority P, the cost required to execute the task, the reward value f ij , and penalty term a; specifically, the matching degree is determined by the matching relationship in S1, the cost includes distance cost and the cost of executing the inspection task, the task reward value will be dynamically adjusted according to the auction information based on the competition degree of the robot to improve the efficiency of task allocation, and the penalty term is to prevent individual robots from being overused and balance the utilization rate of robots in the whole system, so as to improve the auction method to complete the dynamic inspection task allocation of heterogeneous robots in complex environments;

[0010] Advantages of the present invention: The described heterogeneous mobile robot task allocation method for complex inspection environments enables quadruped robots and wheeled robots to fully consider their own capabilities and match them with the task terrain. When allocating tasks, it comprehensively considers the matching degree, the competition degree of robots for tasks, and the utilization rate of robots, so as to ensure that robots are allocated to the designated target inspection area while improving the task allocation efficiency and balancing the utilization rate of robots in the entire system. Brief Description of the Drawings

[0011] Figure 1 It is a schematic diagram of the overall process for multi-heterogeneous mobile robot task allocation;

[0012] Figure 2 It is a flowchart for evaluating the matching of task terrain and robot capabilities; Detailed Embodiments

[0013] The following further elaborates on the embodiments of the present invention in conjunction with the drawings and embodiments.

[0014] An embodiment of the present invention: A heterogeneous mobile robot task allocation method for complex inspection environments; in the case of a large number of inspection tasks and uncertain task location points in complex terrains, a method needs to be set up to reasonably allocate the target inspection task area to robots with different capabilities to perform inspection tasks, such as Figure 1 As shown, the specific content of the present invention includes the following steps:

[0015] S1: Define the robot and task models, and conduct a matching evaluation based on the known matching relationship between the task terrain and robot capabilities;

[0016] S1.1 Construct a heterogeneous robot model; assume that there are two robots performing tasks in the environment: a quadruped robot and a wheeled robot r i (i = 1, 2), and the robot model is constructed as:

[0017] robot = {i, v i , x i , y i , I, T i , M slope , M step , M ditch}

[0018] Where i is the i-th robot, and the coordinate position in the environment is (x i , y i ), which is continuously adjusted as the robot moves. For example, when the robot moves to task j, the coordinate of the robot at this time is the coordinate of this task point, vi V is the running speed of the robot, and I is the current state of the robot, including: busy and idle. If the robot is currently executing a task, it cannot bid for other tasks and needs to wait. T i is the set of tasks that the robot has already executed, M slope and M step and M ditch are the limit values of the robot passing through slopes, obstacles, and gullies respectively. Since quadruped robots and wheeled robots have different physical structures, their abilities to pass through complex terrains are different. Therefore, through geometric-physical level analysis and experimental verification, the above three limit constraint values can be obtained.

[0019] S1.2 Construct a task model; assume that there are M task points t to be completed in the environment j (j = 1, 2,..., M). Bid for the point with the lowest cost among the multiple accessible positions in each inspection task area as the target point of the entrance position of the inspection task (x j , y j ). The tuple for constructing the task model is:

[0020] Task = {j, x j , y j , P, f ij , h j , T slope , T step , T ditch}

[0021] where j is the jth task, the task coordinate position in the environment is (x j , y j ), f ij is the reward value that robot i can obtain for completing task j, h j is the number of robots bidding for this task, P is the task priority, T slope , T step , T ditch are the slope gradient, obstacle size, and gully width in the target inspection task area respectively, and they can be obtained through the perception method of multi-sensor fusion.

[0022] S1.3 Evaluate the matching degree between task terrain and robot capabilities; obtain the obstacle, slope, and gully information T slope , T step , T ditch of the target inspection area through prior environmental perception, as well as the limit constraint values M slope , M step , M ditch of each robot passing through different terrains. Then, the matching degree of this terrain for each robot can be judged, and its overall work process is as follows Figure 2as shown

[0023] The matching degree of the terrain is evaluated by the terrain factor f. First, since slopes with different gradients have different passabilities for different robots, the matching degree of the gradient is measured by the gradient factor f slope , which measures the difficulty of the robot climbing slopes and can be expressed as:

[0024]

[0025] When the gradient in the environment is within the constraint of the robot's climbing gradient threshold, the robot meets the requirements of the terrain and can freely pass through the slope; the larger the gradient factor, the greater the difficulty for the robot to climb the slope, and the worse the matching degree under this terrain; on the contrary, when the gradient in the environment exceeds the constraint of the robot's climbing gradient threshold, the robot cannot meet the climbing requirement, the gradient factor is 1, and it does not match the environment, and this area will be regarded as an impassable area for the robot.

