Maritime task chain service path optimization method and device combining knowledge graph and differential evolution
By combining knowledge graphs and differential evolution algorithms, a unified representation and global optimization of task chains are achieved, solving the problems of inconsistent service dependency descriptions and huge optimization space in existing technologies, and improving the efficiency of task chain generation and resource utilization.
Patent Information
- Application Number
- CN202511837371.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-08
- Publication Date
- 2026-03-17
AI Technical Summary
Existing methods for task chain generation and optimization lack a unified formal representation of services, making it difficult to describe the dependencies and capability constraints between services. This results in a large optimization space, low search efficiency, and difficulty in meeting multiple objective requirements, leading to insufficient practicality and robustness of the optimization results.
By combining knowledge graphs and differential evolution algorithms, a service and task chain representation model based on knowledge graphs is constructed. Logically feasible task chains are generated using rule-based reasoning, and global optimization is performed using differential evolution algorithms to achieve automated reconstruction of service paths.
It improves the efficiency of task chain formation, increases the success rate of execution and resource utilization, reduces task execution costs, and achieves intelligent optimization of multiple objectives.
Smart Images
Figure CN121684233A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the field of system architecture design and optimization, in particular to a sea task chain service path optimization method and device combining knowledge graph and differential evolution. BACKGROUND
[0002] In modern joint operation systems, task execution often needs to be completed by multiple heterogeneous services in collaboration, and is organized and scheduled through task chains. For example, in intelligence surveillance and reconnaissance, command and control, and combat support tasks, services such as detection, identification, tracking, and attack need to be organized into complete task chains according to dependency relationships and constraint conditions.
[0003] Existing task chain generation and optimization methods mainly include path planning methods based on graph search and combination optimization methods based on intelligent algorithms such as genetic algorithms. These methods can to some extent achieve the generation and optimization of task chains, but still have the following shortcomings:
[0004] 1) Lack of unified formalized representation of services: different methods have different ways of abstracting services, making it difficult to accurately describe the dependency relationships, preconditions, and capability constraints between services, and the generation of task chains lacks consistency and scalability, making it difficult to support flexible combination of cross-domain, multi-source, and heterogeneous services.
[0005] 2) Lack of logical pruning mechanism: in the process of task chain generation and optimization, search is often directly performed in a large-scale candidate solution space without making full use of task dependency and constraint information to eliminate unsuitable solutions in advance, resulting in a large optimization space and low search efficiency.
[0006] 3) Globality and multi-objective adaptability of optimization methods are insufficient: in large-scale, multi-constrained, and multi-objective task chain optimization problems, it is easy to fall into local optimization and difficult to simultaneously consider multiple objectives such as task completion time, success rate, and total resource consumption cost, resulting in insufficient practicality and robustness of the optimization results. SUMMARY
[0007] To solve the above problems, the present disclosure provides a task chain service path optimization method and device combining knowledge graph and differential evolution, which can uniformly model services and task chains at the semantic level, realize pruning and reduction of the optimization space through logical reasoning, and then combine an efficient global optimization algorithm to quickly obtain an optimal task chain service path that meets multiple performance requirements.
[0008] The method mainly includes the following steps:
[0009] (1) Construct a service and task chain representation model based on a knowledge graph:
[0010] Task-oriented services (such as detection, identification, support, etc.) are uniformly abstracted as nodes in the knowledge graph, and the calling dependencies, data flow, resource constraints and execution conditions between services are abstracted as edges. The unified modeling of service capabilities and task chain structure is realized through ontology and semantic rules, as shown in Figure 1 .
[0011] The specific steps include:
[0012] First, services are divided into perception, analysis and execution classes according to their functions.
[0013] Protégé is used to build ontology, define service classes and their subclasses, and each class contains attributes such as capability description, input and output data types, preconditions, resource requirements and execution constraints (for example, a detection service can contain attributes such as detection distance, target type, detection accuracy, etc.).
[0014] Then, specific service instances are mapped to the corresponding classes and parameters are set to realize service instantiation and computability.
[0015] The relationship between services includes dependency, data transmission, resource conflict, etc. The knowledge graph forms a computable graph structure, providing a semantic basis for subsequent reasoning and optimization.
