An intelligent engine system for unmanned swarm cognitive networking and its deployment method
By designing an intelligent engine system for unmanned swarm cognitive networking, the problems of bandwidth, network stability, and computing power in intelligent networking of unmanned swarms were solved, realizing the environmental and network cognition of unmanned nodes and improving the adaptability and robustness of unmanned swarm networks.
Patent Information
- Application Number
- CN202310998681.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-08-08
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2043-08-08
AI Technical Summary
Existing technologies cannot effectively solve the problems of bandwidth, network stability, and computing power in intelligent networking in unmanned clusters, and they are highly dependent on base stations, making it impossible to achieve intelligent networking in unmanned clusters.
Design an intelligent engine system for unmanned swarm cognitive networking, including a platform layer and a resource layer, comprising a knowledge base component, a model component, and a resource scheduling component, deployed in a lightweight and a complete manner, respectively. Through collaborative resource scheduling and distributed computing, it enables unmanned aerial vehicle (UAV) nodes to achieve environmental and network cognition.
In unmanned swarm networks with high mobility and frequent network topology changes, intelligent cognitive networking was achieved, improving the adaptability of UAV nodes and the robustness of the network.
Smart Images

Figure CN116866957B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of network switching technology, and in particular relates to an intelligent engine system and its deployment method for unmanned cluster cognitive networking. Background Technology
[0002] The "White Paper on the Development of Intelligent Unmanned Swarm Systems" mentions that the disruptive nature of unmanned swarm technology has attracted great attention from countries around the world. However, there are still no concrete solutions for how to achieve intelligent networking on unmanned drone nodes and how to solve the compatibility between intelligence and the computing, energy and storage resources of unmanned drones.
[0003] In recent years, intelligent engine systems have been proposed in intelligent networking research. These systems are practical platforms for training, learning, inference, and decision-making based on machine learning algorithms. Currently, some are geared towards product modification and game development, such as the product intelligent engine system proposed by Enterprise Service Development Company Priority Responsibility Partnership, and an AIDesigner-based artificial intelligence engine system. Other intelligent engine systems do not specify a particular application direction, such as an AI engine system testing method, platform and terminal, computing and storage devices, an AI engine system optimization method and device, a method, system and device for multi-device sharing of an intelligent engine system, and an intelligent engine system access and invocation method and related platforms, devices, equipment, and media. These designs possess the rudiments of intelligent engine systems and propose testing or optimization methods, or design cloud environments. Their massive computational and data requirements are more suitable for mature and resource-rich ground systems, and they do not focus on how to achieve cognitive networking in unmanned clusters. Some scholars have also proposed achieving intelligence based on cloud platforms, such as an artificial intelligence cloud platform and an intelligent cloud computing management system. These have high requirements for bandwidth, network stability, and computing power, or require base station support, and cannot be implemented in unmanned clusters. Summary of the Invention
[0004] To address the problems of existing technologies that place high demands on bandwidth, network stability, and computing power, or require base station support and are unsuitable for unmanned clusters, this invention provides an intelligent engine system and its deployment method for cognitive networking in unmanned clusters. This intelligent engine system and its deployment method are applicable to unmanned cluster networks with high mobility and frequent network topology changes, achieving intelligent cognitive networking. Furthermore, this invention designs deployment methods for both a lightweight intelligent engine system and a complete intelligent engine system, improving the adaptability of intelligent networking on unmanned aerial vehicle (UAV) nodes.
[0005] To achieve the above-mentioned objectives of this invention, the technical solution adopted is as follows:
[0006] An intelligent engine system for cognitive networking of unmanned swarms is provided. The intelligent engine system is deployed on each UAV node in the unmanned swarm and is used to accept calls from unmanned swarm applications. The system is characterized by comprising a platform layer and a resource layer. The platform layer includes a platform access interface for accepting data input, a knowledge base component, a model component, and a resource scheduling component. The resource layer includes a network resource pool, a spectrum optimization resource pool, a computing resource pool, and a storage resource pool.
[0007] The knowledge base component is used to complete knowledge sharing among different drone nodes within the unmanned cluster and to perform final policy calculation based on a distributed strategy.
[0008] The model component stores recognizable mathematical models, which are used to select the appropriate model and complete the modeling based on the input provided by the unmanned swarm application. The mathematical models include scene models, platform models, and network models.
[0009] The resource scheduling component is used to schedule the resource layer for storing and accessing network resources, spectrum resources, computing resources, and storage resources.