[0026] Secondly, the matching degree of obstacles is measured by the height difference factor f step , which measures the ability of the robot to cross obstacles and can be expressed as:

[0027]

[0028] When the height T of the obstacle in the environment step is within the constraint of the maximum height M that the robot can cross step , the robot meets the requirements of the terrain. The larger the height difference factor, the greater the difficulty for the robot to cross the obstacle, and the worse the matching degree under this terrain; on the contrary, when the height of the obstacle in the environment exceeds the threshold constraint of the maximum height that the robot can cross, the robot cannot meet the obstacle-crossing requirement, the step height difference factor is 1, and it does not match the environment, and this area will be regarded as an impassable area for the robot.

[0029] Then, the matching problem of the gully width can also be measured by a similar method using the gully factor f ditch Therefore, the terrain evaluation factor is defined as:

[0030] f i = ω1f slope + ω2f step + w3f ditch

[0031] where ω1, ω2, and ω3 represent the weights of each item, indicating the degree of emphasis on different features, and the sum is 1. The higher the terrain factor, the greater the degree of mismatch of the robot to the environment and the less suitable it is. On the contrary, it is suitable.

[0032] The target inspection task is to traverse the target area and complete the inspection task of whether there are leaks and potential hazards in the area by carrying temperature sensors and gas sensors, or to reach the designated target task area to complete the inspection tasks of some infrastructure.

[0033] Before designing and improving the auction algorithm, considering the characteristics of heterogeneous mobile robots, the following premises need to be set:

[0034] (1) Since this paper mainly studies the task allocation problem among heterogeneous robots, at this time, it is assumed that except for the target task area in the environment, which contains obstacles, slopes, and gullies, there are no such terrains in other places;

[0035] (2) Due to different structures of heterogeneous robots, their running speeds and abilities to pass through the terrains of the target inspection tasks are different;

[0036] (3) The robots keep running rapidly, and this paper ignores the speed-related problems caused by passing through different terrains;

[0037] (4) To avoid some robots being overused and overloaded, it is necessary to balance the utilization rate of robots;

[0038] (5) Given a set of inspection tasks distributed in various regions and with known environmental terrains, it is stipulated that one task is completed by one robot, and a single robot can only execute one task at the same time;

[0039] (6) Ensure that the total revenue of executing the target task is maximized as much as possible.

[0040] S2: Construct a dynamic task allocation model with optimal revenue, where the revenue function consists of the matching degree p ij , task priority P, the cost required to execute the task, the reward value f ij for completing the task, and the penalty term a; specifically, the matching degree is determined by the matching relationship in S1, the cost includes the distance cost d ij and the cost s ij for executing the inspection task, the task reward value will be dynamically adjusted continuously according to the auction information based on the competition degree of the robots to improve the efficiency of task allocation, and the penalty term is to prevent individual robots from being overused and balance the utilization rate of robots in the whole system, so as to improve the auction method to complete the task allocation of heterogeneous robot inspections in complex environments;

[0041] S2.1 Judges the competition degree of each robot for the task through the improved revenue function; comprehensively considering the robot r i and the task t jThe distance cost, the matching degree between the task terrain and the robot's capabilities, the task priority, the expected reward value that can be obtained by completing the task, the penalty term. The greater the benefit value, the greater the probability that the robot will deliver the task. The specific benefit function is as follows:

[0042]

[0043] The representation methods of the parameters in the formula are as follows:

[0044]

