A hybrid active planning method and system for space tasks based on multi-agent

Abstract

The present invention relates to the field of aerospace mission planning technology, and specifically discloses a method and system for hybrid active planning of space missions based on multi-agents, the method comprising: receiving a first resource allocation scheme generated by each agent according to corresponding space mission planning demand information; optimizing and adjusting all first resource allocation schemes to obtain a target resource allocation scheme for each agent, and generating space experiment application information corresponding to each target resource allocation scheme; when each space experiment application information meets the corresponding demand, obtaining a space mission planning plan generated by each agent according to the corresponding space experiment application information. The present invention solves the design defects of centralized planning, and utilizes a distributed computing system composed of multiple agents to quickly adapt to changes in environmental conditions, make independent and parallel decisions, and effectively improve the efficiency of hybrid active planning of space missions.

Description

Technical Field

[0001] The present invention relates to the technical field of aerospace mission planning, and in particular to a multi-agent-based hybrid active planning method and system for space missions. Background Art

[0002] The purpose of establishing a space station is to support long-term large-scale space science experiments and technical experiments, and to promote the development of space science, life science and other space technologies. And what accomplishes this mission is the various payloads carried by the space station. Payloads refer to instruments and equipment loaded on spacecraft to directly realize specific tasks to be completed by the spacecraft in orbit, and are driven by instructions to complete various target actions. The implementation of space science experiments requires the cooperation of various payloads to complete the target actions according to the instruction sequence. At the same time, the execution of instructions will also consume resources such as power consumption and storage, so it is necessary to uniformly plan various resources before conducting experiments.

[0003] The implementation process of space science experiment tasks is essentially a hybrid active interaction mode, in which humans and machines give each other opinions and lead the tasks. This is because considering the construction goals of the space station, the timing and type of experimental tasks cannot be separated from scientific user planning, which provides initial input for the machine. Similarly, coordinating resource status, recommending executable plans, and executing the final implementation must be done by the machine. However, at present, there is a relative lack of research on the machine's formulation of recommended plans for space experiment tasks. The planning of space experiment tasks still relies on a unified terminal device. When multiple payloads are running at the same time, they will face problems such as short planning cycles, multiple real-time constraints, and complex constraint propagation. It is difficult to propose a universal and efficient constraint conflict resolution strategy in the planning process.

[0004] Therefore, it is urgent to provide a technical solution to solve the above problems. Summary of the invention

[0005] In order to solve the above technical problems, the present invention provides a multi-agent based hybrid active planning method and system for spatial tasks.

[0006] In a first aspect, the present invention provides a multi-agent-based spatial task hybrid active planning method, the technical solution of which is as follows:

[0007] Receiving a first resource allocation plan generated by each agent according to corresponding space mission planning requirement information;

[0008] Optimize and adjust all the first resource allocation schemes to obtain the target resource allocation scheme of each agent, and generate space experiment application information corresponding to each target resource allocation scheme;

[0009] When each space experiment application information meets the corresponding requirements, the space mission planning plan generated by each intelligent agent according to the corresponding space experiment application information is obtained.

[0010] The beneficial effects of the multi-agent-based spatial task hybrid active planning method of the present invention are as follows:

[0011] The method of the present invention solves the design defects of centralized planning. The distributed computing system composed of multiple intelligent agents can quickly adapt to changes in environmental conditions, make independent and parallel decisions, and effectively improve the efficiency of hybrid active planning of space tasks.

[0012] Based on the above scheme, the hybrid active planning method of spatial tasks based on multi-agent of the present invention can also be improved as follows.

[0013] In an optional manner, any space task planning requirement information includes: a set of space tasks to be planned, prerequisites for executing space tasks, a space task decomposition method, resources required for space tasks, and space task planning constraints.

[0014] In an optional manner, any agent includes: an HTN planner, a COP optimizer, and an experiment generator; the step of any agent generating a first resource allocation scheme according to corresponding space mission planning requirement information includes:

[0015] The HTN planner of any of the agents generates a spatial task execution plan based on the spatial task set to be planned, the spatial task execution prerequisites and the spatial task decomposition method input by the user, and sends the plan to the COP optimizer of any of the agents;

[0016] The COP optimizer of any of the intelligent agents generates a first resource allocation plan for any of the intelligent agents based on the resources required for the space mission and the space mission planning constraints.

