Four-network integration architecture for unmanned swarm system

Abstract

Disclosed in the present invention is a four-network integration architecture for an unmanned swarm system. The four-network integration architecture has the capabilities of heterogeneous platform resource pooling, intelligent dynamic computing power allocation, and timely decision planning, so as to maximize the overall benefit. The present invention focuses on abstracting and integrating independent submodules to form a mesh topology of a swarm. The present invention designs a four-network integration architecture for an unmanned swarm system, which comprises a computing power network, a perception network, a decision network and a communication network as core modules. The structure aims to achieve efficient cooperation of all parts in the swarm, thereby improving the overall performance and adaptability of the system. The system integrates environmental perception, a swarm network modeling component, a knowledge base and a resource pool, providing an intelligent environmental perception strategy and a network modeling strategy for the interior of the swarm. Therefore, the perception of environments, tasks and networks by nodes can be facilitated, thereby completing establishment of intelligent networks, so as to ensure the characteristics of the stability and flexibility of networks.

Description

A four-network converged architecture for unmanned cluster systems Technical Field

[0001] This invention relates to a four-network converged architecture for unmanned cluster systems, belonging to the field of unmanned cluster technology. Background Technology

[0002] With the advancement of science and technology and the emergence of artificial intelligence, traditional equipment is rapidly developing towards unmanned operation. Drones, unmanned vehicles, unmanned boats, and other unmanned devices have become a hot research area. Their "unmanned" nature allows them to perform highly complex tasks that humans cannot, and they have been widely applied in military, logistics, transportation, and agriculture. However, due to the limited capabilities of individual unmanned nodes or devices, they struggle to complete tasks independently in many scenarios. Therefore, unmanned swarm collaboration is gradually demonstrating its advantages.

[0003] Unmanned swarms exhibit significantly enhanced coordination and intelligence. They leverage the advantages of flexible networking and information sharing within the swarm to expand situational awareness and mission execution capabilities. Compared to humans, they possess greater survivability, mobility, and environmental adaptability. Using unmanned swarms with low manufacturing and maintenance costs can greatly reduce human casualties in hazardous environments, while swarms with tacit collaborative capabilities can improve mission efficiency. This distributed operation mode of unmanned swarm collaboration overcomes the limitations of localized perception and mission execution of individual unmanned devices, and can play a crucial role in the future of intelligent missions.

[0004] However, with the increasing complexity of application environments and the continuous increase in task workload and difficulty, unmanned swarm platforms are under greater pressure in terms of load capacity, endurance, computing power, and decision-making. This highlights the problems of low capability, low efficiency, and lack of flexibility in the execution of tasks by unmanned swarm platforms. As an important representative of future emerging technologies, unmanned swarm platforms have become an important research direction in both military and civilian fields. In order to improve the ability of unmanned swarm platforms to operate swarm systems, it is necessary to establish a robust distributed heterogeneous framework to cope with the more complex scheduling of swarm computing power, resources, tasks, communication, and other aspects.

[0005] To address these challenges, unmanned swarm platforms need to possess the capabilities of heterogeneous platform resource pooling, intelligent dynamic allocation of computing power, and timely planning and decision-making to maximize overall efficiency. Without a standardized data fusion platform, significant human and computing resources must be consumed to standardize and understand complex tasks, inevitably leading to inefficient task planning.

[0006] Chinese patent publication number CN116860002A discloses a method for scheduling UAV swarm task resources based on a flow network model. First, the task payload entities of the UAV swarm are abstracted, virtualized, and categorized for management, establishing a method for constructing a UAV swarm resource pool. Then, the categorized payloads are connected through task relationships within the swarm to obtain task network links, further resulting in a task-layer network model. Next, the concept of task capability is introduced, assigning task capability as a node capability attribute to task-layer network nodes, which is then converted into task link flow attributes. At the UAV swarm system level, this means that the swarm's task completion status is processed as flow; the flow in the task-layer network represents the swarm's ability to complete tasks. Finally, based on an effective evaluation of the swarm's task capabilities, an improved ant colony algorithm is proposed to find the global optimum and achieve rapid convergence, realizing the optimal swarm resource scheduling strategy. This method can effectively improve the utilization rate of swarm task resource scheduling, but it only performs resource scheduling for the UAV swarm at the task level.