[0010] Preferably, the model component training component provides model input; the scenario model includes an unmanned swarm task scenario model, a network scenario model, and an environmental scenario model; the platform model includes an unmanned aerial vehicle (UAV) dynamics and kinematics model and a formation dynamics and kinematics model; the networking model includes a routing model, a channel access model, a resource awareness model, a resource scheduling model, and a service QoS model.
[0011] Preferably, the resource scheduling component is divided into local resource scheduling and collaborative resource scheduling based on whether the scheduled resource is located locally; local resource scheduling accesses the resource pool of the same intelligent engine system; collaborative resource scheduling accesses the resource pool of another UAV's intelligent engine system through messages, or initiates distributed intelligent computing.
[0012] Preferably, the knowledge base component includes a knowledge transfer module and a strategy library; the knowledge transfer module includes a training sample knowledge transfer submodule, a strategy transfer submodule, and a distributed strategy integration submodule; the strategy library is expanded through knowledge transfer and manual injection, including a situational trend analysis and reasoning strategy library and a network rapid construction and agile adjustment strategy library.
[0013] The distributed strategy integration submodule is a component that, after the resource scheduling component initiates distributed computing through collaborative resource scheduling, receives distributed strategies obtained from the intelligent engine systems deployed on other drone nodes, and integrates all distributed strategies into the final strategy.
[0014] The strategy exists in the form of a "current state - action to be performed" tuple. The current state includes data related to the drone state, environmental situation, and network state. The actions to be performed include prediction information of the drone state, environmental situation, and network state, task planning results, and a set of behaviors that change the drone state and network state.
[0015] The aforementioned situational trend analysis and reasoning strategy library includes environmental cognitive behavior analysis and prediction strategies, electromagnetic environment analysis and prediction strategies, cluster state analysis and prediction strategies, and task analysis and planning strategies.
[0016] The aforementioned network rapid construction and agile adjustment strategy library includes network topology construction strategies, channel access control strategies, network planning strategies, spectrum selection strategies, routing strategies, resource scheduling strategies, and service transmission control strategies.
[0017] Preferably, the platform layer further includes a training component, which includes a supervised learning sub-component, an unsupervised learning sub-component, a reinforcement learning sub-component, and a swarm intelligence sub-component; the training component selects an appropriate learning algorithm based on the input provided by the unmanned swarm application and the characteristics of the training sample set read from the training sample library.
[0018] Preferably, the platform layer further includes a platform management component, which includes a model component management module, a training component management module, a knowledge base component management module, and a resource management module. This module manages the model components, training components, knowledge base components, and resource components, and performs operations such as adding, modifying, and deleting code or data in each resource pool of the knowledge base components, model components, training components, and resource layer, and maintains each component.
[0019] Preferably, the knowledge base component further includes a training sample library, which serves as the input to the training component and stores the training sample set required for the training component to complete learning and training based on machine learning algorithms. The training sample set includes network topology samples, routing samples, network traffic samples, spectrum samples, channel access control samples, service transmission control samples, network planning samples, and task planning samples. The policy library also stores the policies formed after the training component completes learning and training.
[0020] A deployment method for an intelligent engine system for cognitive networking of unmanned swarms is characterized in that the intelligent engine system is deployed on each UAV node in the unmanned swarm, namely a lightweight deployment and a complete deployment; wherein, in the lightweight deployment, the unmanned swarm includes a command node, relay nodes, a cluster head node, and member nodes; the member nodes and the cluster head node deploy a lightweight intelligent engine system, and the command node deploys a complete intelligent engine system; wherein, in the complete deployment, the unmanned swarm includes a cluster head node and member nodes; the member nodes deploy a lightweight intelligent engine system, and the cluster head node deploys a complete intelligent engine system; the lightweight intelligent engine system is the intelligent engine system as described in any one of claims 1 to 5 or the intelligent engine system as described in any one of claims 1 to 8; the complete intelligent engine system is the intelligent engine system as described in any one of claims 1 to 8.
[0021] Preferably, both the lightweight deployment and the full deployment require the following steps:
[0022] (1) The unmanned cluster application calls the intelligent engine system through the platform access interface, and inputs the current situation information, UAV status, network status and policy requirements;
[0023] (2) Based on the current situation information, UAV status, network status, and policy requirements, index the policy base of the knowledge base to find policies that meet the requirements. If the indexing is successful, obtain the policy and proceed to step (9); otherwise, proceed to step (3).