[0016] Through this method, different service types can be modeled under a unified framework, providing a structured and standardized data foundation for task chain generation.
[0017] (2) Feasible task chain generation based on rule-based reasoning:
[0018] The constructed knowledge graph and rule-based reasoning technology are used to generate executable task chains that meet logical constraints, while eliminating unreasonable candidate paths in advance to reduce the optimization space.
[0019] The specific steps include:
[0020] First, define preconditions rules, capability matching rules and resource constraint rules, and design logical pruning rules to eliminate task chains that violate dependency relationships or resource constraints. Example rules are as follows:
[0021] ① Precondition rule: A certain identification service can only be executed after the detection service is completed.
[0022] ② Capability matching rule: The maximum range of detection ≥ target distance.
[0023] ③ Resource constraint rule: Only one task chain can be executed on the same task platform at a time to avoid resource conflicts.
[0024] ④ Logical pruning rule: If there is a mismatch in target type, insufficient resources or service parameters that cannot meet task requirements in the task chain, the path is eliminated.
[0025] Subsequently, the knowledge graph is traversed and reasoned by the Jena inference engine, and a set of feasible task chain sets satisfying the constraint conditions is automatically generated.
[0026] During the task chain generation process, the system starts from the starting service node, gradually expands the task chain according to the dependency relationship, and checks the preconditions, capability matching and resource conflict of each path. The paths that do not meet the conditions are automatically pruned.
[0027] This method ensures that the generated task chain is logically feasible, providing a high-quality candidate solution set for subsequent optimization.
[0028] (3) Service path optimization based on differential evolution:
[0029] The feasible task chain is optimized for multiple objectives, including the shortest task completion time, the highest success rate, and the lowest total cost.
[0030] The specific steps include:
[0031] The task chain path, service selection or resource allocation decision variables are encoded into vector form as individuals for differential evolution algorithm operation. By initialization, multiple candidate solutions are generated, and then mutation, crossover and selection operations are performed, and iterative optimization is performed. The overall process is shown in Figure 2.
[0032] In the multi-objective evaluation function, the total execution time of the task chain, the task success rate and the total cost are calculated, and each candidate solution is evaluated by weighting. For example, when there are multiple feasible task chains, the algorithm can score each chain according to its execution time, success rate and resource consumption cost, and select the optimal or near-optimal solution.
[0033] After iterative convergence, the algorithm can output the optimal task chain and resource allocation scheme, achieving global search optimization and multi-objective consideration, improving task execution efficiency and system reliability.
[0034] Through the organic combination of the above three technologies, the present disclosure realizes the unified representation of task-based services, the generation of logically feasible task chains, and the intelligent optimization of multiple objectives.
[0035] Compared with the prior art, the beneficial effects of the present disclosure are: ① By constructing a knowledge graph-based task chain representation, generating feasible task chains through rule reasoning, and optimizing service paths through differential evolution, the task chain is automatically reconstructed; ② It can effectively improve the task chain construction efficiency and execution success rate, optimize resource utilization, and reduce task execution cost; ③ This achievement has been applied in the service-oriented architecture design and verification of a certain system, and the verification effect is good, showing good engineering applicability and promotion prospects. BRIEF DESCRIPTION OF DRAWINGS
[0036] The above and other objects, features and advantages of the present disclosure will become more apparent from the following detailed description when taken in conjunction with the accompanying drawings, in which like reference characters refer to like parts throughout the figures, and wherein:
[0037] Figure 1 A service and task chain representation method flowchart based on a knowledge graph according to the present disclosure;
[0038] Figure 2 A service path optimization algorithm flowchart based on differential evolution;
[0039] Figure 3 A service classification and instance knowledge graph diagram of a maritime search and rescue system in an exemplary embodiment;
[0040] Figure 4 A service attribute example in an exemplary embodiment;
[0041] Figure 5 A differential evolution convergence process example (small case);
[0042] Figure 6 A differential evolution convergence process example (extended case);
[0043] Figure 7 A general flowchart of a maritime task chain service path optimization method according to the present disclosure. DETAILED DESCRIPTION
[0044] Preferred embodiments of the present disclosure will be described in greater detail below with reference to the accompanying drawings. Although preferred embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that the present disclosure is more thorough and complete, and to fully convey the scope of the present disclosure to those skilled in the art.