[0007] Chinese patent publication number CN111439382A discloses an intelligent combined unmanned aerial vehicle (UAV) system. Its key feature is the ability to combine and separate in mid-air, possessing task perception and scene understanding capabilities. The system includes a complete UAV unit, an energy distribution module, a magnetic connection module, a cluster control and networking module, and a communication reconfiguration and collaboration module. The complete UAV unit comprises a lead UAV and at least one sub-UAV. The lead UAV and at least one sub-UAV, through the energy distribution module, magnetic connection module, cluster control and networking module, and communication reconfiguration and collaboration module, coordinate to complete tasks based on task perception and scene understanding algorithms. While the system includes multiple UAV modules such as the main unit, energy, control, and communication modules, it primarily focuses on the collaborative task completion between UAV clusters and does not describe resource scheduling between clusters.

[0008] As can be seen from the patents above, current patents mainly address resource scheduling at the task level of unmanned swarms, or the existing overall frameworks for unmanned swarms do not address resource scheduling issues. Currently, there is no complete fusion framework for integrating the diverse range of unmanned aerial vehicle (UAV) systems into a heterogeneous resource system.

[0009] Therefore, it is necessary to design a four-network converged architecture for heterogeneous unmanned swarm systems, supported by the cognitive software-defined system architecture design concept, to promote the comprehensive integration of key technologies and task execution capabilities, which is of great significance for the unified monitoring, management and scheduling of tasks executed by unmanned swarm platforms. Summary of the Invention

[0010] The technical problems to be solved by this invention are: 1) the current unmanned swarm platform task execution lacks a unified monitoring and management scheduling framework; 2) existing unmanned swarm resource scheduling methods mainly deal with single resource scheduling, and do not solve the resource scheduling problem of unmanned swarms from multiple aspects such as computing power network, perception network, decision network and communication network.

[0011] To address the aforementioned technical problems, the present invention provides a four-network convergence architecture for unmanned cluster systems. This architecture features heterogeneous platform resource pooling, intelligent dynamic allocation of computing power, and timely planning and decision-making capabilities to maximize overall benefits. Its key feature is that individual sub-modules are abstracted and merged to form a cluster's mesh topology. The four-network convergence architecture is based on sensors, the physical world, actuators, and controllers, with computing power networks, sensing networks, decision-making networks, and communication networks as modules. These modules are interconnected both within and between themselves.

[0012] The modules of the perception network, communication network, and decision network consume the system computing power provided by the computing power network. By pooling heterogeneous resources within the computing power network modules, local resource scheduling, distributed heterogeneous scheduling, hybrid scheduling and optimization, and computing power data slicing based on directed acyclic graphs are established. The computing power network is packaged as a black box to the outside, providing a unified scheduling model for the entire unmanned cluster operating system.

[0013] The perception network forms an understanding of the scene, a perception of the communication link, and a perception of the computing power through three system perception modules: environmental perception, computing power perception, and link perception. Among them, a unified data format designed by humans is used to output scene understanding information to the decision network for subsequent autonomous planning.

[0014] The decision network takes the results of group perception as input and the interconnection of heterogeneous nodes as the basis to generate a task-optimal solution and send it to the controller. The solution executes autonomous planning, strategy generation, task allocation, dynamic task migration, on-demand function reorganization and physical topology reconstruction in sequence to achieve dynamic response to task scenarios and thus improve execution efficiency.

[0015] The communication network ensures that the computing network, sensing network, and decision network form a data closed loop. Through a unified interaction protocol, it realizes the interconnection between heterogeneous nodes and the protocolization of task interaction. Based on physical topology reconstruction, it performs network topology reconstruction to achieve the unification of physical topology and network topology. Finally, it completes the switching of networking protocol and transmission mode to form a complete unmanned cluster network sharing system.

[0016] The unmanned swarm bionic operating system effectively combines computing power networks, sensing networks, decision-making networks, and communication networks. By fully deploying this unmanned swarm bionic operating system on unmanned aerial vehicle (UAV) swarms, it unifies the scheduling model of the swarm computing power network, the data format of the sensing network, the functional interfaces of the decision-making network, and the interaction protocol of the communication network by calling the interfaces of different functional modules.

