Multi-unmanned ship task allocation method based on consistency packet algorithm and Gaussian process model

Abstract

The invention relates to a multi-unmanned ship task allocation method based on a consistency packet algorithm. The method comprises the following steps: 1, allocating initial static tasks; 2, training a Gaussian process model through existing task distribution condition data, and predicting positions where tasks possibly appear so as to determine an area where the tasks are most likely to appear in a next period of time in the scene; 3, detecting tasks are generated at task execution intervals in the static task execution process of the unmanned ship; 4, executing a static task; and 5, when a dynamic task occurs, releasing the distributed task by the unmanned ship system, and carrying out a task distribution process again to realize timely response to an emergent task. A detection task generation algorithm is designed by means of prediction data of task distribution, some detection tasks can be generated in gaps of an original task sequence of the unmanned ship, and the detection tasks can enable the unmanned ship to go to a position with a high task occurrence probability for detection on the basis of not influencing execution of the original task, so that the detection efficiency is improved. And the quick response of the emergency dynamic task is realized.

Description

Technical Field

[0001] The invention relates to the field of unmanned boat task allocation, and in particular to a multi-unmanned boat task allocation method based on a consistency package algorithm and a Gaussian process model. Background Art

[0002] In open scenarios, Academician Zhang Ba (Global Artificial Intelligence and Robotics Summit, 2018) and Professor Zhou Zhihua (Artificial Intelligence and Industry Development Academic Report, 2018) both pointed out that artificial intelligence is currently facing the problem of how to solve the uncertainty and dynamic problems in open scenarios. Similarly, in the application scenarios of multi-unmanned boat systems, how to deal with the dynamics and unknowns of open scenarios poses new challenges to the task allocation method of multi-unmanned boat systems. Therefore, according to the known and unknown nature of the tasks and the time when the tasks appear, the tasks can be divided into two categories: the first category is static known tasks, which already exist at the time of initial task allocation, and the system knows the relevant information of these tasks. For static known tasks, the system will perform an initial allocation process to construct the original task execution sequence of the robot. The second category is dynamic unknown tasks, which do not exist at the time of initial task allocation, but appear randomly in certain locations in the scene after the robot starts to perform the task.

[0003] After long-term research, facing some static known tasks, the existing static task allocation method has been able to solve the multi-task allocation problem of multi-unmanned boat systems. However, with the increasing complexity of task scenarios, new dynamic tasks often appear after the static task allocation is completed and the robot begins to perform the task. In this scenario, how to efficiently respond to the dynamic tasks that appear in the scene has become a key issue in the current field of multi-unmanned boat task allocation.

[0004] Many scholars have conducted corresponding research on scenarios with new dynamic tasks. Chinese Patent Publication No. CN116205464A uses an improved extended consistency beam algorithm to assign tasks to multiple unmanned boats, and obtains a multi-unmanned boat task assignment effect diagram. Chinese Patent Publication No. CN113723805B further decomposes the strict dual-boat completion of the composite task into sub-tasks that must be completed by two types of unmanned boats in cooperation with each other, thereby enabling the unmanned boat task assignment scheme to cope with more complex real-world combat environment task requirements. Chinese Patent Publication No. CN202211418037.7 considers the navigation performance and detection capabilities of different types of surface unmanned boats for the multi-surface unmanned boat area detection task, uses a genetic algorithm to assign a detection area to each surface unmanned boat, and ensures the detection efficiency, so that the entire multi-surface unmanned boat formation has the highest detection efficiency. Although good results have been achieved in their respective scenarios, robots can perform these dynamic tasks to a certain extent. However, a key attribute of dynamic tasks is that they have certain execution time limits. In many current works related to multi-unmanned boat task allocation, only the situation where dynamic tasks exist is considered, but there is a lack of consideration for the existence of dynamic tasks and emergency situations. Especially in scenes like roundups, tasks usually appear suddenly in the scene and are relatively urgent, requiring the unmanned boat system to reach the task point in a short time and start executing the task. If the unmanned boat cannot execute such emergency dynamic tasks in time, it will cause certain revenue losses to the system. Therefore, it is particularly important to improve the responsiveness of the multi-unmanned boat task allocation algorithm to emergency dynamic tasks. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention proposes a multi-unmanned boat task allocation method based on a consistency package algorithm and a Gaussian process model, which is used to solve the problem that the existing task allocation method has poor response ability to emergency dynamic tasks.