[0024] (3) The model component selects the appropriate model for modeling based on the strategy requirements;
[0025] (4) Read training samples that meet the policy requirements from the training sample library of the knowledge base;
[0026] (5) The resource scheduling component performs local resource scheduling and schedules computing resources from the computing resource pool. If the computing resource scheduling is successful, proceed to step (6); otherwise, proceed to step (7).
[0027] (6) Based on the established model and the read training samples, the training component selects a machine learning algorithm from the training component for training to obtain the strategy and proceed to step (9).
[0028] (7) The resource scheduling component performs coordinated resource scheduling and initiates distributed computing to other drone nodes of the intelligent engine system that have deployed training components;
[0029] (8) The knowledge transfer module in the knowledge base component performs distributed strategy integration to obtain the final strategy;
[0030] (9) Output strategies to unmanned cluster applications;
[0031] (10) The platform management component manages the policy through a knowledge base and stores the policy in the policy library;
[0032] (11) After the drone node executes the strategy, it forms knowledge, and the platform management stores the knowledge through the knowledge base.
[0033] The beneficial effects of this invention are as follows:
[0034] 1. This invention targets unmanned swarm networks with high mobility and frequent network topology changes. For the first time, it proposes an intelligent engine system and its detailed design and deployment method for unmanned swarm networks. It integrates environmental cognition and unmanned swarm networking model components, knowledge base, and resource pool for networking. It outputs intelligent environmental cognition strategies and networking strategies to the UAV nodes in the swarm, so that the nodes can perform environmental cognition, task cognition, and network cognition, and complete intelligent networking based on the above cognition, thereby maintaining the robustness and agility of the network.
[0035] 2. This invention considers the insufficient computing, energy, and storage resources of UAV nodes, and designs a lightweight intelligent engine system and its deployment method, as well as a complete intelligent engine system and its deployment method. Based on the above deployment methods, knowledge transfer and collaborative resource scheduling, including distributed computing, are completed to improve the adaptability of intelligent networking on UAV nodes. Attached Figure Description
[0036] Figure 1 This is a structural diagram of an intelligent engine system according to the present invention;
[0037] Figure 2 This is a structural diagram of another intelligent engine system described in this invention;
[0038] Figure 3 This is a structural diagram of the knowledge base component of the present invention;
[0039] Figure 4 This is a schematic diagram illustrating the lightweight deployment of the unmanned cluster head intelligent engine system of the present invention;
[0040] Figure 5 This is a schematic diagram illustrating the complete deployment of the unmanned cluster head intelligent engine system of the present invention. Detailed Implementation
[0041] The embodiments of the present invention will be described below with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention and not for limiting the scope of protection of the present invention.
[0042] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.
[0043] Example 1
[0044] like Figure 1 As shown, an intelligent engine system for cognitive networking of unmanned swarms is provided. The intelligent engine system is deployed on each UAV node in the unmanned swarm and is used to accept calls from unmanned swarm applications. The intelligent engine system includes a platform layer and a resource layer. The platform layer includes a platform access interface for accepting data input, a knowledge base component, a model component, and a resource scheduling component. The resource layer includes a network resource pool, a spectrum optimization resource pool, a computing resource pool, and a storage resource pool.
[0045] The knowledge base component is used to complete knowledge sharing among different drone nodes within the unmanned cluster and to perform final policy calculation based on a distributed strategy.
[0046] The model component stores recognizable mathematical models, which are used to select the appropriate model and complete the modeling based on the input provided by the unmanned swarm application; the mathematical models include scene models, platform models, and network models.
[0047] The resource scheduling component is used to schedule the resource layer for storing and accessing network resources, spectrum resources, computing resources, and storage resources.
[0048] The model component training component provides model input; the scenario model includes unmanned swarm task scenario model, network scenario model and environmental scenario model; the platform model includes UAV dynamics and kinematics model and formation dynamics and kinematics model; the networking model includes routing model, channel access model, resource awareness model, resource scheduling model and service QoS model.
[0049] The resource scheduling component is divided into local resource scheduling and collaborative resource scheduling based on whether the scheduled resources are located locally. Local resource scheduling accesses the resource pool of the same intelligent engine system. Collaborative resource scheduling accesses the resource pool of another UAV's intelligent engine system through messages or initiates distributed intelligent computing.