[0045] First, to prevent a certain task from being over-competed or idle, the difference between the number of robots actually bidding for the task and the number of robots required is used to add a dynamically adjusted task reward value f ij , which means the reward value that robot i can obtain for completing task j. The higher the reward value, the greater the benefit of task j for robot i and the more likely it is to execute the task. β is the adjustment range of the control reward, with a value range of (0,1). Since this paper studies that each task only requires one robot to complete, h j -1 is used to describe the difference between the number of robots actually bidding for the task and the number of robots required. The adjustment concept is: if multiple robots reach the maximum benefit value for the same task, that is, h j >1, then reduce the reward for completing the task, update the task information again for the next round of auction to avoid over-competition. On the contrary, if there is no robot auctioning for a certain task, the task reward will be increased, and the task information will also be updated again to increase competitiveness and improve the execution efficiency of the overall inspection task; in actual applications, according to the real-time task auction situation, reasonably adjust the reward value to prompt the robot to select tasks more efficiently, thereby improving the execution progress of the overall task;

[0046] Secondly, in order to make full use of each robot and balance the utilization rate of all robots, a penalty term a is added to reduce the benefit of the over-utilized robot for subsequent tasks. Among them, μ is the adjustment factor for controlling the penalty intensity, L i is the number of tasks that the robot has already executed, n is the optimal number of tasks that each robot needs to execute, which can be obtained through the number of target inspection areas and a large number of experimental tests. When L i ≤n, it means that no robot is over-utilized at this time and no penalty needs to be added. On the contrary, it means that the robot is over-utilized at this time, then add a penalty term to reduce the task benefit of the robot during the task auction, and the greater the difference, the greater the penalty; in the actual complex inspection environment, adding this penalty term can avoid a certain robot from being over-consumed due to frequent task execution, maintain the stable operation of the entire task allocation system, and is of great significance for large-scale inspection tasks;

[0047] In addition, the specific meanings of other parameters in the formula are as follows: p ij represents the matching degree between robot i and task j, d ij is the distance cost, representing the distance between robot r i and task t j ; s ij is the cost of performing the inspection task in a certain area, ω i is the weight factor; each time the robot will select the task with the largest revenue value for delivery auction, then an optimization model for task allocation that maximizes the total revenue is established as follows:

[0048]

[0049] where x ij represents the task allocation parameter, that is, whether to allocate the j-th task explored to robot i. Specifically, it can be expressed as:

[0050]

[0051] The first constraint in the model is that one task only needs one robot to complete, which is SR. The second constraint is that one robot can only execute the same task at the same time, which is ST.

[0052] S2.2 adopts an auction method improved based on the market method to achieve specific task allocation, so as to solve the above optimization problem and complete the task allocation. The implementation process is as Figure 1 shown, mainly including:

[0053] Step 1: Publish the task area and initialize the parameter information, that is, announce the set of task information θ r , which includes the tasks not allocated in the previous round of auction. Initially, this set is {t1, t2,..., t m}, and the specific information of each task is in the task model. Then traverse the robot set to check whether the current robot status is "idle". If "idle", go to Step 2. If "busy", the robot does not participate in the task bidding in this round of auction and waits for the next round of auction until the robot is "idle" and then calculates the revenue value.

[0054] Step 2: In each round of bidding, all robots bid on multiple accessible positions of all unallocated tasks. The point with the lowest bidding cost among the multiple accessible positions of each task, that is: min(d ij + s ij ), is used as the task point in this area, and thus a set of task information points is received.

[0055] Step 3: Each robot calculates the benefit value that each task can obtain, forming a benefit matrix, whose dimensions are the number of rows of robots and the number of columns of tasks; secondly, select the task with the largest benefit value for delivery auction, that is:

[0056]

[0057] The system traverses all unassigned tasks and processes them as follows according to the list of bidding robots for each task: If the length of the list of bidding robots for task j is 1, that is, only one robot bids, then the robot wins the bid for this task; if the length of the list of bidding robots for task j is greater than 1, that is, the number of robots bidding for this task is greater than 1 (h j > 1), and multiple robots have reached the maximum benefit value for this task, then reduce the reward for completing this task to reduce the attractiveness of this task and avoid excessive competition, and update the task information to conduct the next round of auction; at the same time, if the benefit values of the same robot for multiple tasks are the same, the priority will be judged first. If there is a task with a higher priority, then the robot will win the bid for this task. On the contrary, if the priorities are the same, the distance cost will be compared, and the task with the smaller distance cost will be auctioned first; if the utilization rate of the same robot is too high, a penalty will be added to the benefit to make each robot fully utilized, so as to dynamically adjust the tasks.