[0006] The above-mentioned object of the present invention is achieved by the following technical solutions:

[0007] A method for allocating tasks of multiple unmanned boats based on a consistency pack algorithm and a Gaussian process model, characterized in that it comprises the following steps:

[0008] Step 1: Based on the time constraint characteristics of the static tasks, the consistency package algorithm is used to initially allocate the initially existing static tasks;

[0009] Step 2: Train the Gaussian process model using the existing task distribution data, and use the Gaussian process model to predict where tasks may appear, so as to determine the area in the scene where tasks are most likely to appear in the next period of time;

[0010] Step 3, producing a detection task during a task execution interval that exists when the unmanned boat performs a static task;

[0011] Step 4: Execute static tasks;

[0012] Step 5: When a dynamic task occurs, the unmanned boat system releases the assigned tasks and re-assigns the tasks to achieve timely response to emergency tasks.

[0013] Moreover, in step 1, the time constraint characteristics include the execution time window, the mission start and end time, and the locations of the multiple unmanned boats.

[0014] Furthermore, step 1 includes the following steps:

[0015] Step 1.1 Each unmanned boat will create a mission package and continue to add tasks to the mission package until there are no suitable tasks;

[0016] Step 1.2 After the initial task package construction phase, each unmanned boat forms its own task execution sequence; when multiple unmanned boats win the bid for the same task, a task allocation conflict occurs, and go to step 1.3;

[0017] Step 1.3: Conflict resolution is performed through information exchange between unmanned boats to ensure that each mission is performed by only one unmanned boat;

[0018] Step 1.4 After completing the conflict resolution phase, return to the task package construction phase and re-add all unassigned tasks that meet the time constraints; the two phases of task assignment and conflict resolution are iterated continuously until the bid result no longer changes. Each unmanned boat forms its own task package and corresponding task execution order.

[0019] Furthermore, step 2 includes the following steps:

[0020] Step 2.1 first divide the two-dimensional scene into grid areas, and the size of the grid area is divided according to the needs of the current task and the remaining energy of the unmanned boat;

[0021] Step 2.2: Select the areas where tasks have appeared in the scene as the training set, and other areas where tasks have not appeared as the prediction set, and use the Gaussian process model to model the distribution of tasks;

[0022] Step 2.3 uses the existing task distribution data to train the Gaussian process model to predict the probability of tasks appearing in areas where no tasks appear in the scene.

[0023] Furthermore, step 3 includes the following steps:

[0024] Step 3.1, according to the allocation result of the static task, determine the task execution interval existing in the process of the unmanned boat performing the static task;

[0025] Step 3.2 selects the time when the unmanned boat waits for the task to start from the task execution interval, and then independently generates a series of detection tasks between the original task sequences of each unmanned boat;

[0026] Step 3.3 judges and screens a series of detection tasks generated in step 3.2. The screening criterion is that the generated detection task cannot affect the execution of the original static task. Otherwise, the detection task is regarded as an invalid task and deleted.

[0027] Furthermore, step 4 includes the following steps:

[0028] Step 4.1: According to the initial static task allocation result and the detection task, the unmanned boat starts to perform the task;

[0029] Step 4.2: When the mission is actually being performed, the detection mission will cause the UAV to expand its range of movement, so that the UAV will go to the area with a higher probability of the mission to conduct detection, but it will not stay in the area. That is, the detection mission is considered completed when the UAV reaches the mission location.

