Method for reconnaissance resource matching and optimization based on ontology and monarch butterfly algorithm

Abstract

The application relates to a reconnaissance resource matching and optimization method based on ontology and monarch butterfly algorithms and belongs to the technical field of task resource matching. The method comprises the following steps: 1: based on ontology modeling technology, modeling tasks and resources in the reconnaissance field, and constructing an ontology knowledge base; 2: given task information described in natural language, using natural language processing technology to analyze the task information, extracting key information such as task time, task location, reconnaissance scene and investigation object, and completing the resource preference and execution environment of the task by the user; 3: based on the completed task information, searching and matching are carried out from the resource base based on ontology reasoning; 4: based on the analytic hierarchy process and monarch butterfly algorithms, the matched resource instances are optimized. The application has the beneficial effects that the reconnaissance task information can be automatically extracted and the user preference is integrated; according to the resource attributes, the ability and the user preference information, the resources are matched and optimized, and the matching degree of the task and the resources is improved.

Description

Technical Field

[0001] This invention belongs to the field of mission resource matching technology, and specifically relates to an ontology-based method for reconnaissance resource matching and optimization. Background Technology

[0002] The increasing informatization and intelligentization trends across various fields have led to a richer variety of reconnaissance resources. The demands for cross-domain resource aggregation and collaboration in future reconnaissance scenarios are placing higher requirements on resource integration and matching across different fields. When dealing with heterogeneous resources from various fields, ontology modeling techniques are needed to model and describe tasks and resources, and to build a knowledge base for storage. In the face of reconnaissance missions under modern warfare conditions, how to quickly match suitable reconnaissance resources to these missions is a crucial problem that needs to be solved in the field of reconnaissance resource matching. Therefore, it is necessary to study corresponding resource modeling and representation methods, as well as corresponding matching algorithms, to match the optimal resources required by the mission.

[0003] Current research on task resource matching has yielded some results. Li Chenghai et al. described cloud manufacturing service resources as a quintuple and proposed an algorithm based on resource attribute matching to discover suitable resources for manufacturing tasks; Song Wenbin et al. used RDF technology to describe resources. [2] Gu Fu et al. proposed a framework for describing scientific and technological resources, combining resource classification models, metadata models, ontology models, and other models to describe resources. Regarding ontology-based resource modeling, Lina Lan et al. designed a resource description model for IoT device resources based on ontology modeling technology. This model provides a consistent view of heterogeneous sensing devices for IoT applications in the cloud, addressing the problem of resource sharing among heterogeneous devices. E et al. proposed an OWL-based resource capability ontology modeling method for describing resource capabilities. This method supports representation and automatic reasoning from simple functions to composite functions, providing a method for rapid resource selection and configuration. Analysis of the current research shows that while significant progress has been made in ontology and resource description, the proposed methods still have the following problems:

[0004] 1. Existing modeling techniques focus on modeling resource attributes, but offer limited descriptions of heterogeneous reconnaissance resource capabilities across different domains, hindering resource evaluation, optimization, and integration. The complexity and uncertainty of independently describing cross-domain resource models also impede resource aggregation and collaboration.

[0005] 2. Existing task resource matching methods generally calculate the matching degree between tasks and resources based on their attributes and functions, and select appropriate resources according to the matching degree. This method has low matching efficiency and does not take into account the capability requirements of the task and the capability level of the resources, which may result in the resources being unable to complete the given reconnaissance task. Summary of the Invention

[0006] This invention addresses the problems of inconsistent resource modeling and description across domains, insufficient description of energy capabilities, and low efficiency in resource-task matching by proposing a two-stage task-resource matching and optimization algorithm. First, an ontology-based reconnaissance resource matching method is proposed. Second, for resource optimization in the reconnaissance domain, a resource optimization algorithm based on an improved fuzzy hierarchical analysis method and the Monarch Butterfly algorithm is proposed, ultimately achieving task-resource matching and optimization.

[0007] The above-mentioned technical problems of the present invention are mainly solved by the following technical solutions:

[0008] A method for reconnaissance resource matching and optimization based on ontology and the Monarch Butterfly algorithm is characterized by the following steps:

[0009] Step 1: Model the reconnaissance mission. The modeling mission includes the scenario, platform, sensors, mission type, mission capability requirements, weather concept and the relationships between them.

[0010] Step 2: First, use ontology modeling technology to model resources in the reconnaissance field, including unmanned reconnaissance aircraft and payloads, aerial reconnaissance aircraft, radar, and reconnaissance vehicles, describe the data attributes and capabilities of the modeled resources, and build a knowledge base for the resources.

