A heuristic mission planning method for unmanned vehicles based on multi-point dynamic cluster construction

Abstract

The present invention relates to a heuristic task planning method for unmanned vehicles (UAVs) using a multi-point dynamic cluster structure. The method includes identifying the highest-value tasks, determining the UAVs best suited for those tasks and adding them to the team list. For each task, the benefits of all UAVs are calculated, the top k ranked UAVs are selected, and a task alliance with a certain degree of redundancy is formed. This process is repeated until all assigned tasks are effectively allocated. For UAVs that have not yet been assigned or have completed their assigned alliance tasks, the potential benefits of participating in each remaining task are calculated, and the UAVs are assigned to the tasks with the highest benefits. Compared with existing technologies, the present invention improves the efficiency and effectiveness of task execution, fully utilizes the resources of the UAVs in the team, generates high-quality solutions in a relatively short time, and meets real-time requirements.

Description

Technical Field

[0001] The present invention relates to an autonomous driving technology method, and in particular to a multi-point dynamic cluster construction heuristic task planning method for an unmanned vehicle. Background Art

[0002] MPDA, short for Multipoint Dynamic Aggregation Problem, is an optimization problem that involves the dynamic aggregation of multiple mobile points. It is necessary to determine the best path for a set of mobile points, such as drones, robots, or vehicles, to complete a task within a specific time window.

[0003] Currently, technologies for solving MPDA problems primarily focus on metaheuristic algorithms. Metaheuristic algorithms are a type of heuristic search method based on biological or social behavior in nature, used to solve complex problems. Commonly used metaheuristic algorithms include genetic algorithms, ant colony optimization algorithms, and particle swarm optimization algorithms. These algorithms gradually optimize solutions through local search and policy adjustments. They are simple to implement in solving MPDA problems and perform well in simple scenarios. They can gradually optimize solutions through local search and policy adjustments, finding optimal mobile point paths and task allocation solutions.

[0004] However, these techniques can require significant computational resources and time to solve large-scale problems, and may suffer from problems such as the curse of dimensionality in the search space and slow convergence, limiting their feasibility and efficiency in practical applications. Furthermore, they can become trapped in local optima, with no guarantee of finding a global optimal solution. The selection and tuning of algorithmic parameters significantly impacts performance, but achieving good results often requires experience and extensive experimentation. Summary of the Invention

[0005] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a multi-point dynamic cluster construction heuristic task planning method for unmanned vehicles.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] The present invention provides a multi-point dynamic cluster construction heuristic task planning method for an unmanned vehicle, comprising the following steps:

[0008] Step S1: Identify the task with the highest value from the task list of the unassigned team list, and add the unmanned vehicle with the highest benefit value for achieving the task to the team list of the task;

[0009] Step S2: Calculate the benefits of jointly executing the task by the unmanned vehicles that are not on the task team list and the unmanned vehicles that are on the team list, sort the benefits from high to low, and add the unmanned vehicles in the sorted order from high to low to the task team list one by one until the combined capability value of the unmanned vehicles on the task team list meets the task requirements;

[0010] Step S3: Calculate the time required for the unmanned vehicles that are not on the task team list and the unmanned vehicles that are on the team list to jointly perform the task, select the unmanned vehicle with the shortest time, calculate the relative impact value of the unmanned vehicle on the task based on the shortest time, and compare the relative impact value with a first preset threshold. If the relative impact value is greater than the first preset threshold, the unmanned vehicle is added to the task team list; otherwise, it is not added.

[0011] Step S4: remove the task with the highest value from the task list of the unassigned team list, and determine whether the task list in the current task list is empty. If so, execute step S5; otherwise, return to step S1;

[0012] Step S5: For the unmanned vehicles that have not joined any team list and have completed tasks, calculate their benefits from participating in unfinished tasks and assign them to the task with the highest benefit.