[0017] Preferably, the unmanned swarm bionic operating system, designed according to the principles of resource standardization, functional software implementation, application intelligence, and system reconfigurability, decouples resources, functions, and the platform. By identifying environmental conditions, processing resources and task requirements, it drives the dynamic aggregation of swarm system resources and the reconstruction of the collaborative computing environment, forming a unified perception, collaborative decision-making, coordinated control, and consistent action capability for the unmanned aerial vehicle swarm. This supports the full-domain maintenance and flexible reorganization of the swarm system, shortens task response time, and ensures efficient task completion. The specific operation process is as follows:

[0018] UAV 1, UAV 2 to UAV n exchange sensor data, decision data and control command information through a cluster network;

[0019] The cluster operating system identifies environmental conditions, processes resources and task requirements, and utilizes swarm intelligence algorithm interfaces and swarm intelligence function library interfaces to achieve collaborative scheduling among cluster perception computing, cluster decision-making, and distributed control systems.

[0020] Drive the dynamic aggregation of resources and reconstruction of the collaborative computing environment of the bee colony system to form the ability of unmanned swarms to jointly perceive, make collaborative decisions, coordinate control, and act in unison;

[0021] During the execution of pre-set tasks, the unmanned swarm can achieve autonomous obstacle avoidance and control by constantly sensing the surrounding environment and its own status.

[0022] Preferably, the computing power network is decomposed into a distributed heterogeneous resource scheduling module for indirect resource allocation and a local resource scheduling module for direct resource allocation, performing two systematic scheduling operations. One operation is executed from the unmanned cluster to the unmanned node, which is regarded as indirect computing power resource allocation based on task and role. The other operation is executed from the unmanned node to the resources within the node, which is regarded as direct computing power resource allocation. The indirect resource allocation of the distributed heterogeneous resource scheduling module depends on the specific task and communication network, while the direct resource allocation of the local resource scheduling module depends on the resource distribution within the node.

[0023] Preferably, the distributed heterogeneous resource scheduling module adopts a continuous resource distributed optimization method, specifically including the following steps:

[0024] Step A1: Problem Modeling. The distributed heterogeneous resource scheduling problem is abstracted into a mathematical model. It is assumed that individuals in the cluster can obtain information about neighboring UAVs in real time using the communication network, that is, the resource allocation vector x of neighboring UAV i at a certain moment. i For each UAV platform i, its domain is defined as the set of individuals Ω linked by a communication link. i ,Right now Ω i = {k∈N|(i,k)∈E}, where E represents the connection edge supporting bidirectional communication between drones. Assume a drone swarm consists of N = {1, 2, 3, ..., n} drones, and each drone provides M = {1, 2, 3, ..., m} types of task resources. For different tasks, the task requirement vector is represented as x. d =[x d，1 x d，2 , ..., x d，m ] T x d，j The j-type resources required to complete the task; the task requirement vector xd at a certain moment represents the resources to be allocated; the UAV platform is considered as the entity that receives the corresponding resources; let the set of resource vectors that UAV i can provide be . The feasible resource allocation set is Δ f ={x=[x1, x2, ..., x...]} n ] T};

[0025] Step A2: Define the objective function. Determine the objective function to be optimized. Assume the resource allocation combination of the UAV is x = [x1, x2, ..., x...]. n ], and let the global utility function U: Δ f →[0, +∞) is In the formula, u i Let represent the utility function of the i-th drone. Then the optimal resource allocation problem can be expressed as the following formula:

[0026] In the formula, u i (x i Let ) represent the utility function for resource allocation of the i-th UAV. This represents the maximum value of type j resources required for the i-th drone to complete its mission.

[0027] Step A3: Select an optimization algorithm to find the global optimal solution using a distributed algorithm based on local resource exchange. This algorithm first initializes the allocation value for each drone based on the total resources required for the task. Then, at each time step, each drone calculates its adaptive value using its own state and information from neighboring drones, and calculates the individual in the neighborhood with the largest difference in benefit. Next, each drone asynchronously adjusts its allocated resources based on its own resource state, its own benefit, and the average benefit of neighboring drones. If the resource increment of all individuals in the drone cluster is 0, the algorithm converges, achieving the optimal resource allocation effect globally or locally. Otherwise, it continues to calculate the adaptive value and the neighborhood adaptive value.