[0011] Step 3: Given a task description in natural language, use LTP-based language processing technology to parse the task and extract information including reconnaissance area, reconnaissance time, reconnaissance target, and reconnaissance scene. Then, the user completes the weather conditions and resource preferences of the reconnaissance area.

[0012] Step 4: Based on the task information extracted in Step 3, search for resources with corresponding capabilities from the task ontology model based on knowledge reasoning;

[0013] Step 5: Optimize the matched resource instances based on the analytic hierarchy process and the monarch butterfly algorithm to select the optimal resource.

[0014] Furthermore, step 1 specifically includes the following process:

[0015] Step 1.1: Identify entity classes including the mission scenario and its involved reconnaissance objects, mission capability requirements, platforms, sensors, weather, etc., and complete the data attributes of each class;

[0016] Step 1.2: Analyze the relationships between the entities and establish connections between them.

[0017] Furthermore, step 2 specifically includes the following process:

[0018] Step 2.1: Identify entities and their instances according to the three levels of "abstraction-application-instance", analyze the attributes of the identified entities, and use ontology modeling technology to perform entity modeling;

[0019] Step 2.2: Analyze the capabilities of each resource and associate them with tasks. Based on this, complete the applicable scenarios and task types for each resource.

[0020] Furthermore, given a task description in natural language, step 3, task information parsing and completion, specifically includes the following processes;

[0021] Step 3.1: Parse the user-posted task using the LTP Natural Language Processing Toolkit, extracting entities, attributes, and relationships to form a triple; first, analyze words and phrases, including word segmentation, part-of-speech tagging, and named entity recognition; then, decompose the entire sentence structure through the dependency relationships between words, where dependency relationships express the semantic dependencies between the sentence components;

[0022] Step 3.2: Use the trained language model to extract task information, extract the relationships between specific entities in the information, and then obtain task time and task location information.

[0023] Furthermore, after extracting the task information, step 4 searches for resources with corresponding capabilities from the task ontology model based on knowledge reasoning, specifically including the following sub-steps;

[0024] Step 4.1: Complete the task information obtained in Step 3 by adding the user's resource preference information. Based on the completed task information, query the knowledge base for resources that can meet the task requirements and user preferences.

[0025] Step 4.2: Based on the reconnaissance scenario and reconnaissance object extracted from the task, use SPALQL language to find the parent class of the reconnaissance object from the task ontology model;

[0026] Step 4.3: Based on the parent class found in Step 4.2, further search for the capability requirements corresponding to the reconnaissance mission scenario;

[0027] Step 4.4: Based on the reconnaissance scenario and the reconnaissance target, use SPALQL language to find the reconnaissance platform that can perform the reconnaissance mission from the mission ontology model.

[0028] Step 4.5: Further traverse the subclasses of the reconnaissance platform to obtain the set of all its subclasses;

[0029] Step 4.6: Based on the capability requirements obtained in Step 4.3, find the corresponding sensors from the task ontology model;

[0030] Step 4.7: Based on the task information, weather conditions, and resource preferences completed by the user, filter the platform subclasses obtained in Step 4.5 to obtain a set of resources that meet the requirements.

[0031] Furthermore, step 5 uses an analytic hierarchy process (AHP) and the Monarch Butterfly algorithm to optimize resources, specifically including the following sub-steps;

[0032] Step 5.1 Based on the abstract resource categories obtained in Step 4.6, find specific resource instances under each category; first, according to the reconnaissance radius and flight altitude information of the reconnaissance mission, search the database for resource instances that match the mission requirements, filter out occupied resource instances, and finally return the found resource instances.

[0033] Step 5.2 Normalize the various attributes of the reconnaissance resources. Specifically, use the standard normalization method, first calculating the mean u of each attribute value. i and variance Then, each attribute value is normalized. The formulas for calculating the mean, variance, and normalization are shown in (1), (2), and (3):

[0034]

[0035]

[0036]

[0037] In formulas (1), (2), and (3), i represents the i-th attribute of a resource, j represents the j-th resource, and a ij Let represent the i-th value of the j-th resource instance, and n represent the total number of resources. This represents the normalized value;

[0038] Step 5.3 Calculate the weight of each attribute of the resource; this is done using the Analytic Hierarchy Process (AHP); the specific calculation process is as follows:

[0039] Construct a pairwise comparison matrix A, evaluate the relative importance of the paired attributes, and give the values ​​of the relative importance;

[0040] Calculate the hierarchical single-order weight vector and perform a consistency check;

[0041] Calculate the hierarchical total ranking weight vector and perform a consistency check;

[0042] The formula for calculating the consistency index (CI) is as follows:

[0043]

[0044] Where λ represents the largest eigenvalue of A; to measure the magnitude of CI, the inconsistency ratio CR is defined, and its calculation formula is as follows:

[0045]

[0046] In formula (5), RI is the random consistency index, and its value can be obtained by looking up a table; when the value of CR is less than 0.1, it is considered that the judgment matrix A has passed the consistency test; otherwise, the element values ​​of matrix A need to be adjusted until it passes the test.