[0013] Furthermore, the specific steps of step S1 include: calculating the benefits of each unmanned vehicle that has not been added to the team list for each task that has not been assigned to the team list, sorting the benefits, and the task corresponding to the one with the highest benefit is the task with the highest value, and the corresponding unmanned vehicle is added to the team list of the task.

[0014] Furthermore, the calculation formula for calculating the benefit of each unmanned vehicle that is not included in the team list for each task that is not assigned to the team list is:

[0015]

[0016] Among them, t j represents the time when the autonomous vehicle j arrives at the task point, β j represents the capability value of the unmanned vehicle j, represents the task state curve, α n represents the task state change rate of the current task n, t represents the current time, Represents the state curve of task n, t e Indicates the moment when the task impact value Φ is reached, the benefit of each unmanned vehicle j for each task is f n,j , f n,j =1 / t e .

[0017] Furthermore, the capability value is an inherent parameter of the unmanned vehicle.

[0018] Furthermore, the task impact value Φ is a constant.

[0019] Furthermore, in step S2, the profit of the unmanned vehicles that are not included in the task team list and the unmanned vehicles that are included in the team list jointly performing the task is calculated using the following formula:

[0020]

[0021] Among them, t k represents the time when the unmanned vehicle k that joins the task team list arrives at the task point, β k represents the capability value of the unmanned vehicle k that joins the task team list, t l represents the time when the unmanned vehicle l that is not included in the task team list arrives at the task point, β l Indicates the capability value of the unmanned vehicle l that is not included in the task team list. represents the task state curve, α n represents the task state change rate of the current task n, t represents the current time, Represents the state curve of task n, t e Indicates the moment when the task impact value Φ is reached. The benefit of each unmanned vehicle l that is not included in the task team list for each task is

[0022] Furthermore, in step S3, the time required for the unmanned vehicles that are not included in the task team list and the unmanned vehicles that are included in the team list to jointly perform the task is calculated using the following formula:

[0023]

[0024] in, is the time required for the unmanned vehicle i that is not included in the task team list to jointly execute the task with the unmanned vehicle that is included in the team list, ε is the second preset threshold, α n is the task state change rate of task n, is the capability value of the unmanned vehicle i that is not included in the task team list, is the time when the unmanned vehicle i, which is not included in the task team list, arrives at the task point.

[0025] Furthermore, the relative impact value calculation formula in step S3 is:

[0026]

[0027] in, is the relative impact value of the unmanned vehicle i that is not included in the task team list on the task, Time to complete tasks for the existing team list.

[0028] Furthermore, the formula for calculating the profit in step S5 is the same as the formula for calculating the profit in step S2.

[0029] Furthermore, if the unmanned vehicle is selected for multiple tasks, it will first perform the task with the shortest estimated completion time, and then perform other tasks in sequence.

[0030] Compared with the prior art, the present invention has the following advantages:

[0031] (1) This invention fully utilizes heuristic knowledge, introduces an integer programming framework, and constructs solutions through iteration. Compared to existing technologies, it does not require a large number of iterations and complex algorithms, thus having lower computational complexity and higher real-time performance. Since MPDA problems usually need to complete tasks within a specific time window, real-time performance is a key requirement. By fully utilizing heuristic knowledge and the integer programming framework, this method can generate high-quality solutions in a shorter time, meeting real-time requirements.

[0032] (2) The present invention considers multiple factors to evaluate the value of a task, including the status of the task, the location of the UGV and the task point, the capabilities of the UGV, and the task sequence. This comprehensive consideration can more accurately assess the importance of tasks and assign them, thereby improving the efficiency and effectiveness of task execution.