[0045] The present disclosure provides a task chain service path optimization method and device combining a knowledge graph and differential evolution. The method realizes semantic modeling of task-based services and their dependency relationships through a knowledge graph, generates a logically feasible task chain in combination with rule-based reasoning, and uses a differential evolution algorithm to globally optimize and reconstruct the service path.
[0046] The general flowchart of the method is shown in FIG. 1, and mainly includes the following steps: Figure 7
[0047] S1, task-based service collection and category division;
[0048] S2, based on a knowledge graph, semantic unified modeling of task-based services and their dependency relationships;
[0049] S3, generating a set of logically feasible task chains based on rule reasoning;
[0050] S4, globally optimizing and reconstructing the service path based on the differential evolution algorithm to form an optimal task chain service path.
[0051] In an exemplary embodiment, an offshore unmanned search and rescue system is verified. The system is composed of unmanned surface ships, unmanned aerial vehicles and unmanned underwater vehicles, covering target search, target identification, environmental reconnaissance, rescue execution, task coordination and command control, communication support and other functions, and realizing multi-platform collaborative operation.
[0052] For this system, the main steps of the task chain service path optimization method combining knowledge graph and differential evolution are as follows:
[0053] 1. The OWL ontology is used to semantically model the services and platforms
[0054] In this embodiment, task-based services (such as detection, identification, support, etc.) are abstracted as nodes in the knowledge graph, and the calling dependencies, data flow, resource constraints and execution conditions between services are abstracted as edges. The unified modeling of service capabilities and task chain structure is realized through ontology and semantic rules. The process is shown in the accompanying Figure 1
[0055] (1) First, collect task-based services and classify them according to functions.
[0056] In this embodiment, the overall architecture includes four types of services:
[0057] 1) Perception-based services: responsible for acquiring environmental and target information. For example, unmanned surface ship A and unmanned surface ship D provide sea target detection and floating object detection services, using radar or sonar sensors to detect targets and obstacles; unmanned underwater vehicle C and unmanned underwater vehicle E provide underwater sonar detection services to obtain underwater target position information; unmanned aerial vehicle B provides weather monitoring services to collect wind, wave height and visibility data. Each perception service sets execution time, coverage range, success rate indicators, and specifies deployment on a specific platform.
[0058] 2) Analysis-based services: process and analyze perception data. For example, the target identification service of unmanned surface ship A uses image recognition or radar analysis to determine the target type; the environmental analysis service of unmanned aerial vehicle B analyzes the weather data to determine the sea state; the risk assessment service of unmanned surface ship A combines the target type and sea state analysis results to determine the risk level. Analysis-based services usually rely on pre-posed perception-based services to complete the acquisition of input data.
[0059] 3) Execution service: directly complete the rescue operation. For example, the near-shore rescue service of unmanned surface ship A deploys lifesaving equipment, and the ship guiding service guides the distressed ship, both of which rely on the risk assessment results and have resource conflict constraints during execution, ensuring that the same platform does not perform conflicting tasks at the same time.
[0060] 4) Command coordination service: used for task scheduling and coordination between platforms. For example, the command center provides task scheduling services, integrates task state information of various platforms, and outputs a task allocation scheme.
[0061] (2) Use the Protégé tool to build an ontology, define service classes and their subclasses, and each class contains attributes such as capability description, input and output data types, preconditions, resource requirements, and execution constraints. For example, the detection service can include detection distance, target type, detection accuracy, and other attributes;
[0062] Then map the specific service instance to the corresponding class and set the parameters to realize service instantiation and computability; the relationship between services includes dependency, data transfer, resource conflict, etc., and the knowledge graph forms a computable graph structure, providing a semantic foundation for subsequent reasoning and optimization.
[0063] Through the above OWL ontology modeling, the deployment relationship, dependency relationship, and resource conflict relationship between platforms and services are clearly expressed, as shown in Figure 3 and Figure 4 , providing basic data support for subsequent knowledge graph-based task chain optimization and decision-making.