[0050] like Figure 3 As shown, the knowledge base component includes a knowledge transfer module and a strategy library; the knowledge transfer module includes a training sample knowledge transfer submodule, a strategy transfer submodule, and a distributed strategy integration submodule; the strategy library is expanded through knowledge transfer and manual injection, including a situational trend analysis and reasoning strategy library and a network rapid construction and agile adjustment strategy library.
[0051] The distributed strategy integration submodule is a component that, after the resource scheduling component initiates distributed computing through collaborative resource scheduling, receives distributed strategies obtained from the intelligent engine systems deployed on other drone nodes, and integrates all distributed strategies into the final strategy.
[0052] The strategy is presented in the form of a "current state - action to be performed" tuple. The current state includes data related to the UAV state, environmental situation, and network state. The actions to be performed include prediction information of the UAV state, environmental situation, and network state, task planning results, and a set of behaviors that change the UAV state and network state.
[0053] The aforementioned situational trend analysis and reasoning strategy library includes environmental cognitive behavior analysis and prediction strategies, electromagnetic environment analysis and prediction strategies, cluster state analysis and prediction strategies, and task analysis and planning strategies; the aforementioned network rapid construction and agile adjustment strategy library includes network topology construction strategies, channel access control strategies, network planning strategies, spectrum selection strategies, routing strategies, resource scheduling strategies, and service transmission control strategies.
[0054] like Figure 2 As shown, the platform layer also includes a training component, which is one of the core components of the intelligent engine system. It serves as a carrier for loading machine learning algorithms and completes learning based on these algorithms. Through iterative processes, it summarizes historical data to form environmental awareness and networking strategies. The training component includes supervised learning sub-components, unsupervised learning sub-components, reinforcement learning sub-components, and swarm intelligence sub-components, loading supervised learning algorithms, unsupervised learning algorithms, reinforcement learning algorithms, and swarm intelligence algorithms respectively. Based on the input provided by the unmanned swarm application and the characteristics of the training sample set read from the training sample library, the training component selects an appropriate machine learning algorithm. For example, if the goal is to optimize the strategy, the reinforcement learning component can be selected for training; if the training sample set has labels, the supervised learning component can be selected for training.
[0055] like Figure 2 As shown, the platform layer also includes a platform management component, which includes a model component management module, a training component management module, a knowledge base component management module, and a resource management module. This module manages the model components, training components, knowledge base components, and resources, and performs operations such as adding, modifying, and deleting code or data in each resource pool of the knowledge base components, model components, training components, and resource layer, and maintains each component.
[0056] like Figure 3 As shown, the knowledge base component also includes a training sample library, which serves as the input to the training component and stores the training sample set required for the training component to complete learning and training based on machine learning algorithms. The training sample set includes network topology samples, routing samples, network traffic samples, spectrum samples, channel access control samples, service transmission control samples, network planning samples, and task planning samples. The policy library also stores the policies formed after the training component completes learning and training.
[0057] A deployment method for an intelligent engine system for cognitive networking in unmanned swarms, wherein the intelligent engine system is deployed on each UAV node in the unmanned swarm, and the deployment methods are lightweight deployment and full deployment; for example... Figure 4 As shown, in the lightweight deployment, the unmanned cluster includes a command node, relay nodes, a cluster head node, and member nodes. The cluster head node is connected to the command node through relay nodes. To reduce costs, both member nodes and the cluster head node deploy lightweight intelligent engine systems. When the cluster head node calls its intelligent engine system to output a policy, its lightweight intelligent engine system cannot call the training components for training due to the lack of some or all training components or insufficient sample database. Therefore, the cluster head node's lightweight intelligent engine system initiates distributed computation to the command node, which has a fully deployed intelligent engine system, through coordinated resource scheduling. After the command node completes the computation by calling its intelligent engine system, it distributes the obtained policy to the cluster head node through policy migration. The cluster head node then distributes the policy to all member nodes through policy migration by calling its intelligent engine system. Similarly, when a member node calls its intelligent engine system to output a policy, its lightweight intelligent engine system cannot call the training components for training. Therefore, its lightweight intelligent engine system first initiates distributed computation to the cluster head node through coordinated resource scheduling, and the cluster head node then initiates distributed computation to the command node.