[0033] (3) This invention introduces appropriate alliance redundancy to cope with the long-tail phenomenon of task states. We strengthen resource allocation by recruiting additional suitable UGVs to join the team. Even if these additional UGVs have weaker capabilities, they can still significantly speed up the completion time of the task. This strategy allows us to more fully utilize the resources of the UGVs in the team, thereby effectively improving the efficiency of overall task execution. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a schematic diagram of the process of the present invention;

[0035] Figure 2 is a graph of the value function under various circumstances;

[0036] Figure 3 This is a diagram of the long-tail phenomenon in the MPDA task;

[0037] Figure 4 The task planning scheme diagram for the typical example 1 is shown for different algorithms;

[0038] Figure 5 The diagram shows the task planning scheme of different algorithms for typical example 2. DETAILED DESCRIPTION

[0039] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0040] Example 1:

[0041] This invention relates to a multi-point dynamic assembly (MPDA) mission optimization planning method for unmanned ground vehicles (UGVs). The goal is to determine the ideal mission execution sequence for each UGV, ensuring that all mission points quickly fall below a preset state threshold ε. This involves optimizing the UGV's mission execution sequence so that all mission points can quickly reach a predetermined target state. This target state is evaluated using a specific threshold.

[0042] Specifically, assuming that the task execution sequence of the i-th UGV is represented as X i , then the planning scheme of all UGVs can be expressed as , the task planning problem can be modeled as the following integer programming framework:

[0043]

[0044] where t j (X) represents the completion time of task j, The maximum completion time is determined by the task with the longest completion time among all tasks. The constraint indicates that task j is completed when its status value drops below the threshold.

[0045] In order to avoid repeated movement of the UGV between different mission points, this method assumes that the UGV remains stationary after arriving at the assembly point until the state of the corresponding mission point reaches the expected value, and then immediately moves to the next assembly point.

[0046] This method proposes an evaluation function that takes into account multiple factors such as the task status, the location of the UGV and the task point, the UGV capability β, and the task sequence. When the UGV performs the task, the task status will decline or its state change rate will slow down. This refers to the change of the task status over time, indicating that the task is being executed, such as Figure 2 As shown in the figure, the dashed line represents the task state change when the UGV is not executing the task, and the solid line represents the task state curve when the UGV is executing the task. The gap between the two state curves is defined as the impact value Φ. Φ is considered the cumulative impact of the UGV on the task state, which refers to the change in the task state caused by the UGV executing the task. In the following, Φ is fixed to compare the time required to achieve the same impact on the task. The shorter the time, the greater the impact on the task and the greater the task benefit.

[0047] Assume that UGVi(i=1,2,...,N G ) has been executed at task n, and the time when it reaches the task point is recorded as t i , at this time the task status curve of task n for

[0048]

[0049] Assume that UGVi(i=1,2,...,N G ) has been executed at task n, and the time when it reaches the task point is recorded as t i , the capability value is β i The ability of a single UGV to complete a task is limited and different. In many cases, multiple agents need to collaborate to reach the target task point to perform the task. For example, in disaster relief, the robot's firefighting ability and search ability are limited. This value is given in advance based on expert experience. The task state curve of task n α n is the task state change rate of the current task n.

[0050] Now there is another UGV, such as UGV j , it is necessary to determine whether it also performs task n. j When it goes to perform task n, the time it arrives at the task point is t j , then the state curve of task n becomes

[0051]

[0052] When UGV j performs task n, the time it arrives at the task point is t j , the capability value is β j , the state curve of task n is

[0053] UGV j The impact value Φ on task n is

[0054]

[0055] The impact value of UGVj on task n is Φ,t e Refers to the moment when the task impact value Φ is reached

[0056] Let Φ be a constant, UGV j The value function f for task n n，j Defined as 1 / t e .f n，j Bigger means UGV j The greater the impact on task n, the greater the task benefit.

[0057] This method incorporates three heuristics into the algorithm:

[0058] 1) For a certain task point, the benefit value f obtained by completing it n The larger it is, the more important it is and the more priority it should be allocated;

[0059] 2) Based on the mission benefit value, build an alliance of UGVs to complete the MPDA mission to complete the MPDA mission as quickly as possible;

[0060] 3) Through a certain degree of alliance redundancy, the long tail phenomenon of task status can be avoided. The long tail phenomenon occurs when the task point status is declining, such as Figure 3 As shown in the figure, if the appropriate UGV is recruited into the team at this time, the task completion time will be greatly accelerated.