[0064] 2. Based on rule reasoning, generate a set of logically feasible task chain service paths
[0065] Use the constructed knowledge graph and rule reasoning technology to generate executable task chains that meet logical constraints, and eliminate unreasonable candidate paths in advance to reduce the optimization space. Specifically, it includes:
[0066] First, define pre-condition rules, capability matching rules, and resource constraint rules, and design logical pruning rules to eliminate task chains that violate dependency relationships or resource constraints; then, through the Jena inference engine, traverse and reason the knowledge graph to automatically generate a set of feasible task chains that meet the constraint conditions. During task chain generation, the system starts from the starting service node, expands the task chain according to the dependency relationship, and checks each path for preconditions, capability matching, and resource conflict. Paths that do not meet the conditions are automatically pruned.
[0067] In this embodiment, on the basis of step 1, inference rules are constructed as follows:
[0068] 1) Precondition rule
[0069] For example, the rescue service can only be executed after the identification service is completed, written in Jena rule form:
[0070] [Precondition rule: (?x rdf:type :identification service) (?y rdf:type :rescue service)
[0071] (:dependent on?y?x) (:completed?x true)->(:executable?y true)]
[0072] 2) Capability matching rule
[0073] For example, coverage matching: the maximum coverage of the rescue platform ≥ target distance, written in Jena rule form:
[0074] [Coverage matching: (?s rdf:type :rescue service) (:coverage?s?r)
[0075] (?t rdf:type :distress target) (:target distance?t?d)
[0076] greaterThanOrEqual(?r,?d)-> (:coverage feasible?s true)]
[0077] Rescue capability matching: the number of lifesaving equipment ≥ the number of people in distress, written in Jena rule form:
[0078] [Rescue capability matching: (?s rdf:type :rescue service) (:number of rescue resources?s?n)
[0079] (?t rdf:type :distress target) (:number of people in distress?t?m)
[0080] greaterThanOrEqual(?n,?m) -> (:resource feasible?s true)]
[0081] 3) Resource constraint rule
[0082] For example, the same unmanned surface ship / unmanned aerial vehicle can only perform one service at the same time, written in Jena rule form:
[0083] [Resource constraint: (?s1 rdf:type :service) (?s2 rdf:type :service)
[0084] (:deployed in?s1?p) (:deployed in?s2?p) notEqual(?s1,?s2)
[0085] (:Not simultaneously executed?s1?s2) -> (:Resource conflict?s1?s2)]
[0086] 4) Logical pruning rules
[0087] For example, if the target type does not match the service applicable type (e.g.: life raft launching service can only be effective for "person in distress", not for "ship fire"), the task chain is not feasible, written in the form of Jena rules:
[0088] [Logical pruning: (?chain rdf:type :task chain) (:contains service?chain?s)
[0089] (:target type?t1) (:service applicable target type?s?t2)
[0090] notEqual(?t1,?t2)-> (:infeasible chain?chain)]
[0091] Thus, the set of feasible task chain service paths is formed, as shown in the following examples:
[0092] ① Sea surface target detection unmanned surface ship A → target identification unmanned surface ship A → weather monitoring unmanned aerial vehicle B → environmental analysis unmanned aerial vehicle B → risk assessment unmanned surface ship A → near shore rescue unmanned surface ship A;
[0093] ② Sea surface target detection unmanned surface ship D → floating object detection unmanned surface ship D → sonar underwater detection unmanned underwater vehicle E → target identification unmanned surface ship A → environmental analysis unmanned aerial vehicle B → risk assessment unmanned surface ship A → near shore rescue unmanned surface ship A.
[0094] 3. Based on the differential evolution algorithm, the service path is globally optimized and reconstructed to form the optimal task chain service path. The overall process is shown in FIG. 2.
[0095] In this embodiment, the differential evolution algorithm is adopted, and the optimal task chain service path formed by considering the execution time, success rate and total cost and other objective functions is:
[0096] Optimal task chain: ['sea surface target detection unmanned surface ship A', 'target identification unmanned surface ship A', 'weather monitoring unmanned aerial vehicle B', 'environmental analysis unmanned aerial vehicle B', 'risk assessment unmanned surface ship A', 'near shore rescue unmanned surface ship A'].