[0030] Furthermore, step 5 includes the following steps:

[0031] Step 5.1 The multi-unmanned boat system releases the existing task execution sequence and re-assigns the task;

[0032] Step 5.2: Add the newly emerged task distribution to the task distribution dataset and retrain to obtain a new task allocation Gaussian process model;

[0033] Step 5.3 The unmanned boat system responds immediately to the emergency dynamic task that occurs, restarts a new round of task execution process, and thus performs a cyclic task response process.

[0034] The advantages and positive effects of the present invention are:

[0035] 1. Aiming at the problem of rapid discovery and response to dynamic unknown tasks in the application of unmanned boat system, the present invention designs a multi-task dynamic allocation strategy and method for multi-unmanned boat collaboration to meet the respective time constraints in the execution of static known tasks and dynamic unknown tasks.

[0036] 2. The present invention uses the location data of historical tasks to train the Gaussian process model, and uses the historical data of task occurrence to predict the location where the task may appear in the future. And gives the probability of the task appearing at each location. These data can provide a forward-looking basis for the allocation and execution planning of tasks.

[0037] 3. The present invention designs a detection task generation algorithm with the help of the predicted data of task distribution. The algorithm can generate some detection tasks in the gaps of the original task sequence of the unmanned boat. These detection tasks can enable the unmanned boat to detect at locations with a higher probability of task occurrence without affecting the execution of the original tasks, thereby achieving rapid response to emergency dynamic tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is the static task allocation process of the present invention;

[0039] Figure 2 It is the task distribution prediction process of the present invention;

[0040] Figure 3 It is the detection task generation process of the present invention;

[0041] Figure 4 is a flow chart of the detection task generation algorithm of the present invention;

[0042] Figure 5 It is an example task execution timing diagram of the present invention. DETAILED DESCRIPTION

[0043] The structure of the present invention is further described below with reference to the accompanying drawings and by way of examples. It should be noted that the present examples are descriptive rather than restrictive.

[0044] In current practical applications, the speed of unmanned boats is limited, and emergency dynamic tasks need to be executed in a short time. The time it takes to move to the task location becomes a key bottleneck that limits the unmanned boats from executing such tasks. If the locations where these emergency dynamic tasks may occur can be predicted, through the frequent visits of unmanned boats to these locations, there is a high probability that the time it takes for the unmanned boats to reach the task point can be effectively reduced after the task occurs, thereby improving the successful execution rate of emergency dynamic tasks. In reality, whether it is a task caused by natural phenomena or man-made events, it often reflects a certain spatial correlation. Therefore, the location of the task can be predicted based on the spatial position relationship between tasks. Combined with the existing multi-unmanned boat task allocation algorithm, the emergency tasks that may occur in the future are taken into account in the early stage of task allocation, and they are given special attention. This can improve the timely response capability of the multi-unmanned boat system to emergency dynamic tasks to a certain extent, and can also further improve the overall benefits of the system.

[0045] Based on the above-mentioned ideas for solving the problem, the present invention is aimed at the field of multi-unmanned boat task allocation, aiming to improve the existing task allocation algorithm by using the consistency package algorithm and the Gaussian process model. In order to effectively improve the response capability of the multi-unmanned boat system to emergency dynamic tasks, the present invention proposes a multi-unmanned boat task allocation method based on the consistency package algorithm and the Gaussian process model, which mainly includes the steps of static task allocation, task distribution space prediction and detection task generation. The specific steps are as follows:

[0046] Step 1: Based on the time constraint characteristics of static tasks, the consistency package algorithm is used to initially allocate the initial static tasks:

[0047] The initial allocation of tasks is made based on the time constraint characteristics of static tasks, including the execution time window, the start and end time of the task, and the location of multiple unmanned boats. The goal of the allocation is to maximize the benefits of the unmanned boats performing the tasks. The benefits will take different forms depending on the nature of the tasks. For example, the number of rescuers in a rescue mission, the number of successful resupply in a resupply mission, etc. It includes the following steps:

[0048] Step 1.1 Each unmanned boat will create a mission package and continue to add tasks to the mission package until there are no suitable tasks.