[0058] Step 4: In this round of auction, the robots that win the bid are combined with the tasks. The robot wins the bid for the task and updates the execution status to "busy", indicating that it is executing or preparing to execute the task. After the task is completed, update the robot's coordinate position to the diagonal point of the entrance position, and remove the task from the task list until all tasks are assigned and meet the constraints to maximize the total benefit, that is, the task assignment of the heterogeneous robots is completed.

Claims

1. A heterogeneous mobile robot task allocation method for complex inspection environments, characterized in that: The following steps are involved: S1: Build robot and task models to establish the matching relationship between target task terrain and robot capabilities; S2: Construct a dynamic task allocation model with optimal benefits, where the benefit function is determined by the matching degree p ij , task priority P, the cost of executing the task, and the reward value f for completing the task ij , and penalty term a; specifically, the matching degree is determined by the matching relationship in S1, the cost includes the distance cost and the cost of executing the inspection task, the task reward value will be dynamically adjusted according to the degree of competition among the robots and the auction information to improve the efficiency of task allocation, and the penalty term is to prevent individual robots from being overused and balance the utilization rate of the robots in the entire system, so as to improve the auction method to complete the allocation of heterogeneous robot inspection tasks in complex environments.

2. According to the robot and task model described in S1 of claim 1, a matching relationship between the target task terrain and the robot capability is established, characterized in that: The robot and task model is constructed based on the known capabilities and environmental information. The obstacle height, slope size, and gully width in the task model are then matched with the limit capabilities of each robot in the robot model to obtain a matching relationship, including: Robot model: robot = {i,v i ,x i ,y i ,I,T i ,M slope ,M step ,M ditch Specifically, there are two robots in the environment that perform tasks: a quadruped robot and a wheeled robot. i (i=1,2), where i is the i-th robot, v i is the robot's running speed, and the coordinate position in the environment is (x i ,y i ) As the robot moves, its coordinate position is constantly adjusted. I is the current state of the robot, including: busy, idle, T i is the set of tasks that the robot has performed, M slope 、M step 、M ditch are the limit values ​​of the robot passing through slopes, obstacles, and gullies respectively; Task model: Task = {j, x j ,y j ,P,f ij ,h j ,T slope ,T step ,T ditch Specifically, there are M task areas t to be completed in the environment j (j=1,2,...,M), the point with the lowest bid cost among multiple accessible locations in each task area is used as the entry location target point of the task (x j ,y j ), where j is the jth task, f ij is the profit reward value that robot i can obtain by completing task j, h j is the number of robots bidding for the task, P is the task priority, T slope , T step , T ditch They are the slope of the slope, the size of the obstacle, and the width of the gully under the task; The matching relationship between the robot and the target task terrain is evaluated using the terrain factor f. First, because slopes of different slopes have different passability for different robots, the slope matching problem is evaluated using the slope factor f. slope , which measures the difficulty of the robot climbing, as shown in formula (1); when the slope in the environment is within the robot climbing slope threshold constraint, the robot meets the requirements of the terrain and can freely pass through the slope; the larger the slope factor, the greater the difficulty of the robot climbing, and the worse the matching degree under the terrain; on the contrary, when the slope in the environment exceeds the robot climbing slope threshold constraint, the robot cannot meet the climbing requirements, the slope factor is 1, then it does not match the environment, and the area will be regarded as an inaccessible area for the robot; The matching problem of gully width and the matching degree of obstacles can also be solved by using the gully factor f in a similar way. ditch , height difference factor f step Therefore, the terrain assessment factor is defined as f i =ω1f slope +ω2f step +w3f ditch , where ω1, ω2, and ω3 represent the weight of each item, indicating the importance attached to different features, and the sum is 1. The higher the terrain factor, the greater the degree of mismatch between the robot and the environment, and the less suitable it is. On the contrary, it is suitable.