[0017] In an optional manner, the step of generating a spatial task execution plan by the HTN planner of any agent according to the set of spatial tasks to be planned, the spatial task execution prerequisites and the spatial task decomposition method input by the user includes:

[0018] Initialize and set an empty planning task list, and determine whether any spatial task in the to-be-planned spatial task set is an atomic task;

[0019] If any of the space tasks is an atomic task, the instructions of the space task are determined and added to the planned task list, and according to the status list information deleted or added after the space task is executed, the current network status is updated and the space task is deleted from the planned task list;

[0020] If any of the space tasks is not an atomic task, the state information of the space task is matched with the space task execution prerequisite; if the match is successful, the space task is divided into multiple sub-space tasks using the space task decomposition method and added to the planned task list; if the match is unsuccessful, the space task is added to the planned task list;

[0021] Repeat the process of determining whether any spatial task in the set of spatial tasks to be planned is an atomic task, until each spatial task in the set of spatial tasks to be planned is determined, a final list of planned tasks is obtained, and the final list of planned tasks is determined as the spatial task execution plan.

[0022] In an optional manner, the step of generating a first resource allocation scheme of any of the agents by the COP optimizer of any of the agents according to the resources required for the space mission and the space mission planning constraints includes:

[0023] The COP optimizer of any of the intelligent agents determines a distributed constraint optimization algorithm according to the spatial task planning constraints, and generates a first resource allocation plan for any of the intelligent agents according to the resources required for the spatial task and the distributed constraint optimization algorithm.

[0024] In an optional manner, the step of optimizing and adjusting all first resource allocation schemes to obtain a target resource allocation scheme for each agent includes:

[0025] Based on the first-in-first-out strategy, all the first resource allocation plans are optimized and adjusted to obtain the target resource allocation plan for each agent.

[0026] In an optional manner, the method further includes:

[0027] The spatial task planning plan of each agent is output and displayed.

[0028] In an optional manner, the method further includes:

[0029] When any space experiment application information does not meet the corresponding requirements, based on the updated space task planning requirement information obtained by each intelligent agent, return to the step of receiving the first resource allocation plan generated by each intelligent agent according to the corresponding space task planning requirement information until the space task planning plan generated by each intelligent agent according to the corresponding space experiment application information is obtained.

[0030] In a second aspect, the present invention provides a multi-agent-based spatial task hybrid active planning system, the technical solution of which is as follows:

[0031] Includes: receiving module, optimization module and planning module;

[0032] The receiving module is used to: receive a first resource allocation plan generated by each agent according to corresponding space mission planning requirement information;

[0033] The optimization module is used to optimize and adjust all the first resource allocation schemes to obtain the target resource allocation scheme of each agent, and generate space experiment application information corresponding to each target resource allocation scheme;

[0034] The planning module is used to obtain the space mission planning plan generated by each intelligent agent according to the corresponding space experiment application information when the space experiment application information meets the corresponding requirements.

[0035] The beneficial effects of the multi-agent-based spatial task hybrid active planning system of the present invention are as follows:

[0036] The system of the present invention solves the design defects of centralized planning. The distributed computing system composed of multiple intelligent agents can quickly adapt to changes in environmental conditions, make independent and parallel decisions, and effectively improve the efficiency of hybrid active planning of space tasks.

[0037] Based on the above scheme, the multi-agent-based spatial task hybrid active planning system of the present invention can also be improved as follows.

[0038] In an optional manner, any space task planning requirement information includes: a set of space tasks to be planned, prerequisites for executing space tasks, a space task decomposition method, resources required for space tasks, and space task planning constraints.

[0039] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The accompanying drawings are only used to illustrate the embodiments and are not to be considered as limiting the present invention. In addition, the same reference symbols are used to represent the same components throughout the accompanying drawings. In the accompanying drawings:

[0041] Figure 1 A schematic diagram of a flow chart of an embodiment of a multi-agent-based spatial task hybrid active planning method of the present invention;

[0042] Figure 2 This is a schematic diagram of the overall architecture;

[0043] Figure 3 It is a flowchart of the hierarchical task network algorithm;

[0044] Figure 4This is a schematic diagram of the decomposition of the hierarchical task network algorithm;

[0045] Figure 5 This is a schematic diagram of the principle of multi-agent resource allocation problem;

[0046] Figure 6 Schematic diagram of model transformation and search space in the solution algorithm;

[0047] Figure 7 Pseudo-code diagram for the agent's local constraint processing flow;

[0048] Figure 8 It is a schematic diagram of the resource information interaction process between intelligent agents;

[0049] Fig. 9 It is a structural schematic diagram of an embodiment of a multi-agent based spatial task hybrid active planning system of the present invention. DETAILED DESCRIPTION

[0050] Exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited to the embodiments set forth herein.