[0049] Step 1.2 After the initial task package construction phase, each unmanned boat forms its own task execution sequence; when multiple unmanned boats win the bid for the same task, a task allocation conflict occurs, and go to step 1.3.

[0050] Step 1.3: In order to finally obtain a conflict-free allocation result, conflicts are resolved through information exchange between unmanned boats, that is, to ensure that each task is performed by only one unmanned boat.

[0051] Step 1.4 After completing the conflict resolution phase, return to the task package construction phase and re-add all unassigned tasks that meet the time constraints; the two phases of task assignment and conflict resolution are continuously iterated until the bid result no longer changes; each unmanned boat forms its own task package and corresponding task execution order.

[0052] Step 2: Task distribution space prediction

[0053] The Gaussian process model is trained using the existing task distribution data. Several hyperparameters can be determined by maximizing the marginal likelihood function. The Gaussian process model is used to predict the possible locations of tasks to determine the areas in the scene where tasks are most likely to appear in the next period of time. The steps include:

[0054] Step 2.1 First, divide the two-dimensional scene into grid areas. The size of the grid area depends on the task prediction accuracy. If a more accurate task distribution prediction is required, a smaller grid area is required, but this may bring more data sampling points, which means a greater amount of calculation for the unmanned boat. Therefore, the size of the grid area needs to be divided according to the needs of the current task and the remaining energy of the unmanned boat.

[0055] Step 2.2: Select the areas where tasks have appeared in the scene as the training set, and use other areas where tasks have not appeared as the prediction set, and use the Gaussian process model to model the distribution of tasks.

[0056] Step 2.3 uses the existing task distribution data to train the Gaussian process model to predict the probability of tasks appearing in areas where no tasks appear in the scene.

[0057] Step 3: Generate detection tasks

[0058] Step 3.1: According to the allocation result of the static task, determine the task execution interval of the unmanned boat when performing the static task.

[0059] Step 3.2 selects the time when the unmanned boat waits for the task to start from the task execution interval, and then independently generates a series of detection tasks between the original task sequences of each unmanned boat.

[0060] Step 3.3 judges and screens a series of detection tasks generated in step 3.2. The screening criterion is that the generated detection task cannot affect the execution of the original static task. Otherwise, the detection task is regarded as an invalid task and deleted.

[0061] Step 4: Execute static tasks

[0062] Step 4.1: According to the initial static task allocation results and detection tasks, the unmanned boat starts to perform the task.

[0063] Step 4.2: When the mission is actually being performed, the detection mission will cause the UAV to expand its range of movement, so that the UAV will go to the area with a higher probability of the mission to conduct detection, but it will not stay in the area. That is, the detection mission is considered completed when the UAV reaches the mission location.

[0064] Step 5: Task re-planning to respond to dynamic tasks

[0065] When a dynamic task occurs, the unmanned boat system needs to release the assigned task and re-assign the task to ensure that it can respond to the sudden task more promptly. The specific steps include:

[0066] Step 5.1 The multi-unmanned boat system releases the existing task execution sequence and re-assigns the task.

[0067] Step 5.2 adds the newly emerged task distribution to the task distribution dataset and retrains to obtain a new task allocation Gaussian process model.

[0068] Step 5.3 The unmanned boat system responds immediately to the emergency dynamic task that occurs, restarts a new round of task execution process, and thus performs a cyclic task response process.

[0069] Although the embodiments and drawings of the present invention are disclosed for illustrative purposes, those skilled in the art will appreciate that various substitutions, changes and modifications are possible without departing from the spirit and scope of the present invention and the appended claims. Therefore, the scope of the present invention is not limited to the contents disclosed in the embodiments and drawings.