[0047] Step 5.4 Generate an initial population. Each chromosome in the population is composed of the attributes of the resource, where each gene locus represents an attribute of the resource. In the process of generating the population, an orthogonal design method is used to generate the initial population.

[0048] Step 5.5 After the initial population is generated, the population is divided into two smaller subpopulations, NP1 and NP2. The population size in NP1 and NP2 is set using the following formulas:

[0049] NP1 = ceil(p·NP) (6)

[0050] NP2 = NP - NP1 (7)

[0051] Step 5.6 Adjust the position of the population; in formula (6), p represents the migration rate of the monarch butterfly, which is set to 0.417 based on existing techniques. The position update method for each individual in the population is determined by formulas (8) and (9):

[0052]

[0053]

[0054] in Let x represent the value at the (t+1)th iteration. i The k-th dimension is used to determine the position of Monarch butterfly i, and t represents the iteration number; r1 and r2 are generated randomly; when updating the individual position, the update mechanism is determined by the parameter R; if R≤p, then formula (8) is executed to update the position, otherwise formula (9) is used to update the position:

[0055] R = rand * peri (10)

[0056] Where peri represents the migration period, which is set to 1.2 based on existing technology, and rand is a random number in [0,1].

[0057] If rand ≤ p, the position update formula is:

[0058]

[0059] If rand > p, then the position update formula is:

[0060]

[0061] If rand > p, then further, if rand > BAR, the position of the monarch butterfly can be updated using the following formula:

[0062]

[0063] Where BAR represents the adjustment rate, its size is equal to p, and the random step size dx is calculated as follows:

[0064]

[0065] The parameter α is calculated using formula (15):

[0066]

[0067] S max This represents the maximum step size, and t represents the number of iterations;

[0068] Step 5.7 Construct an evaluation function to assess the overall capability of the resources. The constructed evaluation function is as follows:

[0069]

[0070]

[0071] maxf3=β1·f1+β2·f2 (18)

[0072] st0<ω i ,β1,β2<1 (19)

[0073] In formula (16), ω i Let represent the weight of the i-th resource attribute, and m represent the total number of attributes; Formula (17) represents the correlation between resource attributes and task requirement attributes, where The attributes that indicate the task requirements. The attributes of the resource are represented; in formula (18), β1 and β2 represent the weights of the two evaluation functions;

[0074] Step 5.8 Perform genetic operations on the population and evaluate individuals using an evaluation function. The algorithm stops calculating when it reaches the maximum number of iterations. Individuals are sorted according to their evaluation values; the larger the value, the better the individual.

[0075] Compared with the prior art, the present invention has the following beneficial effects:

[0076] This invention can automatically extract reconnaissance mission information and integrate user preferences; and based on resource attributes, capabilities and user preference information, it can match and optimize resources, which is conducive to improving the matching degree between missions and resources. Attached Figure Description

[0077] Figure 1 It is a task ontology model.

[0078] Figure 2 It is an ontology model of unmanned reconnaissance aircraft resources. Detailed Implementation

[0079] The following is a specific embodiment of applying the method of the present invention to a reconnaissance case: Here, the reconnaissance of ground tanks is taken as an example, and the specific implementation process of the present invention will be described in detail with reference to the accompanying drawings.

[0080] A reconnaissance resource matching and optimization method based on ontology and the Monarch Butterfly algorithm includes the following steps:

[0081] Step 1: Model the reconnaissance mission. This involves modeling common reconnaissance missions, primarily focusing on concepts such as mission scenarios, platforms, sensors, mission types, mission capability requirements, weather, and the relationships between them.

[0082] Step 2: Modeling resources in the reconnaissance domain. First, ontology modeling techniques are used to model commonly used resources in the reconnaissance domain, such as unmanned reconnaissance aircraft and their payloads, aerial reconnaissance aircraft, radar, and reconnaissance vehicles. This mainly describes the data attributes and capabilities of these resources. Taking unmanned aerial vehicles (UAVs) as an example, the main modeling focuses on their data attributes, including takeoff weight, service ceiling, maximum loiter time, maximum flight speed, reconnaissance radius, payload, and their capabilities, such as the ability of UAVs to reconnoiter both ground and aerial targets. This process builds a knowledge base of commonly used resources.