[0051] Figure 1 FIG. 1 is a flow chart showing an embodiment of a method for hybrid active planning of a space task based on multi-agents provided by the present invention, which is executed by a processor. Figure 1 As shown, the following steps are included:

[0052] S1. Receive the first resource allocation plan generated by each intelligent agent according to the corresponding space mission planning requirement information.

[0053] The agent refers to an intelligent entity that can perceive the environment, make decisions, and guide the controller to complete specific mission objectives. In this embodiment, it refers to the ground terminal used by the payload to carry out scientific experiments. Figure 2 As shown, each agent includes: HTN (Hierarchical Task Network) planner, COP optimizer and experiment generator. The space task planning requirement information includes: the set of space tasks to be planned, the prerequisites for space task execution, the space task decomposition method, the resources required for space tasks and the space task planning constraints.

[0054] S2. Optimize and adjust all the first resource allocation schemes to obtain the target resource allocation scheme of each intelligent agent, and generate space experiment application information corresponding to each target resource allocation scheme.

[0055] The resource information is shared and interacted among multiple agents through optimization and adjustment. The target resource allocation scheme is the resource allocation scheme after optimization and adjustment.

[0056] S3. When each space experiment application information meets the corresponding requirements, obtain the space mission planning plan generated by each intelligent agent according to the corresponding space experiment application information.

[0057] Among them, when the user confirms that the space experiment application information of each intelligent agent meets the corresponding needs, the space mission planning plan generated by each intelligent agent according to the corresponding space experiment application information is obtained.

[0058] It should be noted that the processor used in this embodiment is equivalent to a dynamic database for storing solution state data and control data in the solution process, and the blackboard is used by default in this embodiment. Each intelligent agent performs hybrid active planning of space tasks according to the space task planning plan.

[0059] In an optional manner, the step of generating a first resource allocation solution by any agent according to corresponding space mission planning requirement information includes:

[0060] The HTN planner of any of the intelligent agents generates a spatial task execution plan based on the spatial task set to be planned, the spatial task execution prerequisites and the spatial task decomposition method input by the user, and sends it to the COP optimizer of any of the intelligent agents.

[0061] The space task set to be planned is a space experiment task to be carried out that is planned by the user in units of weeks or months, and the space task set to be planned includes multiple space tasks.

[0062] Specifically, Figure 3 As shown, the HTN planner of any agent generates a space task execution plan according to the space task set to be planned, the space task execution prerequisite and the space task decomposition method input by the user, including:

[0063] Initialize an empty planning task list and determine whether any spatial task in the to-be-planned spatial task set is an atomic task.

[0064] The HTN planner refers to a module that uses hierarchical task network technology to perform automatic planning calculations. It gives the dependencies between actions in the form of a network, obtains the planning tasks, task decomposition methods, and task execution prerequisites input by the user, recommends executable solutions for the generated tasks, and sends them to the COP optimizer. Before initializing an empty planning task list, the following points need to be explained:

[0065] ①An atomic task refers to a task that cannot be decomposed any further. In this model, it refers to an executable control instruction, which is expressed as (action(para), P, D, A); action(para) indicates the name of the atomic task, para indicates the input parameter of the atomic task, which is the event parameter value contained in the instruction; P indicates the prerequisite for the execution of the atomic task, which can be judged based on multiple indicators such as telemetry, engineering parameters, application data, images, characterization analysis, etc.; D indicates the status list information that needs to be deleted after the atomic task is executed; A indicates the status list information that needs to be added after the atomic task is executed. D and A are used to maintain the operating status, resource consumption and other status information of each payload in the space station system after the atomic task is executed, which is the basis for the ground end of the payload to decide whether to execute it in the end.

[0066] ② If Figure 4 As shown, a composite task is a task that can be further decomposed, also known as a method. A composite task can be decomposed into composite tasks or atomic tasks, using (task, P i ,T i ,…) expression; task represents the name of the composite task (spatial task); each P i It represents a prerequisite for a composite task, which can be judged based on multiple indicators such as telemetry, engineering parameters, application data, images, characterization analysis, etc. i It represents a set of partially ordered subtasks (decomposable spatial tasks), corresponding to satisfying P i Decomposition tasks performed under prerequisites.

[0067] ③The HTN planning method model can be represented by a triple (S, T, E), where S represents the initial planning state, T represents the set of spatial tasks to be planned, which is a set of spatial experimental tasks submitted by users; E represents the knowledge domain information, which is the task decomposition method set Method and instruction set C provided to users.