[0097] Since the total feasible chain of the service path in this embodiment is less, the convergence curve is as follows: Figure 5As shown, directly into the optimal state. If the task chain length is extended to 10, and each task has 3-4 optional service instances, a convergence curve as shown in Figure 6
[0098] It can be seen that, by means of knowledge graph-based service semantic modeling, reasoning rule constraint and differential evolution optimization, the present disclosure can generate a feasible task chain service path under the conditions of multiple platforms, multiple tasks and multiple targets, and output an optimal scheme; at the same time, the method has good scalability, and is not only suitable for a maritime unmanned search and rescue system composed of unmanned surface ships, unmanned underwater vehicles and unmanned aerial vehicles under a service-oriented architecture, but also can be applied to rapid construction and reorganization of task chain links of other complex systems under a service-oriented architecture.
[0099] The technical solutions described above are only exemplary embodiments of the present application. For those skilled in the art, on the basis of the application disclosed herein, various types of improvements or modifications can be easily made without being limited to the methods described in the above embodiments. Therefore, the above description is only preferred, and is not limiting.
Claims
1. A method for optimizing maritime task chain service path by combining knowledge graph and differential evolution, characterized in that, The method comprises the following steps: S1, task type service collection and category division; S2, based on the knowledge graph, the task type service and its dependent relationship are uniformly semantically modeled; S3, based on rule reasoning, a set of logically feasible task chain service paths is generated; S4, based on the differential evolution algorithm, the service path is globally optimized and reconstructed to form the optimal task chain service path.
2. The method of claim 1, wherein, In the step S1, the category division of the task type service comprises: 1) perception type service: used for obtaining environment and target information, each perception service is provided with execution time, coverage range, success rate index, and is explicitly deployed on a specific platform; 2) analysis type service: used for processing and analyzing the obtained perception data; 3) execution type service: used for executing specific actions; the same platform will not execute conflicting tasks at the same time; 4) command coordination service: used for task scheduling and coordination between platforms.
3. The method of claim 1, wherein, The step S2 specifically comprises: The task type service is uniformly abstracted as a node in the knowledge graph, the service ontology is established by using the Protégé tool, the service class and its subclasses are defined, and the attributes of each class include one or more of the following: capability description, input and output data type, preconditions, resource demand and execution constraint; The specific service instance is mapped to the corresponding class, and parameters are set for it to realize service instantiation and computability; The calling dependency, data flow, resource constraint and execution condition between services are abstracted as edges, and the service representation based on the knowledge graph is generated through the ontology and semantic rules.
4. The method of claim 1, wherein, The step S3 specifically comprises: The pre-condition rule, the capability matching rule and the resource constraint rule are defined, and the logical pruning rule is designed to eliminate the task chain that violates the dependent relationship or resource constraint; Through the Jena inference machine, the knowledge graph is traversed and inferred to automatically generate a set of feasible task chains that meet the constraint conditions; In the task chain generation process, starting from the starting service node, the task chain is gradually expanded according to the dependent relationship, and the pre-condition, capability matching and resource conflict of each path are checked, and the path that does not meet the condition is automatically pruned; Thus, a set of logically feasible task chain service paths is formed.
5. The method according to any one of claims 1 to 4, characterized in that, The step S4 specifically comprises: The decision variables such as task chain path, service selection or resource allocation are encoded into vector form as individuals of the differential evolution algorithm for operation; A plurality of candidate solutions are initialized, and then mutation, crossover and selection operations are performed to continuously iterate and optimize; In the multi-objective evaluation function, the total execution time of the task chain, the task success rate and the total cost are calculated respectively, and each candidate solution is evaluated by weighting.
6. A device for optimizing service path of maritime task chain by combining knowledge graph and differential evolution by using the method of any one of claims 1-5. It comprises: A task type service knowledge graph construction module is used for semantically and uniformly modeling the task type service and its dependent relationship based on the knowledge graph; A rule reasoning module is used for generating a set of logically feasible task chains based on rule reasoning; A differential evolution module is used for globally optimizing and reconstructing the service path based on the differential evolution algorithm to form the optimal task chain service path.