[0083] Step 3, Task Information Parsing and Completion. Given task information described in natural language, LTP-based language processing technology is used to parse the task resources, extracting information such as "reconnaissance area," "reconnaissance time," "reconnaissance target," and "reconnaissance scenario." Then, the user completes the weather conditions and resource preferences for the reconnaissance area.

[0084] Step 4: Based on the task information extracted in Step 3, search for resources with corresponding capabilities from the task ontology model based on knowledge reasoning;

[0085] Step 5: Optimize the matched resource instances based on the analytic hierarchy process and the monarch butterfly algorithm to select the optimal resource.

[0086] Step 1 specifically includes the following process:

[0087] Step 1.1 Identify the classes and relationships involved in common tasks in the domain. Identify common task scenarios and the entity classes involved, such as reconnaissance objects, task capability requirements, platforms, sensors, and weather, and complete the data attributes of each class.

[0088] Step 1.2 Entity Relationship Construction. Analyze the relationships between various entities and establish connections between them. For example, if an airborne platform can conduct reconnaissance of ground targets, then the relationship between "airborne platform" and "ground target" is "reconnaissance".

[0089] Step 2 specifically includes the following process:

[0090] Step 2.1 Identify domain resource entities and their attributes. Identify entities and their instances according to the three levels of "abstraction-application-instance", analyze the attributes of the identified entities, and use ontology modeling techniques to perform entity modeling.

[0091] Step 2.2 Resource Capability Modeling. Analyze the capabilities of each resource and associate them with tasks. Based on this, complete the applicable scenarios and task types for each resource, paying special attention to the impact of weather on resources.

[0092] Given a task description in natural language, step 3, task information parsing and completion, specifically includes the following process;

[0093] Step 3.1 Lexical Analysis of the Task Sentence. The user-submitted task is parsed using the LTP Natural Language Processing Toolkit, extracting entities, attributes, and relations to form a triple. First, words and phrases are analyzed, including word segmentation, part-of-speech tagging, and named entity recognition. After lexical analysis, the entire sentence structure is deconstructed through dependency relations between words. These dependency relations express the semantic dependencies between sentence components, such as subject-verb, verb-object, and modifier-type structural relationships.

[0094] Step 3.2 Extraction of Key Task Information. After sentence analysis, the trained language model is used to extract task information, including relationships between specific entities, such as subordinate relationships, kinship relationships, and synonym relationships, thereby obtaining task time and location information.

[0095] After extracting the task information, step 4 uses knowledge reasoning to search for resources with corresponding capabilities from the task ontology model, which includes the following sub-steps.

[0096] Step 4.1 Task Information Completion. The task information obtained in Step 3 is completed by adding the user's resource preference information. Based on the completed task information, resources that can meet the task requirements and user preferences are queried from the knowledge base.

[0097] Step 4.1 Based on the reconnaissance scenario and reconnaissance object extracted from the task, use SPALQL language to find the parent class of the reconnaissance object from the task ontology model;

[0098] Step 4.2 Based on the parent class found in Step 4.1, further search for the capability requirements corresponding to the reconnaissance mission scenario;

[0099] Step 4.3 Based on the reconnaissance scenario and the reconnaissance target, use SPALQL language to find the reconnaissance platform that can perform the reconnaissance mission from the mission ontology model;

[0100] Step 4.4 Further traverse the subclasses of the reconnaissance platform to obtain the set of all its subclasses;

[0101] Step 4.5 Based on the capability requirements obtained in Step 4.2, find the corresponding sensors from the task ontology model;

[0102] Step 4.6 Based on the task information, weather conditions, and resource preferences completed by the user, filter the platform subclasses obtained in Step 4.4 to obtain a set of resources that meet the requirements;

[0103] Step 5 uses an analytic hierarchy process (AHP) and the Monarch Butterfly algorithm to optimize resources, specifically including the following sub-steps;

[0104] Step 5.1 Obtain resource instances. Based on the abstract resource categories obtained in Step 4.6, find specific resource instances under each category. First, based on information such as the reconnaissance radius and flight altitude of the reconnaissance mission, search the database for resource instances that match the mission requirements, filter out occupied resource instances, and finally return the found resource instances;

[0105] Step 5.2 Normalize the various attributes of the reconnaissance resources. Specifically, use the standard normalization method, first calculating the mean u of each attribute value. i and variance Then, each attribute value is normalized. The formulas for calculating the mean, variance, and normalization are shown in (1), (2), and (3):

[0106]

[0107]

[0108]

[0109] In formulas (1), (2), and (3), i represents the i-th attribute of a resource, j represents the j-th resource, and a ij Let represent the i-th value of the j-th resource instance, and n represent the total number of resources. This represents the normalized value.