Claims

1. A multi-unmanned boat task allocation method based on consistency package algorithm and Gaussian process model, characterized in that: The steps include: Step 1: Based on the time constraint characteristics of the static tasks, the consistency package algorithm is used to initially allocate the initially existing static tasks; Step 2: Train the Gaussian process model using the existing task distribution data, and use the Gaussian process model to predict where tasks may appear, so as to determine the area in the scene where tasks are most likely to appear in the next period of time; Step 3, producing a detection task during a task execution interval that exists when the unmanned boat performs a static task; Step 4: Execute static tasks; Step 5: When a dynamic task occurs, the unmanned boat system releases the assigned tasks and re-assigns the tasks to achieve timely response to emergency tasks.

2. The method for allocating tasks of multiple unmanned boats based on consistency package algorithm and Gaussian process model according to claim 1 is characterized in that: In step 1, the time constraint characteristics include the execution time window, the mission start and end time, and the location of the multiple unmanned boats.

3. The multi-unmanned boat task allocation method based on consistency package algorithm and Gaussian process model according to claim 1 is characterized in that: Step 1 includes the following steps: Step 1.1 Each unmanned boat will create a mission package and continue to add tasks to the mission package until there are no suitable tasks; Step 1.2 After the initial task package construction phase, each unmanned boat forms its own task execution sequence; when multiple unmanned boats win the bid for the same task, a task allocation conflict occurs, and go to step 1.3; Step 1.3: Conflict resolution is performed through information exchange between unmanned boats to ensure that each mission is performed by only one unmanned boat; Step 1.4 After completing the conflict resolution phase, return to the task package construction phase and re-add all unassigned tasks that meet the time constraints; the two phases of task assignment and conflict resolution are iterated continuously until the bid result no longer changes. Each unmanned boat forms its own task package and corresponding task execution order.

4. The multi-unmanned boat task allocation method based on consistency package algorithm and Gaussian process model according to claim 1 is characterized in that: Step 2 includes the following steps: Step 2.1 first divide the two-dimensional scene into grid areas, and the size of the grid area is divided according to the needs of the current task and the remaining energy of the unmanned boat; Step 2.2: Select the areas where tasks have appeared in the scene as the training set, and other areas where tasks have not appeared as the prediction set, and use the Gaussian process model to model the distribution of tasks; Step 2.3 uses the existing task distribution data to train the Gaussian process model to predict the probability of tasks appearing in areas where no tasks appear in the scene.

5. The method for allocating tasks of multiple unmanned boats based on consistency package algorithm and Gaussian process model according to claim 1 is characterized in that: Step 3 includes the following steps: Step 3.1, according to the allocation result of the static task, determine the task execution interval existing in the process of the unmanned boat performing the static task; Step 3.2 selects the time when the unmanned boat waits for the task to start from the task execution interval, and then independently generates a series of detection tasks between the original task sequences of each unmanned boat; Step 3.3 judges and screens a series of detection tasks generated in step 3.

2. The screening criterion is that the generated detection task cannot affect the execution of the original static task. Otherwise, the detection task is regarded as an invalid task and deleted.

6. The method for allocating tasks of multiple unmanned boats based on consistency package algorithm and Gaussian process model according to claim 1 is characterized in that: Step 4 includes the following steps: Step 4.1: According to the initial static task allocation result and the detection task, the unmanned boat starts to perform the task; Step 4.2: When the mission is actually being performed, the detection mission will cause the UAV to expand its range of movement, so that the UAV will go to the area with a higher probability of the mission to conduct detection, but it will not stay in the area. That is, the detection mission is considered completed when the UAV reaches the mission location.

7. The method for allocating tasks of multiple unmanned boats based on consistency package algorithm and Gaussian process model according to claim 1 is characterized in that: Step 5 includes the following steps: Step 5.1 The multi-unmanned boat system releases the existing task execution sequence and re-assigns the task; Step 5.2: Add the newly emerged task distribution to the task distribution dataset and retrain to obtain a new task allocation Gaussian process model; Step 5.3 The unmanned boat system responds immediately to the emergency dynamic task that occurs, restarts a new round of task execution process, and thus performs a cyclic task response process.

Images