[0058] like Figure 5As shown, in the complete deployment, the unmanned cluster includes a cluster head node and member nodes. The cluster head node deploys a complete intelligent engine system, while its member nodes deploy lightweight intelligent engine systems. When the cluster head node calls its intelligent engine system to output a strategy, the cluster head node can complete intelligent computation and obtain the strategy by calling its own intelligent engine system without needing to coordinate resource scheduling. When a member node calls its intelligent engine system to output a strategy, its lightweight intelligent engine system cannot call the training component for training. Therefore, its lightweight intelligent engine system initiates distributed computation to the cluster head node first through coordinated resource scheduling.
[0059] Both the lightweight deployment and the full deployment require the following steps:
[0060] (1) The unmanned cluster application calls the intelligent engine system through the platform access interface, and inputs the current situation information, UAV status, network status and policy requirements;
[0061] (2) Based on the current situation information, UAV status, network status, and policy requirements, index the policy base of the knowledge base to find policies that meet the requirements. If the indexing is successful, obtain the policy and proceed to step (9); otherwise, proceed to step (3).
[0062] (3) The model component selects the appropriate model for modeling based on the strategy requirements;
[0063] (4) Read training samples that meet the policy requirements from the training sample library of the knowledge base;
[0064] (5) The resource scheduling component performs local resource scheduling and schedules computing resources from the computing resource pool. If the computing resource scheduling is successful, proceed to step (6); otherwise, proceed to step (7).
[0065] (6) Based on the established model and the read training samples, the training component selects a machine learning algorithm from the training component for training to obtain the strategy and proceed to step (9).
[0066] (7) The resource scheduling component performs coordinated resource scheduling and initiates distributed computing to other drone nodes of the intelligent engine system that have deployed training components;
[0067] (8) The knowledge transfer module in the knowledge base component performs distributed strategy integration to obtain the final strategy;
[0068] (9) Output strategies to unmanned cluster applications;
[0069] (10) The platform management component manages the policy through a knowledge base and stores the policy in the policy library;
[0070] (11) After the drone node executes the strategy, it forms knowledge, and the platform management component stores the knowledge through the knowledge base component management module.
[0071] In step (7) of this embodiment, distributed computing must not be initiated to drone nodes of a lightweight intelligent engine system that lacks all training components. The intelligent engine system with training components can be a lightweight intelligent engine system that deploys some or all training components, or a complete intelligent engine system.
[0072] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the claims of the present invention.
Claims
1. An intelligent engine system for cognitive networking in unmanned swarms, wherein the intelligent engine system is deployed on each UAV node in the unmanned swarm and is used to receive calls from unmanned swarm applications, characterized in that: The intelligent engine system includes a platform layer and a resource layer. The platform layer includes a platform access interface for receiving data input, a knowledge base component, a model component, and a resource scheduling component. The resource layer includes a network resource pool, a spectrum optimization resource pool, a computing resource pool, and a storage resource pool. The knowledge base component is used to complete knowledge sharing among different drone nodes within the unmanned cluster and to perform final policy calculation based on a distributed strategy. The model component stores recognizable mathematical models, which are used to select the appropriate model and complete the modeling based on the input provided by the unmanned swarm application; the mathematical models include scene models, platform models, and network models. The resource scheduling component is used to schedule the resource layer, including the storage and retrieval of network resources, spectrum resources, computing resources, and storage resources. The scenario models include unmanned swarm mission scenario models, network scenario models, and environmental scenario models; the platform models include unmanned aerial vehicle (UAV) dynamics and kinematics models and formation dynamics and kinematics models; the networking models include routing models, channel access models, resource awareness models, resource scheduling models, and service QoS models.
2. The intelligent engine system for unmanned cluster cognitive networking according to claim 1, characterized in that: The resource scheduling component is divided into local resource scheduling and collaborative resource scheduling based on whether the scheduled resources are located locally. Local resource scheduling accesses the resource pool of the same intelligent engine system. Collaborative resource scheduling accesses the resource pool of another UAV's intelligent engine system through messages or initiates distributed intelligent computing.
3. The intelligent engine system for unmanned cluster cognitive networking according to claim 1, characterized in that: The knowledge base components include a knowledge transfer module and a strategy library; The knowledge transfer module includes a training sample knowledge transfer submodule, a policy transfer submodule, and a distributed policy integration submodule. The strategy library is expanded through knowledge transfer and manual injection, including a situational trend analysis and reasoning strategy library and a network rapid construction and agile adjustment strategy library.