[0061] like Figure 1 As shown, based on the evaluation function and heuristic rules proposed in the method, the detailed execution process of this method is listed as follows:

[0062] Step S1: Identify the task with the highest value from the task list of the unassigned team list, and add the UGV with the highest benefit value for achieving the task to the team list of the task;

[0063] Step S1 specifically includes: calculating the benefit of each UGV not included in the team list for each task not assigned to the team list, and the impact value Φ of the UGV on the task: Let Φ be a constant, the upper bound of the integral is t e It can be easily obtained that the value function value of UGV for the task is f n，j The larger the value, the greater the mission benefit. Sort the benefits, and the mission with the highest benefit is the mission with the highest value, and the corresponding UGV is added to the team list of the mission.

[0064] Among them, the task with the highest value is denoted as n * , the UGV that is most suitable for completing this task is denoted as j * , N G Refers to the number of UGVs, N T Refers to the number of MPDA, f n，j Refers to the value function value of UGV j for task n.

[0065] The calculation formula for calculating the benefit of each unmanned vehicle that is not included in the team list for each task that is not assigned to the team list is:

[0066]

[0067]

[0068] Among them, t j represents the time when the autonomous vehicle j arrives at the task point, β j represents the capability value of the unmanned vehicle j, represents the task state curve, α n represents the task state change rate of the current task n, t represents the current time, Represents the state curve of task n, t e Indicates the moment when the task impact value Φ is reached, the benefit of each unmanned vehicle j for each task is f n,j , f n,j =1 / t e .

[0069] Step S2: Calculate the benefits of the UGVs that are not included in the task team list and the UGVs that are included in the team list jointly performing the task Sort the revenue from high to low, and select the top k UGVs to join the team list for the mission;

[0070] In step S2, the profit of the unmanned vehicles that are not included in the task team list and the unmanned vehicles that are included in the team list jointly performing the task is calculated using the following formula:

[0071]

[0072] Among them, t k represents the time when the unmanned vehicle k that joins the task team list arrives at the task point, β k represents the capability value of the unmanned vehicle k that joins the task team list, t l represents the time when the unmanned vehicle l that is not included in the task team list arrives at the task point, β l Indicates the capability value of the unmanned vehicle l that is not included in the task team list. represents the task state curve, α n represents the task state change rate of the current task n, t represents the current time, Represents the state curve of task n, t e Indicates the moment when the task impact value Φ is reached. The benefit of each unmanned vehicle l that is not included in the task team list for each task is

[0073] Step S3: Form a task alliance with a certain redundancy and calculate the time required for the UGVs that are not included in the task team list and the UGVs that are included in the team list to jointly execute the task. The formula is:

[0074]

[0075] in, is the time required for the unmanned vehicle i that is not included in the task team list to jointly execute the task with the unmanned vehicle that is included in the team list, ε is the second preset threshold, α n is the task state change rate of task n, is the capability value of the unmanned vehicle i that is not included in the task team list, is the time when the unmanned vehicle i, which is not included in the task team list, arrives at the task point.

[0076] Select the UGV with the shortest time and calculate the relative impact of the UGV on the task based on the shortest time required Every one not in The UGV (e.g. UGVi) calculates its ability to complete the task n together with the existing alliance * The time, recorded as The corresponding UGV is i c , Refers to the completion time of the existing alliance's tasks. The UGV (e.g. UGVi) calculates its ability to complete the task n together with the existing alliance * The time, recorded as The corresponding UGV is i c , Refers to the task completion time of the existing alliance. Compare the relative impact value with the first preset threshold TS. If it is greater than TS, the UGV is added to the team list of the task to avoid the long tail phenomenon. If it is less than or equal to TS, it is not added.