[0068] ④ Method Method describes the decomposition process of non-atomic tasks. The definition of the method is consistent with that of the composite task, which is Method = {name, pre1, subtask1, pre2, subtask2, ...}, where name represents the method name; pre i Indicates the prerequisite for executing the i-th spatial task, expressed as a logical expression; subtask i Indicates a sublist of tasks that can be decomposed if the prerequisite is met. A conditional selection structure is adopted, that is, if condition 1 is met, the method is decomposed according to the subspace task list subtask1; if condition 1 is not met, but condition 2 is met, the method is decomposed according to the subspace task list subtask2, and so on:

[0069]

[0070] Specifically, an empty planning task list P is initialized and set, and it is determined whether any spatial task t in the set of spatial tasks to be planned is an atomic task.

[0071] If any of the space tasks is an atomic task, the instructions of the space task are determined and added to the planned task list. According to the status list information deleted or added after the execution of the space task, the current network status is updated and the space task is deleted from the planned task list.

[0072] Specifically, if the spatial task t is an atomic task, select the instruction c that can complete the task t in the current state from the instruction set C in the knowledge domain E, and add the instruction c to the planning list P. According to the state list information deleted or added after the atomic task is executed, update the current network state S, and delete the task t from the planning task list after the spatial task t is decomposed.

[0073] If any of the spatial tasks is not an atomic task, the status information of the spatial task is matched with the prerequisite for the execution of the spatial task; if the match is successful, the spatial task is divided into multiple sub-space tasks using the spatial task decomposition method and added to the planned task list; if the match is unsuccessful, the spatial task is added to the planned task list.

[0074] Among them, the status information of the space mission includes: telemetry, engineering parameters, and analysis conclusions of the application data image intelligent characterization platform.

[0075] Specifically, if the space task t is not an atomic task, obtain the state information such as telemetry, engineering parameters, application data images and intelligent characterization platform analysis conclusions, and match them with the space task execution prerequisites in the task method set Method in the knowledge domain E provided by the scientific user; if the match is successful (the state information meets the prerequisites), then use the space task decomposition method to decompose the space task t into the space task execution prerequisites preconditions that meet the method Method i Subtask i , and replace task t in the planning task list with the subspace task, and the composite task decomposition is completed once, otherwise failure is returned.

[0076] Repeat the process of determining whether any spatial task in the set of spatial tasks to be planned is an atomic task, until each spatial task in the set of spatial tasks to be planned is determined, a final list of planned tasks is obtained, and the final list of planned tasks is determined as the spatial task execution plan.

[0077] Specifically, it is repeatedly determined whether each spatial task in the set of spatial tasks to be planned is an atomic task until all tasks in the task list T are executed, that is, all spatial tasks are decomposed into atomic tasks, the entire planning process is completed, and the final planning task list P is output, and the final planning task list is determined as the spatial task execution plan.

[0078] It should be noted that in the final planning task list P, P represents all the operation instructions required to complete all the spatial tasks in the spatial task set T to be planned. If during the planning process, a task that cannot be decomposed or does not conform to the planning definition is encountered, it cannot be decomposed using the above planning method and exception handling is required.

[0079] The COP optimizer of any of the agents generates a first resource allocation plan for any of the agents based on the resources required for the space mission and the space mission planning constraints. Specifically:

[0080] The COP optimizer of any of the intelligent agents determines a distributed constraint optimization algorithm according to the spatial task planning constraints, and generates a first resource allocation plan for any of the intelligent agents according to the resources required for the spatial task and the distributed constraint optimization algorithm.

[0081] Among them, the COP optimizer refers to the module that solves the constrained optimization problem. It obtains the system resource status and calls the optimization algorithm according to the planning constraints entered in advance by the scientific user to process the constraints locally in the intelligent agent to generate a resource allocation plan.

[0082] It should be noted that the process of processing through the distributed constraint optimization algorithm is as follows:

[0083] ① Establish a multi-agent resource allocation problem model.

[0084] The multi-agent resource allocation problem can be modeled as a DCOP model, which includes the following elements. The multi-agent set composed of the payload ground terminal is denoted as A = {A1, A2, ..., A n}; The set of variables owned by all agents A in the resource allocation problem planning model is denoted as R = {R1, R2, ..., R P}, are the various resources consumed by multi-agents to perform tasks, such as variables such as experimental status, time, and resources; the set of spatial tasks to be planned recommended by the HTN planning method for each agent is denoted as F i ={T i1 ,T i2 ,…,T ic}; F i represents the set of spatial tasks to be planned for the ith agent, T icrepresents the cth spatial task to be planned in the set of spatial tasks to be planned of the ith agent. The range of variables is denoted as V = {V1, V2, …, V p}, R j ∈V j , is the upper limit of various resources owned by the space station; the constraints between variables are recorded as C = {C1,C1,…,C m}.