[0110] Step 5.3 Calculate the weight of each attribute of the resource. The Analytic Hierarchy Process (AHP) is used for this calculation. The specific calculation steps are as follows:

[0111] (1) Construct a pairwise comparison matrix A and evaluate the relative importance of the pairwise attributes, and give the value of the relative importance.

[0112] (2) Calculate the hierarchical single sorting weight vector and perform a consistency check.

[0113] (3) Calculate the total hierarchical ranking weight vector and perform a consistency check.

[0114] The formula for calculating the consistency index (CI) is as follows:

[0115]

[0116] Where λ represents the largest eigenvalue of A. To measure the magnitude of CI, the inconsistency ratio CR is defined, and its calculation formula is:

[0117]

[0118] In formula (5), RI is the random consistency index, the value of which can be obtained by looking up a table. When the value of CR is less than 0.1, it is considered that the judgment matrix A has passed the consistency test; otherwise, the element values ​​of matrix A need to be adjusted until it passes the test.

[0119] Step 5.4 generates an initial population. Each chromosome in the population is composed of resource attributes, where each gene locus represents a resource attribute. An orthogonal design method is used to generate the initial population.

[0120] Step 5.5 After the initial population is generated, the population is divided into two smaller subpopulations, NP1 and NP2. The population size in NP1 and NP2 is set using the following formulas:

[0121] NP1 = ceil(p·NP) (6)

[0122] NP2 = NP - NP1 (7)

[0123] Step 5.6 Adjust the position of the population. In formula (6), p represents the migration rate of the monarch butterfly, which is set to 0.417 based on existing techniques. The position update method for each individual in the population is determined by formulas (8) and (9):

[0124]

[0125]

[0126] in Let x represent the value at the (t+1)th iteration. i The k-th dimension is used to determine the position of Monarch butterfly i, and t represents the iteration number. r1 and r2 are generated randomly. When updating the individual position, the update mechanism is determined by the parameter R; if R≤p, then formula (8) is executed to update the position, otherwise formula (9) is used to update the position:

[0127] R = rand * peri (10)

[0128] Where peri represents the migration period, which is set to 1.2 based on existing technology, and rand is a random number in [0, 1].

[0129] If rand ≤ p, the position update formula is:

[0130]

[0131] If rand > p, then the position update formula is:

[0132]

[0133] If rand > p, then further, if rand > BAR, the position of the monarch butterfly can be updated using the following formula:

[0134]

[0135] Where BAR represents the adjustment rate, its size is equal to p, and the random step size dx is calculated as follows:

[0136]

[0137] The parameter α is calculated using formula (15):

[0138]

[0139] S max This represents the maximum step size, and t represents the number of iterations.

[0140] Step 5.7 Construct an evaluation function to assess the overall capability of the resources. The constructed evaluation function is as follows:

[0141]

[0142]

[0143] max f3=β1·f1+β2·f2 (18)

[0144] st0<ω i ,β1,β2<1 (19)

[0145] In formula (16), ω i Let represent the weight of the i-th resource attribute, and m represent the total number of attributes. Formula (17) represents the correlation between resource attributes and task requirement attributes, where The attributes that indicate the task requirements. The attributes of the resource are represented. In formula (18), β1 and β2 represent the weights of the two evaluation functions.

[0146] Step 5.8 Perform genetic operations on the population and evaluate individuals using an evaluation function. The algorithm stops calculating when it reaches the maximum number of iterations. Individuals are sorted according to their evaluation values; the larger the value, the better the individual.

[0147] The following reference Figure 1 and Figure 2 Further explanation of this embodiment:

[0148] Step 1 involves constructing the task ontology model. The search for tasks involves entity classes such as task scenario, objects, platform, sensors, task capability requirements, and weather. The model analyzes the integration relationship between the platform and sensors, the impact of weather on sensors, and object attributes such as task capability requirements. The constructed ontology model is shown below. Figure 1 As shown.

[0149] Step 2 involves constructing the UAV resource ontology model. First, define the UAV and its subclasses, complete their data attributes, and analyze their capabilities to construct the UAV resource ontology model. The constructed UAV ontology model is as follows: Figure 2 As shown.

[0150] Step 3 involves extracting key information from a given natural language description of a task. For example, the task description might be: "Due to continuous heavy rains, a landslide occurred near Qiangjiuling. A search operation targeting people on the ground near Qiangjiuling is being conducted on June 3rd to determine if any people are trapped in the area." First, the task is analyzed using the techniques described in Step 3. The key information extracted is as follows:

[0151] Task Time: June 3rd

[0152] Mission Location: Qiangjiuling, China

[0153] Target audience: Ground personnel

[0154] Mission Overview: Conduct a ground personnel search mission.