[0077] Step S4: remove the task with the highest value from the task list of the unassigned team list, and execute step S1 until the task list of the unassigned team list is empty;

[0078] Step S5: For UGVs that have not joined any team list and have completed tasks, calculate the benefits of participating in unfinished tasks. , assign it to the task with the highest benefit. The calculation formula of the benefit in step S5 is the same as the formula for calculating the benefit in step S2.

[0079] When executing tasks, the process is carried out in a similar order. If a UGV is selected for multiple tasks, it will first perform the task with the shortest estimated completion time, and then perform other tasks in sequence.

[0080] Example 2:

[0081] The parts not mentioned in this embodiment are the same as those in embodiment 1.

[0082] Two typical cases were tested. In the typical case 1, there are four mission points and four UGVs. All mission points are located at [1,1], [1,-1], [-1,-1], and [-1,1], and their change rate is 0.1. All UGVs are located at [0,0], and their capability value is 0.7. Each UGV only needs to go to the mission point closest to it to complete the mission. As expected, all four algorithms can make the UGV go to the four mission points respectively. The results are as follows: Figure 4 shown.

[0083] Example 3:

[0084] The parts not mentioned in this embodiment are the same as those in embodiment 1.

[0085] In the typical case 2, there are four mission points and four UGVs, and the initial positions of the mission points and UGV are the same as in the case 1. The change rate of all mission points is 0.6, and the capability values ​​of UGV are 0.1, 0.1, 0.2 and 0.3. The change rate of the mission is greater than that of any N G -1 UGV capability value combination, in other words, all UGVs need to complete each mission point in sequence. Figure 5 The task planning schemes of three algorithms are presented in . Figure 5 It can be seen that this method allows all UGVs to complete each task point in sequence and thus complete all tasks; while the solutions generated by EDA and MA-MLS cause each UGV to be deadlocked at each task point and unable to continue to complete the task.

[0086] Example 4:

[0087] The parts not mentioned in this embodiment are the same as those in embodiment 1.

[0088] In this invention, capability values ​​are inherent attributes of each unmanned ground vehicle (UGV). For example, in a firefighting mission, different UGVs possess varying firefighting capabilities, resulting in different capability values. MPDA missions, in this context, correspond to firefighting missions at different locations, and these missions also require different UGV capability values. This invention utilizes a heuristic task planning method based on multi-point dynamic clustering of unmanned vehicles to assess task value, taking into account multiple factors, including task status, the location of the UGV relative to the task point, the UGV's capabilities, and the task sequence. This comprehensive consideration allows for more accurate assessment of task importance and allocation, thereby improving the efficiency and effectiveness of task execution.

[0089] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and such modifications or substitutions are intended to be within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be subject to the scope of protection of the claims.

Claims

1. A multi-point dynamic cluster construction heuristic task planning method for unmanned vehicles, characterized by: The following steps are involved: Step S1: Identify the task with the highest value from the task list of the unassigned team list, and add the unmanned vehicle with the highest benefit value for achieving the task to the team list of the task; Step S2: Calculate the benefits of jointly executing the task by the unmanned vehicles that are not on the task team list and the unmanned vehicles that are on the team list, sort the benefits from high to low, and add the unmanned vehicles in the sorted order from high to low to the task team list one by one until the combined capability value of the unmanned vehicles on the task team list meets the task requirements; Step S3: Calculate the time required for the unmanned vehicles that are not on the task team list and the unmanned vehicles that are on the team list to jointly perform the task, select the unmanned vehicle with the shortest time, calculate the relative impact value of the unmanned vehicle on the task based on the shortest time, and compare the relative impact value with a first preset threshold. If the relative impact value is greater than the first preset threshold, the unmanned vehicle is added to the task team list; otherwise, it is not added. Step S4: remove the task with the highest value from the task list of the unassigned team list, and determine whether the task list in the current task list is empty. If so, execute step S5; otherwise, return to step S1; Step S5: For the unmanned vehicles that have not joined any team list and have completed tasks, calculate their benefits from participating in unfinished tasks and assign them to the task with the highest benefit.