[0085] Due to the limitations of communication range and user privacy, each agent cannot obtain global constraints and can only process constraints based on its own information or local information. The essence of the multi-agent resource allocation problem is the resources that the agents compete for. Specifically, the HTN planning recommends the experimental plan for each agent to obtain a task set. The resource requirements of the agent can be extracted based on the resource consumption of each task. There are constraints between resources, and there is competition for the same resource between agents. This can lead to a series of inequalities about resource constraints. That is: g(R ji ) is about the variable R ji function. Figure 5 For example, the spatial task sets recommended by the HTN planning method for each agent to be planned are T1 = {T 11 ,T 12 ,T 13}、T2={T 21}、T3={T 31 、T 32}, where the space task set to be planned by agent A1 needs to consume three types of resources: R1, R2, and R3, and the constraints are: The set of spatial tasks to be planned by agent A2 needs to consume two types of resources, R1 and R2, with the following constraints: The space task set to be planned by agent A3 needs to consume two types of resources R3, and the constraints are: {2≤R 33 ≤10; Considering the space station resource constraints, the constraints between agents are:

[0086] The purpose of optimizing the multi-agent resource allocation problem model is to reasonably allocate resources, resolve resource conflicts between agents, and make the experimental task plan meet the needs of scientific users to the greatest extent. Therefore, the present invention defines an effect function set F = {f1, f2, ..., f m}, effect function f j (R j ) is about the variable R jThe larger the function value, the higher the user's satisfaction with the variable value. Agents can maximize the global objective function by assigning values ​​to variables. Let F represent the sum of a set of effect functions. The objective function of the DCOP problem is:

[0087] ②Solving the multi-agent resource allocation problem model.

[0088] The solution process of the DCOP algorithm is mainly divided into two stages: the first stage is that the agents perform constraint processing locally, that is, first solve the COP of each agent. First, the model seen needs to be preprocessed, the model transformation and search space construction are performed, and then the solution strategy is designed for local solution. The second stage is to consider the scope of agent message transmission and optimize the constraints between agents.

[0089] Regarding the model transformation in the first stage of the solution process, the COP problem is transformed into a constraint diagram. Figure 6 As shown in (a), a node represents a resource, and an edge represents a constraint relationship between two resources. The search space in the first stage of the solution process is to determine the order relationship between variables. A pseudo tree is constructed based on the constraint graph, such as Figure 6 As shown in (b), the algorithm can perform parallel searches on independent branches.

[0090] Regarding the solution strategy in the first stage of the solution process, the present invention selects the branch and bound search algorithm, and its search strategy is to first define the resource variable list L as empty. Select a root node R0, add the value of R0 to the resource list L, calculate the function value of f(R0), and calculate the sum of the function value of each child node and the function value of f(R0) in all child node sets S of R0. According to the size of the sum, select the node with the largest sum from the current child nodes as the expansion node R. add . Then expand the node R add As the search starting point, the next expansion node is selected from its child nodes in the above manner, and the above process is repeated until the optimal resource allocation solution L that satisfies the constraints is obtained. The pseudo code of the algorithm is as follows Figure 7 shown.

[0091] Regarding the second stage of the solution process, Figure 8As shown in the figure, by introducing the blackboard model structure, the first-in-first-out strategy is adopted to eliminate resource conflicts between agents. The blackboard data structure is a parallel resource sharing data structure that realizes the sharing and interaction of resource information among multiple agents. When multiple agents apply for the same resource information from the blackboard data structure, the blackboard data structure will first publish the resource information to the agent that submitted the resource application earliest. After obtaining the resource information, the agent uses the COP optimizer to perform constraint processing locally to generate the target resource allocation plan, and then writes it to the blackboard data structure, that is, changes the data content on the blackboard, and then publishes the resource information to other agents again according to the first-in-first-out strategy until the resource applications of all agents are processed and the target resource allocation plan of each agent is obtained.

[0092] It should be noted that the multi-agent task allocation problem model is a distributed constraint optimization satisfaction problem, and the solution method is the branch-and-bound search algorithm. In addition, this problem model can also use other algorithms such as depth-first search, best-first search, backtracking, etc. Specifically:

[0093] The depth-first search algorithm starts from the root node of the pseudo-tree, traverses its first child node each time, and when traversing to a leaf node, goes back one step to its parent node, and then traverses other child nodes of the parent node, and repeats this process until all nodes are traversed.

[0094] The best-first search algorithm is a heuristic search algorithm (Heuristic Algorithm), which is based on the breadth-first search algorithm. The difference is that it relies on an evaluation function to evaluate the nodes to be traversed, and selects nodes with small costs to traverse until the target point is found.