[0155] Mission scenario: Search for ground targets

[0156] Step 4 involves matching abstract resources to the extracted tasks. Based on the extracted task information and user-completed information, platform resources that meet the task requirements and user preferences are queried from the ontology.

[0157] Step 4.1 Based on the task scenario "Search for ground targets", start searching in the ontology model and find its subclasses as "Search for fixed targets" and "Search for ground moving targets". Since the object is ground personnel, it belongs to "Search for ground moving targets". Using the SPARQL query language, it is found that "Search for ground moving targets" is the parent class of "ground personnel".

[0158] Step 4.2 utilizes the SPARQL query language to search for the IRI of the "Search for Ground Moving Targets" task, which requires the ability to detect, locate, and identify targets. Based on the subclasses included in "Search for Ground Moving Targets," it is determined that the target is ground personnel, which matches the extracted task information.

[0159] Step 4.3 utilizes the SPARQL query language to query the task ontology model using the "Search for Ground Targets" query. The query reveals that "Airborne Platform" can search for ground targets, meaning it has the capability to search for "Ground Personnel," a subclass of "Search for Ground Targets." "Airborne Platform" has two subclasses, "Unmanned Aerial Vehicle" and "Firefighting Aircraft," indicating that these two subclasses have the corresponding capability to search for "Ground Personnel."

[0160] Step 4.4: There are many types of drones, such as low-altitude, medium-altitude, and high-altitude, as well as short-range, medium-range, and long-range. Which type is needed depends on the mission radius and flight altitude. The specific rules are as follows:

[0161] The activity radius is 0-15km Ultra-short-range drones Flight altitude: 0-100m ultra-low altitude drones Activity radius: 15-50km Short-range drones Flight altitude 100-1000m low-altitude drones Activity radius 50-200km Short-range drones Flight altitude 1000-7000m Medium-altitude unmanned aerial vehicles Activity radius 200-800km Medium-range unmanned aerial vehicles Flight altitude 7000-18000m High-altitude drones Activity radius greater than 800km Long-range drones Flight altitude greater than 18,000m Ultra-high altitude drones

[0162] Step 4.5: For the UAV to perform the corresponding search task, it needs to carry the appropriate payload. Based on the task capability requirements obtained in step two, it needs to discover, locate, and identify the target system, and then retrieve the required payload from the UAV itself.

[0163] Discovery: Visible light sensor, infrared sensor

[0164] Positioning: Laser rangefinder

[0165] Identification: Visible light sensor, infrared sensor

[0166] Step 4.6 filters sensors based on weather information. Assuming the weather is clear and it's nighttime, the task ontology model indicates that photoelectric sensors are not suitable at night; therefore, photoelectric sensors are filtered out. The final sensor list is as follows:

[0167] Discovery: Infrared sensor

[0168] Positioning: Laser rangefinder

[0169] Identification: Infrared sensor

[0170] Step 5 uses an analytic hierarchy process (AHP) and the Monarch Butterfly algorithm to optimize resources. The specific implementation steps are as follows:

[0171] Step 5.1: Based on the combination of the UAV and its onboard sensors obtained in step 4.6, search the resource database for resources that match the task requirements, filter out unavailable resources based on their status, and return the searched resource instances, as shown in Table 1.

[0172] Table 1 Matching resource instances

[0173]

[0174] Step 5.2 is executed to normalize the attribute values ​​of the resource instance. The normalized attribute values ​​are as follows:

[0175] (0.000021, 0.000075, 0.022021, 0.004521)

[0176] (0.000024, 0.000083, 0.012635, 0.001087)

[0177] (0.000011, 0.000091, 0.042236, 0.002804)

[0178] (0.000041, 0.000248, 0.022021, 0.001774)

[0179] (0.000033, 0.000114, 0.045124, 0.005322)

[0180] (0.000005, 0.000017, 0.027796, 0.001201)

[0181] (0.000056, 0.000013, 0.116600, 0.004178)

[0182] (0.000010, 0.000110, 0.029962, 0.000515)

[0183] Next, proceed to step 5.3, using the analytic hierarchy process (AHP) to calculate the weight of each attribute, resulting in the following judgment matrix:

[0184] (1,2,1 / 4,1 / 3)

[0185] (1 / 2, 1, 1 / 7, 1 / 5)

[0186] (4,7,1,1 / 2)

[0187] (3,5,2,1)

[0188] The calculated weights are (0.198, 0.108, 0.615, 0.755).