[0028] Preferably, the local resource scheduling module of the computing power network adopts a method based on a directed acyclic graph model, specifically including the following steps:

[0029] Step B1: Merging of multi-directed acyclic graphs:

[0030] First, merge multiple directed acyclic graph task graphs. The merging method is to add a virtual entry task node and an exit task node. The entry task node also serves as the parent node of the entry task node of each sub-directed acyclic graph, and the exit node also serves as the child task node of the exit task node of each sub-directed acyclic graph. Then, update the communication overhead and computation cost between the virtual entry task node and exit task node and the relevant task nodes.

[0031] Step B2: Task sorting:

[0032] The time lag value of the cost calculated on internal resources is used as a parameter to sort the tasks in the directed acyclic graph using a fair ranking method. The larger the time lag value, the higher the priority assigned to the task, and vice versa.

[0033] Step B3: Internal Resource Selection and Task Scheduling

[0034] Based on the sorted task set, the tasks are scheduled in descending order of priority, and the tasks are scheduled to be executed on the internal resource hardware with the earliest completion time according to the task scheduling strategy in the HEFT algorithm.

[0035] Preferably, in the sensing network, the entire unmanned sensing module is divided into three system sensing module scenarios according to task attributes: scene sensing, computing power sensing, and link sensing, wherein:

[0036] The sensing network acquires key parts of external information and understands the scene through sensing modules. It has the ability to identify and understand elements, structures and changes in various scenes, and can identify natural geographical features and complex task environment information.

[0037] The sensing network analyzes the state within the cluster, monitors and collects sensor data through the computing power sensing module, and enables the sensing network to have better pattern recognition and trend prediction capabilities.

[0038] The link awareness module is responsible for ensuring the smoothness of data transmission and the stability of communication. By monitoring the network connection status and optimizing the data transmission path, it ensures that the sensed information can be transmitted to the decision network in a timely and reliable manner.

[0039] The operation process of the sensing network in the cluster system is as follows:

[0040] The input consists of data transmitted from sensors in the physical domain and data transmitted from other nodes in the information domain cluster via the communication network. Upon reaching the cognitive domain, the data is distributed to three system perception modules via a classifier for inference using distributed intelligent algorithms. Based on the current perception device data, corresponding perception algorithms are deployed to analyze the natural geographical features, physical obstacle features, and target detection features contained in the current task environment. These features are then processed in a unified data format, and the necessary interactive information is output to the communication network for data exchange. The perceived information serves as part of the input for the decision network in subsequent task planning.

[0041] Based on the operating system's hardware resource usage, the computing power perception module monitors the load status of currently running programs in real time, predicts and retains a portion of cache resources, and collects statistics on idle resources and hands them over to the computing power network for pooling and redistribution.

[0042] Combining the currently obtained local sensing information and the cluster network status, the link sensing module analyzes the overall status of the current cluster, forming overall situational information for the task scenario. This situational information is then used as another input to the decision network via the cluster operating system.

[0043] Preferably, the input of the decision network originates from the results of unmanned swarm perception. Through in-depth analysis of the swarm perception data, the decision network gains insight into the current context of task execution and captures the underlying patterns and trends. The intelligent input based on swarm perception provides the system with a more comprehensive and real-time information foundation, supporting the subsequent generation of task-oriented optimal solutions. The task-oriented optimal solutions utilize two key technologies: dynamic task migration and on-demand function reconfiguration. Dynamic task migration empowers the system with the ability to respond dynamically in different task scenarios. On-demand function reconfiguration adds another layer of adaptability to the decision network, enabling the system to dynamically reorganize and configure its functional modules according to the nature and requirements of the task to provide optimal execution efficiency.

[0044] The operation process of the decision network in the cluster system is as follows:

[0045] Group perception data acquisition: First, the perception nodes in the cluster system collect various task-related data, which are then sent to the central processing unit of the decision network as input information.

[0046] Group perception results analysis: The decision network performs in-depth analysis of the collected group perception data;

[0047] Dynamic task migration: Based on the group perception results, the decision network assesses the characteristics and requirements of the current task scenario. Based on these assessments, the system decides whether dynamic task migration is necessary. If so, the system will adjust resource allocation, optimize network topology, or reconfigure task allocation strategy to adapt to the new environment.