[0095] The backtracking algorithm is actually a search and attempt process similar to enumeration. It mainly searches for the solution of the problem during the search and attempt process. When it is found that the solution conditions are no longer met, it "backtracks" and returns to try other paths. The backtracking method is a selection search method that searches forward according to the selection conditions to achieve the goal. However, when exploring a certain step, it is found that the original choice is not optimal or cannot achieve the goal, so it goes back one step and chooses again.

[0096] In an optional manner, the method further includes:

[0097] The spatial task planning plan of each agent is output and displayed.

[0098] Specifically, the spatial task planning plan of each agent is output and displayed through the blackboard.

[0099] In an optional manner, the method further includes:

[0100] When any space experiment application information does not meet the corresponding requirements, based on the updated space task planning requirement information obtained by each intelligent agent, return to the step of receiving the first resource allocation plan generated by each intelligent agent according to the corresponding space task planning requirement information until the space task planning plan generated by each intelligent agent according to the corresponding space experiment application information is obtained.

[0101] Specifically, when any space experiment application information does not meet the corresponding requirements, based on the constraints or space tasks added by the user (C = C∪c z Or adjust task F i =T iz ∪F i , c z is the new constraint, T iz is the new space task corresponding to the ith agent), returns to the step of receiving the first resource allocation plan generated by each agent according to the corresponding space task planning requirement information, until the space task planning plan generated by each agent according to the corresponding space experiment application information is obtained.

[0102] The technical solution of this embodiment proposes a hybrid active planning method for space tasks based on multiple agents, which regards the ground end of the payload that integrates communication, computing, and data processing capabilities as an agent, and obtains the optimal solution to the problem through information exchange and autonomous decision-making between agents. At the same time, the planning process is more flexible, and scientific users can participate in the entire decision-making process, supporting users to gradually add constraints and adjust tasks to meet users' diverse experimental needs. The technical solution of this embodiment solves the design defects of centralized planning, and uses a distributed computing system composed of multiple agents to quickly adapt to changes in environmental conditions, make independent and parallel decisions, and effectively improve the efficiency of hybrid active planning of space tasks.

[0103] Fig. 9 FIG. 2 shows a schematic diagram of a structure of an embodiment of a multi-agent-based spatial task hybrid active planning system 200 provided by the present invention. Fig. 9 As shown, the system 200 includes: a receiving module 210, an optimization module 220 and a planning module 230;

[0104] The receiving module 210 is used to: receive a first resource allocation scheme generated by each agent according to corresponding space mission planning requirement information;

[0105] The optimization module 220 is used to optimize and adjust all the first resource allocation schemes to obtain the target resource allocation scheme of each agent, and generate space experiment application information corresponding to each target resource allocation scheme;

[0106] The planning module 230 is used to obtain the space mission planning plan generated by each intelligent agent according to the corresponding space experiment application information when the space experiment application information meets the corresponding requirements.

[0107] In an optional manner, any space task planning requirement information includes: a set of space tasks to be planned, prerequisites for executing space tasks, a space task decomposition method, resources required for space tasks, and space task planning constraints.

[0108] In an optional manner, any agent includes: an HTN planner, a COP optimizer, and an experiment generator; and the process of any agent generating a first resource allocation solution according to corresponding space mission planning requirement information includes:

[0109] The HTN planner of any of the agents generates a spatial task execution plan based on the spatial task set to be planned, the spatial task execution prerequisites and the spatial task decomposition method input by the user, and sends the plan to the COP optimizer of any of the agents;

[0110] The COP optimizer of any of the intelligent agents generates a first resource allocation plan for any of the intelligent agents based on the resources required for the space mission and the space mission planning constraints.

[0111] In an optional manner, the HTN planner of any agent generates a spatial task execution plan according to the set of spatial tasks to be planned, the spatial task execution prerequisites and the spatial task decomposition method input by the user, including:

[0112] Initialize and set an empty planning task list, and determine whether any spatial task in the to-be-planned spatial task set is an atomic task;

[0113] If any of the space tasks is an atomic task, determine the instruction of the space task and add it to the planned task list, update the current network state and delete the space task from the planned task list according to the state list information deleted or added after the space task is executed;

[0114] If any of the space tasks is not an atomic task, the state information of the space task is matched with the space task execution prerequisite; if the match is successful, the space task is divided into multiple sub-space tasks using the space task decomposition method and added to the planned task list; if the match is unsuccessful, the space task is added to the planned task list;

[0115] Repeat the process of determining whether any spatial task in the set of spatial tasks to be planned is an atomic task, until each spatial task in the set of spatial tasks to be planned is determined, a final list of planned tasks is obtained, and the final list of planned tasks is determined as the spatial task execution plan.