[0189] Finally, by executing steps 5.4, 5.5, and 5.6, the scores for all resource instances are shown in Table 3.

[0190] Table 2 Resource Instance Scoring

[0191] 1 Drone 1 0.8965 2 Drone 2 0.8365 3 Drone 3 0.8765 4 Drone 4 0.8829 5 5 drones 0.8991 6 Drone 6 0.9095 7 Drone 7 0.8813 8 8 drones 0.9032

[0192] The specific embodiments described herein are merely illustrative of the spirit of the invention. Those skilled in the art to which this invention pertains may make various modifications or additions to the described specific embodiments or use similar methods to substitute them, without departing from the spirit of the invention or exceeding the scope defined by the appended claims.

Claims

1. A method for reconnaissance resource matching and optimization based on ontology and the Monarch Butterfly algorithm, characterized in that, Includes the following steps: Step 1: Model the reconnaissance mission. The modeling mission includes the scenario, platform, sensors, mission type, mission capability requirements, weather concept and the relationships between them. Step 2: First, use ontology modeling technology to model resources in the reconnaissance field, including unmanned reconnaissance aircraft and payloads, aerial reconnaissance aircraft, radar, and reconnaissance vehicles, describe the data attributes and capabilities of the modeled resources, and build a knowledge base for the resources. Step 3: Given a task description in natural language, use LTP-based language processing technology to parse the task and extract information including reconnaissance area, reconnaissance time, reconnaissance target, and reconnaissance scene. Then, the user completes the weather conditions and resource preferences of the reconnaissance area. Step 4: Based on the task information extracted in Step 3, search for resources with corresponding capabilities from the task ontology model based on knowledge reasoning; Step 5: Optimize the matched resource instances based on the analytic hierarchy process and the monarch butterfly algorithm to select the optimal resource; After extracting the task information, step 4 uses knowledge reasoning to search for resources with corresponding capabilities from the task ontology model, which includes the following sub-steps. Step 4.1: Complete the task information obtained in Step 3 by adding the user's resource preference information. Based on the completed task information, query the knowledge base for resources that can meet the task requirements and user preferences. Step 4.2: Based on the reconnaissance scenario and reconnaissance object extracted from the task, use SPALQL language to find the parent class of the reconnaissance object from the task ontology model; Step 4.3: Based on the parent class found in Step 4.2, further search for the capability requirements corresponding to the reconnaissance mission scenario; Step 4.4: Based on the reconnaissance scenario and the reconnaissance target, use SPALQL language to find the reconnaissance platform that can perform the reconnaissance mission from the mission ontology model. Step 4.5: Further traverse the subclasses of the reconnaissance platform to obtain the set of all its subclasses; Step 4.6: Based on the capability requirements obtained in Step 4.3, find the corresponding sensors from the task ontology model; Step 4.7: Based on the task information, weather conditions, and resource preferences completed by the user, filter the platform subclasses obtained in Step 4.5 to obtain a set of resources that meet the requirements; Step 5 includes the following sub-steps: Step 5.1 Based on the abstract resource categories obtained in Step 4.6, find specific resource instances under each category; first, according to the reconnaissance radius and flight altitude information of the reconnaissance mission, search the database for resource instances that match the mission requirements, filter out occupied resource instances, and finally return the found resource instances. Step 5.2 Normalizes the various attributes of the reconnaissance resources, specifically using a standard normalization method. Step 5.3 Calculate the weight of each attribute of the resource; this is done using the analytic hierarchy process (AHP). Step 5.4 Generate an initial population. Each chromosome in the population is composed of the attributes of the resource, where each gene locus represents an attribute of the resource. In the process of generating the population, an orthogonal design method is used to generate the initial population. Step 5.5 After the initial population is generated, the population is divided into two smaller subpopulations, NP1 and NP2. Step 5.6 Adjust the location of the population; Step 5.7 Construct an evaluation function to assess the overall capability of the resources. The constructed evaluation function is as follows: （16） （17） max (18) s.t. 0< (19) In formula (16), i Let represent the weight of the i-th resource attribute, and m represent the total number of attributes; formula (17) represents the correlation between resource attributes and task requirement attributes, where The attributes that indicate the task requirements. Represents the attributes of the resource; in formula (18), and Represents the weights of the two evaluation functions; Step 5.8 Perform genetic operations on the population and evaluate individuals using an evaluation function. Stop the calculation when the algorithm reaches the maximum number of iterations. Sort the individuals according to the evaluation value. The larger the value, the better the individual.