4. The intelligent engine system for unmanned cluster cognitive networking according to claim 3, characterized in that: The distributed strategy integration submodule is a component that, after the resource scheduling component initiates distributed computing through collaborative resource scheduling, receives distributed strategies from the intelligent engine systems deployed on other drone nodes and integrates all distributed strategies into the final strategy. The strategy exists in the form of a "current state - action to be performed" tuple. The current state includes data related to the UAV state, environmental situation and network state. The actions to be performed include prediction information of UAV state, environmental situation and network state, task planning results and a set of behaviors that change the UAV state and network state. The aforementioned situational trend analysis and reasoning strategy library includes environmental cognitive behavior analysis and prediction strategies, electromagnetic environment analysis and prediction strategies, cluster state analysis and prediction strategies, and task analysis and planning strategies. The aforementioned network rapid construction and agile adjustment strategy library includes network topology construction strategies, channel access control strategies, network planning strategies, spectrum selection strategies, routing strategies, resource scheduling strategies, and service transmission control strategies.
5. The intelligent engine system for unmanned cluster cognitive networking according to claim 3, characterized in that: The platform layer also includes a training component, which includes a supervised learning sub-component, an unsupervised learning sub-component, a reinforcement learning sub-component, and a swarm intelligence sub-component. The training component selects an appropriate learning algorithm based on the input provided by the unmanned swarm application and the features of the training sample set read from the training sample library.
6. The intelligent engine system for unmanned cluster cognitive networking according to claim 5, characterized in that: The platform layer also includes a platform management component, which includes a model component management module, a training component module, a knowledge base component module, and a resource management component module. This module manages the model components, training components, knowledge base components, and resources, and performs operations such as adding, modifying, and deleting code or data in each resource pool of the knowledge base components, model components, training components, and resource layer, while maintaining each component.
7. The intelligent engine system for unmanned cluster cognitive networking according to claim 5, characterized in that: The knowledge base component also includes a training sample library, which serves as the input to the training component and stores the training sample set required for the training component to complete learning and training based on machine learning algorithms. The training sample set includes network topology samples, routing samples, network traffic samples, spectrum samples, channel access control samples, service transmission control samples, network planning samples, and task planning samples. The policy library also stores the policies formed after the training component completes learning and training.
8. A deployment method for an intelligent engine system for unmanned cluster cognitive networking, characterized in that: The intelligent engine system is deployed on each UAV node in the unmanned swarm, using two deployment methods: lightweight deployment and full deployment. In the lightweight deployment, the unmanned swarm includes a command node, relay nodes, a cluster head node, and member nodes. The member nodes and cluster head node deploy the lightweight intelligent engine system, while the command node deploys the full intelligent engine system. In the full deployment, the unmanned swarm includes a cluster head node and member nodes. The member nodes deploy the lightweight intelligent engine system, while the cluster head node deploys the full intelligent engine system. The lightweight intelligent engine system is the intelligent engine system described in any one of claims 1 to 4; the full intelligent engine system is the intelligent engine system described in any one of claims 5 to 7.
9. The deployment method of an intelligent engine system for unmanned cluster cognitive networking according to claim 8, characterized in that: Both the lightweight deployment and the full deployment require the following steps: (1) The unmanned cluster application calls the intelligent engine system through the platform access interface, and inputs the current situation information, UAV status, network status and policy requirements; (2) Based on the current situation information, UAV status, network status, and policy requirements, index the policy base of the knowledge base to find policies that meet the requirements. If the indexing is successful, obtain the policy and proceed to step (9); otherwise, proceed to step (3). (3) The model component selects the appropriate model for modeling based on the strategy requirements; (4) Read training samples that meet the policy requirements from the training sample library of the knowledge base; (5) The resource scheduling component performs local resource scheduling and schedules computing resources from the computing resource pool. If the computing resource scheduling is successful, proceed to step (6); otherwise, proceed to step (7). (6) Based on the established model and the read training samples, the training component selects a machine learning algorithm from the training component for training to obtain the strategy and proceed to step (9). (7) The resource scheduling component performs coordinated resource scheduling and initiates distributed computing to other drone nodes of the intelligent engine system that have deployed training components; (8) The knowledge transfer module in the knowledge base component performs distributed strategy integration to obtain the final strategy; (9) Output strategies to unmanned cluster applications; (10) The platform management component manages the policy through a knowledge base and stores the policy in the policy library; (11) After the drone node executes the strategy, it forms knowledge, and the platform management stores the knowledge through the knowledge base.
Citation Information
Patent Citations
Dynamic strategy-driven intelligent network resource planning and configuration method
CN109560968A
Artificial intelligence model online training method and system
WO2022222558A1