[0116] In an optional manner, the process of the COP optimizer of any of the agents generating the first resource allocation scheme of any of the agents according to the resources required for the space mission and the space mission planning constraints includes:

[0117] The COP optimizer of any of the intelligent agents determines a distributed constraint optimization algorithm according to the spatial task planning constraints, and generates a first resource allocation plan for any of the intelligent agents according to the resources required for the spatial task and the distributed constraint optimization algorithm.

[0118] In an optional manner, the optimization module 220 is specifically used to:

[0119] Based on the first-in-first-out strategy, all the first resource allocation plans are optimized and adjusted to obtain the target resource allocation plan for each agent.

[0120] In an optional manner, the device further includes: a display module; the display module is used to:

[0121] The spatial task planning plan of each agent is output and displayed.

[0122] In an optional manner, the method further includes: an iterative planning module; the iterative planning module is used to:

[0123] When any space experiment application information does not meet the corresponding requirements, based on the updated space task planning requirement information obtained by each intelligent agent, return to the step of receiving the first resource allocation plan generated by each intelligent agent according to the corresponding space task planning requirement information until the space task planning plan generated by each intelligent agent according to the corresponding space experiment application information is obtained.

[0124] The technical solution of this embodiment solves the design defects of centralized planning. The distributed computing system composed of multiple agents can quickly adapt to changes in environmental conditions and make independent and parallel decisions, which can effectively improve the efficiency of hybrid active planning of space tasks.

[0125] The above-mentioned parameters and steps for each module to implement corresponding functions in the multi-agent-based spatial task hybrid active planning system 200 of this embodiment can refer to the parameters and steps in the embodiment of the multi-agent-based spatial task hybrid active planning method above, and will not be repeated here.

[0126] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. Similarly, in order to simplify the present invention and help understand one or more of the various inventive aspects, in the above description of exemplary embodiments of the present invention, the various features of the embodiments of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof. Wherein, the claims that follow the specific embodiment are hereby expressly incorporated into the specific embodiment, wherein each claim itself is a separate embodiment of the present invention.

[0127] It should be noted that the above embodiments illustrate the present invention rather than limit it, and that those skilled in the art may devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference symbol between brackets shall not be construed as a limitation on the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "one" or "an" preceding an element does not exclude the presence of a plurality of such elements. The present invention may be implemented by means of hardware comprising a number of different elements and by means of a suitably programmed computer. In a unit claim enumerating a number of devices, several of these devices may be embodied by the same hardware item. The use of the words first, second, and third, etc. does not indicate any order. These words may be interpreted as names. The steps in the above embodiments, unless otherwise specified, should not be understood as limitations on the order of execution.

Claims

1. A hybrid active planning method for spatial tasks based on multi-agents, characterized in that: include: Receiving a first resource allocation plan generated by each agent according to corresponding space mission planning requirement information; Optimize and adjust all the first resource allocation schemes to obtain the target resource allocation scheme of each agent, and generate space experiment application information corresponding to each target resource allocation scheme; When all space experiment application information meets the corresponding requirements, obtain the space mission planning plan generated by each agent according to the corresponding space experiment application information; The planning requirement information of any space task includes: the set of space tasks to be planned, the prerequisites for executing the space tasks, the decomposition method of the space tasks, the resources required for the space tasks and the planning constraints of the space tasks; Any intelligent agent includes: an HTN planner, a COP optimizer and an experiment generator; the steps of generating a first resource allocation scheme by any intelligent agent according to corresponding space mission planning requirement information include: The HTN planner of any of the agents generates a spatial task execution plan based on the spatial task set to be planned, the spatial task execution prerequisites and the spatial task decomposition method input by the user, and sends the plan to the COP optimizer of any of the agents; The COP optimizer of any of the intelligent agents generates a first resource allocation plan for any of the intelligent agents according to the resources required for the space mission and the space mission planning constraints; The step of generating a spatial task execution plan by the HTN planner of any agent according to the spatial task set to be planned, the spatial task execution prerequisite and the spatial task decomposition method input by the user comprises: Initialize and set an empty planning task list, and determine whether any spatial task in the to-be-planned spatial task set is an atomic task; If any of the space tasks is an atomic task, determine the instruction of the space task and add it to the planned task list, update the current network state and delete the space task from the planned task list according to the state list information deleted or added after the space task is executed; If any of the space tasks is not an atomic task, the state information of the space task is matched with the space task execution prerequisite; if the match is successful, the space task is divided into multiple sub-space tasks using the space task decomposition method and added to the planned task list; if the match is unsuccessful, the space task is added to the planned task list; Repeat the step of determining whether any spatial task in the set of spatial tasks to be planned is an atomic task, until each spatial task in the set of spatial tasks to be planned is determined, a final list of planned tasks is obtained, and the final list of planned tasks is determined as the spatial task execution plan; The step of generating a first resource allocation scheme of any of the intelligent agents by the COP optimizer of any of the intelligent agents according to the resources required for the space mission and the space mission planning constraints comprises: The COP optimizer of any of the intelligent agents determines the distributed constraint optimization algorithm according to the space task planning constraints, and generates a first resource allocation plan for any of the intelligent agents according to the resources required for the space task and the distributed constraint optimization algorithm; wherein the COP optimizer refers to a module for solving constraint optimization problems, which obtains the system resource status, calls the optimization algorithm according to the planning constraints entered in advance by the scientific user, and processes the constraints locally in the intelligent agent to generate a resource allocation plan.