2. The method for reconnaissance resource matching and optimization based on ontology and Monarch Butterfly algorithm according to claim 1, characterized in that, Step 1 specifically includes the following process: Step 1.1: Identify entity classes including the mission scenario and its involved reconnaissance objects, mission capability requirements, platforms, sensors, weather, etc., and complete the data attributes of each class; Step 1.2: Analyze the relationships between the entities and establish connections between them.

3. The method for reconnaissance resource matching and optimization based on ontology and Monarch Butterfly algorithm according to claim 1, characterized in that, Step 2 specifically includes the following process: Step 2.1: Identify entities and their instances according to the three levels of "abstraction-application-instance", analyze the attributes of the identified entities, and use ontology modeling technology to perform entity modeling; Step 2.2: Analyze the capabilities of each resource and associate them with tasks. Based on this, complete the applicable scenarios and task types for each resource.

4. The method for reconnaissance resource matching and optimization based on ontology and Monarch Butterfly algorithm according to claim 1, characterized in that, Given a task description in natural language, step 3, task information parsing and completion, specifically includes the following process; Step 3.1: Parse the user-posted task using the LTP Natural Language Processing Toolkit, extracting entities, attributes, and relationships to form a triple; first, analyze words and phrases, including word segmentation, part-of-speech tagging, and named entity recognition; then, decompose the entire sentence structure through the dependency relationships between words, where dependency relationships express the semantic dependencies between the sentence components; Step 3.2: Use the trained language model to extract task information, extract the relationships between specific entities in the information, and then obtain the task time and task location information.

5. The method for reconnaissance resource matching and optimization based on ontology and Monarch Butterfly algorithm according to claim 1, characterized in that, Step 5 uses an analytic hierarchy process (AHP) and the Monarch Butterfly algorithm to optimize resources, specifically including the following sub-steps; Step 5.1 Based on the abstract resource categories obtained in Step 4.6, find specific resource instances under each category; first, according to the reconnaissance radius and flight altitude information of the reconnaissance mission, search the database for resource instances that match the mission requirements, filter out occupied resource instances, and finally return the found resource instances. Step 5.2 First calculate the mean of each attribute value. and variance Then, each attribute value is normalized. The formulas for calculating the mean, variance, and normalization are shown in (1), (2), and (3): （1） （2） （3） In formulas (1), (2), and (3), i represents the i-th attribute of a resource, and j represents the j-th resource. Let represent the i-th value of the j-th resource instance, and n represent the total number of resources. This represents the normalized value; Step 5.3 Calculate the weight of each attribute of the resource; this is done using the Analytic Hierarchy Process (AHP); the specific calculation process is as follows: Construct a pairwise comparison matrix A, evaluate the relative importance of the paired attributes, and give the values ​​of the relative importance; Calculate the hierarchical single-order weight vector and perform a consistency check; Calculate the hierarchical total ranking weight vector and perform a consistency check; The formula for calculating the consistency index (CI) is as follows: （4） in Let A represent the largest eigenvalue of A; to measure the magnitude of CI, the inconsistency ratio CR is defined, and its calculation formula is as follows: （5） In formula (5), RI is the random consistency index, and its value can be obtained by looking up a table; when the value of CR is less than 0.1, it is considered that the judgment matrix A has passed the consistency test; otherwise, the element values ​​of matrix A need to be adjusted until it passes the test. Step 5.4 Generate an initial population. Each chromosome in the population is composed of the attributes of the resource, where each gene locus represents an attribute of the resource. In the process of generating the population, an orthogonal design method is used to generate the initial population. Step 5.5 After the initial population is generated, the population is divided into two smaller subpopulations, NP1 and NP2. The population size in NP1 and NP2 is set using the following formulas: （6） NP2 = NP - NP1 (7) Step 5.6 Adjust the position of the population; in formula (6), p represents the migration rate of the monarch butterfly, which is set to 0.417 with reference to existing technology. The position update method of each individual in the population is determined by formulas (8) and (9): （8） （9） in Indicates the time at the (t+1)th iteration The k-th dimension is used to determine the position of Monarch Butterfly i, and t represents the number of iterations; and The position is generated randomly; when updating the individual position, the update mechanism is determined by the parameter R; if R≤p, then formula (8) is executed to update the position, otherwise formula (9) is used to update the position: R = rand * peri (10) Where peri represents the migration period, which is set to 1.2 based on existing technology, and rand is a random number in [0,1]. If rand ≤ p, the position update formula is: （11） If rand > p, then the position update formula is: （12） If rand > p, then further, if rand > BAR, the position of the monarch butterfly can be updated using the following formula: （13） Where BAR represents the adjustment rate, its size is equal to p, and the random step size dx is calculated as follows: （14） parameter Calculate using formula (15): （15） This represents the maximum step size, and t represents the number of iterations.

Images