[0048] Functional reconfiguration on demand: At the same time, the decision-making network reconfigures functions on demand according to the nature and requirements of the current task;

[0049] Optimal solution generation: After dynamic migration and functional reorganization, the decision network generates task-optimal decision solutions by integrating the results of group perception information, dynamic task migration, and on-demand functional reorganization.

[0050] Execution and feedback: The generated decision plan is communicated to the execution unit of the cluster system. The execution unit performs the corresponding tasks according to the plan. At the same time, the system continuously senses the environment and collects data during the execution process, providing feedback information for the next round of decision-making, forming a closed loop.

[0051] Preferably, in the communication network, the UAV communication module is deployed on each intelligent carrier within the cluster, and data transmission is carried out through the self-organizing network between nodes, integrating the business units, software radio operation framework, scene perception, computing, networking communication, and the cluster-side perception, networking and scheduling within the UAV cluster nodes.

[0052] The entire communication module is divided into two functional modules based on their functional attributes: task interaction and heterogeneous interconnection. The task interaction module is responsible for the autonomous, efficient, and accurate planning and generation of unmanned swarm tasks, environmental conditions, task coordination, and task implementation. Through local information interaction and long-range correlation, it enables distributed self-organization of unmanned swarms under communication constraints, forming a systematic interoperability capability for unmanned swarms. The heterogeneous interconnection module is responsible for the interconnection and interoperability of information between different nodes in the unmanned swarm self-organizing network. The unmanned swarms establish a self-organizing network through a unified communication module.

[0053] The operation process of the communication network in the trunking system is as follows:

[0054] The unmanned cluster is equipped with a communication module to start the route discovery function. It performs adaptive networking based on the signal-to-noise ratio of the links between different devices, selects the node with the better signal-to-noise ratio as the next hop node, and outputs the unmanned cluster network topology.

[0055] The design of the interaction mechanism and protocol for unmanned cluster tasks after network deployment is completed includes the following steps:

[0056] Step C1: Planning and Generation of Synergistic Information:

[0057] Tacit collaborative information is pre-designed and generated by the unmanned swarm task planning system according to the mission objectives and requirements, and transformed into task action instructions for unmanned swarm execution to promote mission execution. Tacit collaborative information interaction refers to the process of sharing and interacting the strategic policies of the unmanned swarm system within the group to achieve collaborative task allocation and coordination.

[0058] Step C2: Generation and processing of ad hoc messages:

[0059] Contingency response information refers to the standardized and structured information or algorithms required for unmanned swarms to handle emergencies based on changes in the situation during mission execution. Contingency response information interaction refers to the unmanned swarm system relying on tacit and collaborative information and situational selection to form a distributed action plan for the group, driving the unmanned swarm to complete a series of actions.

[0060] Step C3: Interop Message Generation and Processing:

[0061] Interoperability messages are formatted messages that enable network interconnection, information exchange, and system interoperability between platforms within an unmanned swarm and between unmanned systems and other heterogeneous platforms. This enables basic functions such as situational awareness sharing, command and control, task coordination, and intelligence distribution during unmanned system mission execution. Interoperability message interaction refers to the distributed action plan formed by the unmanned swarm system based on ad hoc information exchange. It involves formulating interoperability instructions for the unmanned system platform's own control units, sensor units, and equipment units, and then transmitting interoperability messages through the unmanned swarm network to achieve coordination among formations, sensors, and equipment.

[0062] Preferably, the tacit collaborative information exchange in step C1 specifically includes the following steps:

[0063] Step C1.1: Unmanned Cluster Task Information Planning:

[0064] Based on the analysis and task decomposition of unmanned swarm mission objectives in typical mission concepts, strategic policies are planned.

[0065] Step C1.2: Unmanned Cluster Task Process Planning:

[0066] The planning of unmanned swarm task scheduling and execution processes is carried out. To explore the generation and evolution of unmanned swarm control, task transition relationship planning involves planning the conditions and methods for task transitions.

[0067] Step C1.3: Output of the unmanned cluster task solution:

[0068] In order to form a loop from task planning scheme to scheme-based messages and then to semi-physical simulation testing, and generate standard format unmanned swarm task schemes; during the planning of task schemes, each element node of the system should automatically associate with the scheme-based message data in the unmanned swarm data element dictionary, and automatically store the planned relevant data into the scheme-based message case set database.