2. The multi-agent-based spatial task hybrid active planning method according to claim 1 is characterized in that: The step of optimizing and adjusting all the first resource allocation schemes to obtain the target resource allocation scheme of each agent includes: Based on the first-in-first-out strategy, all the first resource allocation plans are optimized and adjusted to obtain the target resource allocation plan for each agent.

3. The multi-agent-based spatial task hybrid active planning method according to claim 2 is characterized in that: Also includes: The spatial task planning plan of each agent is output and displayed.

4. The multi-agent based spatial task hybrid active planning method according to claim 1, characterized in that: Also includes: When any space experiment application information does not meet the corresponding requirements, based on the updated space task planning requirement information obtained by each intelligent agent, return to the step of receiving the first resource allocation plan generated by each intelligent agent according to the corresponding space task planning requirement information until the space task planning plan generated by each intelligent agent according to the corresponding space experiment application information is obtained.

5. A multi-agent based spatial task hybrid active planning system, characterized in that: include: Receiving module, optimization module and planning module; The receiving module is used to: receive a first resource allocation plan generated by each agent according to corresponding space mission planning requirement information; The optimization module is used to optimize and adjust all the first resource allocation schemes to obtain the target resource allocation scheme of each agent, and generate space experiment application information corresponding to each target resource allocation scheme; The planning module is used to: when each space experiment application information meets the corresponding requirements, obtain the space task planning plan generated by each agent according to the corresponding space experiment application information; The planning requirement information of any space task includes: the set of space tasks to be planned, the prerequisites for executing the space tasks, the decomposition method of the space tasks, the resources required for the space tasks and the planning constraints of the space tasks; In an optional manner, any agent includes: an HTN planner, a COP optimizer, and an experiment generator; and the process of any agent generating a first resource allocation solution according to corresponding space mission planning requirement information includes: The HTN planner of any of the agents generates a spatial task execution plan based on the spatial task set to be planned, the spatial task execution prerequisites and the spatial task decomposition method input by the user, and sends the plan to the COP optimizer of any of the agents; The COP optimizer of any of the intelligent agents generates a first resource allocation plan for any of the intelligent agents according to the resources required for the space mission and the space mission planning constraints; In an optional manner, the HTN planner of any agent generates a spatial task execution plan according to the set of spatial tasks to be planned, the spatial task execution prerequisites and the spatial task decomposition method input by the user, including: Initialize and set an empty planning task list, and determine whether any spatial task in the to-be-planned spatial task set is an atomic task; If any of the space tasks is an atomic task, determine the instruction of the space task and add it to the planned task list, update the current network state and delete the space task from the planned task list according to the state list information deleted or added after the space task is executed; If any of the space tasks is not an atomic task, the state information of the space task is matched with the space task execution prerequisite; if the match is successful, the space task is divided into multiple sub-space tasks using the space task decomposition method and added to the planned task list; if the match is unsuccessful, the space task is added to the planned task list; Repeat the step of determining whether any spatial task in the set of spatial tasks to be planned is an atomic task, until each spatial task in the set of spatial tasks to be planned is determined, a final list of planned tasks is obtained, and the final list of planned tasks is determined as the spatial task execution plan; The process of the COP optimizer of any of the intelligent agents generating a first resource allocation scheme of any of the intelligent agents according to the resources required for the space mission and the space mission planning constraints includes: The COP optimizer of any of the intelligent agents determines the distributed constraint optimization algorithm according to the space task planning constraints, and generates a first resource allocation plan for any of the intelligent agents according to the resources required for the space task and the distributed constraint optimization algorithm; wherein the COP optimizer refers to a module for solving constraint optimization problems, which obtains the system resource status, calls the optimization algorithm according to the planning constraints entered in advance by the scientific user, and processes the constraints locally in the intelligent agent to generate a resource allocation plan.

Images