2. The heuristic task planning method for multi-point dynamic cluster construction of unmanned vehicles according to claim 1 is characterized in that: The specific steps of step S1 include: calculating the benefits of each unmanned vehicle that is not included in the team list for each task that is not assigned to the team list, sorting the benefits, and the task corresponding to the one with the highest benefit is the task with the highest value, and the corresponding unmanned vehicle is added to the team list of the task.

3. The heuristic task planning method for multi-point dynamic cluster construction of unmanned vehicles according to claim 2 is characterized in that: The calculation formula for calculating the benefit of each unmanned vehicle that is not included in the team list for each task that is not assigned to the team list is: Among them, t j represents the time when the autonomous vehicle j arrives at the task point, β j represents the capability value of the unmanned vehicle j, represents the task state curve, α n represents the task state change rate of the current task n, t represents the current time, Represents the state curve of task n, t e Indicates the moment when the task impact value Φ is reached, the benefit of each unmanned vehicle j for each task is f n,j , f n,j =1 / t e .

4. The method for heuristic mission planning of multi-point dynamic cluster construction for unmanned vehicles according to claim 1 is characterized in that: The capability value is an inherent parameter of the unmanned vehicle.

5. The method for heuristic mission planning of multi-point dynamic cluster construction for unmanned vehicles according to claim 3 is characterized in that: The task impact value Φ is a constant.

6. The method for heuristic mission planning of multi-point dynamic cluster construction for unmanned vehicles according to claim 3 is characterized in that: In step S2, the profit of the unmanned vehicles that are not included in the task team list and the unmanned vehicles that are included in the team list jointly performing the task is calculated using the following formula: Among them, t k represents the time when the unmanned vehicle k that joins the task team list arrives at the task point, β k represents the capability value of the unmanned vehicle k that joins the task team list, t l represents the time when the unmanned vehicle l that is not included in the task team list arrives at the task point, β l Indicates the capability value of the unmanned vehicle l that is not included in the task team list. represents the task state curve, α n represents the task state change rate of the current task n, t represents the current time, Represents the state curve of task n, t e Indicates the moment when the task impact value Φ is reached. The benefit of each unmanned vehicle l that is not included in the task team list for each task is 7. The heuristic mission planning method for multi-point dynamic cluster construction of unmanned vehicles according to claim 1 is characterized in that: In step S3, the time required for the unmanned vehicles that are not included in the task team list and the unmanned vehicles that are included in the team list to jointly perform the task is calculated using the following formula: in, is the time required for the unmanned vehicle i that is not included in the task team list to jointly execute the task with the unmanned vehicle that is included in the team list, ε is the second preset threshold, α n is the task state change rate of task n, is the capability value of the unmanned vehicle i that is not included in the task team list, is the time when the unmanned vehicle i, which is not included in the task team list, arrives at the task point.

8. The heuristic mission planning method for multi-point dynamic cluster construction of unmanned vehicles according to claim 1 is characterized in that: The relative impact value calculation formula in step S3 is: in, is the relative impact value of the unmanned vehicle i that is not included in the task team list on the task, Time to complete tasks for the existing team list.

9. The method for heuristic mission planning of multi-point dynamic cluster construction for unmanned vehicles according to claim 1, characterized in that: The calculation formula for the profit in step S5 is the same as the formula for calculating the profit in step S2.

10. The heuristic mission planning method for multi-point dynamic cluster construction of unmanned vehicles according to claim 1 is characterized in that: If the unmanned vehicle is selected for multiple tasks, it will first perform the task with the shortest estimated completion time, and then perform other tasks in sequence.

Images