[0069] The emergency response information generation and processing in step C2 specifically includes the following steps:

[0070] Step C2.1: Perform rule construction, including the construction of rules for unmanned groups and individual rules:

[0071] Unmanned swarm rule construction includes the construction of formation rules and algorithms, formation transformation rules and algorithms, leader succession rules and algorithms, and detection rules and algorithms for unmanned swarms; unmanned individual rule construction includes the construction of safe flight rules and algorithms, search payload usage rules and algorithms, and execution rules for unmanned individuals.

[0072] Step C2.2: Execution rule management, visually edit and manage execution rules and algorithms, and customize knowledge according to the task requirements of the unmanned cluster;

[0073] Step C2.3: Behavior flow editing. Add atomic behaviors to the individual and group models of the unmanned cluster, and build the behavior state flow. Behavior editing includes atomic behaviors, method lists, and behavior state flow editing. Among them, atomic behaviors are common behaviors that cannot be further divided. The method list combines multiple atomic behaviors to form an atomic behavior flow. Through behavior state flow editing, atomic behaviors or combined behaviors can be set in series, and trigger condition attributes can be set.

[0074] The interoperability message generation process in step C3 specifically includes the following steps:

[0075] Step C3.1: Interoperability message editing and formulation. Edit and formulate the message commands that the unmanned system control unit, sensor unit, and equipment unit may transmit during the simulation. When editing the interoperability message, automatically associate it with the interoperability message data in the data element dictionary to ensure that the message name, data item name, bit position, number of bits, and other information are consistent with the data element dictionary.

[0076] Step C3.2: Message compilation. The interoperability message compilation plugin is invoked to digitize the information transmitted by the unmanned system control unit, sensor unit, and equipment unit, and a demonstration is executed in the simulation test environment. The interoperability message editing module edits and defines the interoperability message case content and stores it in the interoperability message case set database. The interoperability message compilation plugin is then invoked again to convert the interoperability message content into a bit data stream, which is then demonstrated through simulation.

[0077] Step C3.3: Message command simulation operation. The interoperability message transmission process is demonstrated and verified through the data link model of the task simulation unit. In the simulation system, the interoperability messages are transmitted by calling the data link models of the unmanned system control unit and sensor unit. The message transmission content and process are displayed in the visualization subsystem.

[0078] Compared with the prior art, the present invention has the following beneficial effects:

[0079] 1. This invention focuses on abstracting and merging independent sub-modules to form a clustered mesh topology. This invention designs a four-network converged architecture for an unmanned cluster system, including a computing network, a perception network, a decision-making network, and a communication network as core modules. This structure aims to achieve efficient collaboration among various parts within the cluster, improving the overall performance and adaptability of the system. This system integrates environmental perception and cluster network modeling components, a knowledge base, and a resource pool, providing intelligent environmental perception and network modeling strategies for the cluster. This helps nodes perceive the environment, tasks, and network, thereby completing the intelligent network construction and ensuring the network maintains robust and flexible characteristics.

[0080] 2. To address the problem of insufficient computing, energy, and storage resources, this invention designs a distributed heterogeneous resource scheduling module for indirect resource allocation and a local resource scheduling module for direct resource allocation. Each module employs a different resource allocation method, achieving collaborative resource scheduling, including distributed computing, to improve the adaptability of intelligent networking in cluster systems. Attached Figure Description

[0081] Figure 1 is a diagram of a four-network converged architecture for an unmanned cluster system according to the present invention.

[0082] Figure 2 is a diagram of the unmanned cluster bionic operating system architecture of the present invention;

[0083] Figure 3 is a diagram of the computing power network architecture of the present invention;

[0084] Figure 4 is a diagram of the sensing network architecture of the present invention;

[0085] Figure 5 is a diagram of the decision network architecture of this invention;

[0086] Figure 6 is a diagram of the communication network architecture of the present invention;

[0087] Figure 7 is a flowchart of the distributed heterogeneous resource scheduling module method of the computing power network of the present invention;

[0088] Figure 8 is a flowchart of the local resource scheduling module method of the computing power network of the present invention;

[0089] Figure 9 is a flowchart of the communication network unmanned cluster task interaction mechanism and protocol method of the present invention.