3. The task allocation model according to S2 in claim 1, characterized in that: Based on the optimization concept, a dynamic task allocation model with the maximum total benefit is constructed, including: The improved benefit function is used to judge the best choice of each robot for the task, taking into account the cost, the matching degree between the task terrain and the robot's ability, the task priority, the expected reward value for completing the task, and the adaptive penalty item. The larger the benefit value, the greater the probability that the robot will deliver the task. The specific benefit function As shown in formula (2): The parameters in the formula are expressed as follows: First, to prevent a task from being over-competed or idle, a dynamically adjusted task reward value f is added based on the difference between the number of robots actually bidding for the task and the number of robots required. ij , which means the reward value that robot i can get for completing task j. The higher the reward value, the greater the benefit of robot i to task j and the more likely it is to perform the task. β is the amplitude of the control reward adjustment and the range is (0,1). Since this paper studies that each task only requires one robot to complete, h is used. j -1 is used to describe the difference between the number of robots actually bidding for the task and the number of robots required. The adjustment concept is: if multiple robots achieve the maximum benefit value for the same task, that is, h j >1, the reward for completing the task will be reduced, and the task information will be updated for the next round of auction to avoid excessive competition. On the contrary, if there is no robot auctioning for a task, its task reward will be increased, and the task information will be updated to increase competitiveness and improve the execution efficiency of the overall inspection task. In actual applications, the reward value is reasonably adjusted according to the real-time task auction situation to encourage the robot to select tasks more efficiently, thereby improving the overall task execution progress. Secondly, in order to make full use of each robot and balance the utilization of all robots, a penalty term a is added to reduce the benefits of over-utilized robots on subsequent tasks, where μ is the adjustment factor for controlling the penalty intensity, L i is the number of tasks that the robot has performed, and n is the optimal number of tasks that each robot needs to perform, which can be obtained through the number of target inspection areas and a large number of experimental tests. i When ≤n, it means that no robot is over-utilized and no penalty is added. On the contrary, it means that the robot is over-utilized. In this case, adding a penalty item will reduce the task income of the robot in the auction task, and the greater the difference, the greater the penalty. In the actual complex inspection environment, adding this penalty item can avoid excessive loss of a robot due to frequent task execution, maintain the stable operation of the entire task allocation system, which is of great significance for large-scale inspection tasks. In addition, the specific meanings of other parameters in the formula are: ij represents the matching degree between robot i and task j, d ij The robot r is the distance cost i With the task j The distance between ij The cost of performing inspection tasks in a certain area, i is the weight factor; each time the robot will select the task with the largest profit value for delivery auction, then the task allocation optimization model that maximizes the total profit is established as shown in formula (3): where x ij Represents the task allocation parameters: That is, whether to assign the explored jth task to robot i; Step 1: Publish the task area and initialize parameter information, that is, publish the task information set θ that has not been assigned r , which includes the tasks that were not assigned in the previous round of auction. Initially, the set is {t1,t2,...,t m }, the specific information of each task is in the task model, and the robot set is traversed to check whether the current robot status is "idle". If it is "idle", go to step 2. If it is "busy", the robot will not participate in the task bidding in this round of auction and wait for the next round of auction until the robot is "idle" to calculate the revenue value; Step 2: In each round of bidding, all robots bid for multiple accessible positions of all unassigned inspection tasks. The point with the lowest bidding cost among multiple accessible positions of each task is: min(d ij +s ij ) as the task point of the area, thereby receiving the task information point set; Step 3: Each robot calculates the revenue value that can be obtained from each task to form a revenue matrix, whose dimensions are the number of robots in rows and the number of tasks in columns; secondly, the task with the largest revenue value is selected for delivery auction, that is: The system traverses all unassigned tasks and processes as follows according to the bidding robot list of each task: if the length of the bidding robot list of task j is 1, that is, only one robot has submitted it, then the robot bids for the task; if the length of the bidding robot list of task j is greater than 1, that is, multiple robots have reached the maximum benefit value for the task, then the reward for completing the task is reduced to reduce the attractiveness of the task and avoid excessive competition, and the task information is updated for the next round of auction; at the same time, if the same robot has the same benefit value for multiple tasks, the priority will be determined first. If there is a task with a higher priority, it will be auctioned to that task. Otherwise, if the priorities are the same, the distance costs will be compared and the task with a smaller distance cost will be auctioned first; Step 4: The robots that have successfully bid in this round of auction are combined with tasks. The robot bids for the task and updates the execution status to "busy", indicating that the task is being executed or is about to be executed. After the task is completed, the robot's coordinate position is updated to the diagonal point of the entrance position, and the task is removed from the task list until all tasks are assigned and the constraints are met to maximize the total revenue. In other words, the task assignment of the heterogeneous robot is completed.

Images