[0090] Figure 10 is a flowchart of the tacit collaborative information interaction method of the present invention;

[0091] Figure 11 is a flowchart of the emergency response information generation and processing method of the present invention;

[0092] Figure 12 is a flowchart of the interoperability message generation and processing method of the present invention. Detailed Implementation

[0093] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the teachings of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined by the appended claims.

[0094] Example 1:

[0095] Please refer to Figure 1, which is a diagram of a four-network convergence architecture for an unmanned cluster system according to the present invention. This embodiment provides the architectural components of the present invention, which specifically include:

[0096] This invention abstracts and merges individual sub-modules to form a clustered mesh topology, designing a four-network fusion architecture for an unmanned swarm system based on sensors, the physical world, actuators, and controllers, with computing networks, sensing networks, decision networks, and communication networks as modules. The modules within and between each other are interconnected, and this invention features a functional redesign of each module.

[0097] The operation of the modules in the perception network, communication network, and decision network all consume system computing power. Heterogeneous resources within these modules are pooled and reorganized to establish local resource scheduling, distributed heterogeneous scheduling, hybrid scheduling and optimization based on directed acyclic graphs, and computing power data slicing. Externally, this is packaged as a black box, providing a unified scheduling model for the entire unmanned cluster operating system. The perception network, through three system perception modules—environmental perception, computing power perception, and link perception—forms an understanding of the scene, a perception of communication links (protocol switching and transmission modes), and a perception of computing power, respectively. A manually designed unified data format is used as scene understanding information output to the decision network for subsequent autonomous planning. The decision network, using the group perception results as input and based on the interconnection of heterogeneous nodes, can generate task-optimal solutions and send them to the controller. These solutions sequentially execute autonomous planning, policy generation, task allocation, dynamic task migration, on-demand function reorganization, and physical topology reconstruction, achieving dynamic response to task scenarios and improving execution efficiency. The communication network is a key factor in ensuring the formation of a data loop among the three networks. By designing a unified interaction protocol, the interconnection and task interaction between heterogeneous nodes are protocolized. Based on physical topology reconstruction, network topology reconstruction is performed to achieve the unification of physical and network topologies. Ultimately, functions such as networking protocol and transmission mode switching are completed, forming a complete unmanned swarm network sharing system. The unmanned swarm operating system effectively combines the above-mentioned computing network, sensing network, decision-making network, and communication network. The system is fully deployed on the unmanned swarm, unifying the scheduling model of the swarm computing network, the data format of the sensing network, the functional interfaces of the decision-making network, and the interaction protocol of the communication network by calling the interfaces of different functional modules.

[0098] Example 2:

[0099] Please refer to Figure 2, which is a diagram of the unmanned cluster biomimetic operating system architecture of the present invention. This architecture is included in the four-network converged architecture of Embodiment 1, and the specific technical solution is as follows:

[0100] Unmanned aerial vehicles (UAVs) 1 through 2 interact via a swarm network, exchanging sensor data, decision data, and control commands. The swarm operating system identifies environmental conditions, processes resources and task requirements, and utilizes swarm intelligence algorithm and function library interfaces to achieve coordinated scheduling between swarm perception computing, swarm decision-making, and the distributed control system. This drives dynamic aggregation of swarm system resources and reconstruction of the collaborative computing environment, forming a unified perception, collaborative decision-making, coordinated control, and consistent action capability for the unmanned swarm. During the execution of pre-set tasks, the unmanned swarm continuously perceives its surrounding environment and its own status, enabling autonomous obstacle avoidance and control.

[0101] Example 3:

[0102] Please refer to Figure 3, which is a diagram of the computing power network architecture of the present invention. This architecture is included in the four-network converged architecture of Embodiment 1, and the specific technical solution is as follows:

[0103] The computing power network can be decomposed into a distributed heterogeneous resource scheduling module for indirect resource allocation and a local resource scheduling module for direct resource allocation. It performs two systematic scheduling processes: one from the unmanned cluster to the unmanned nodes (which can be considered indirect computing power resource allocation based on task and role assignment), and the other from the unmanned nodes to their internal resources (which can be considered direct computing power resource allocation). The indirect resource allocation of the distributed heterogeneous resource scheduling module depends on the specific task and communication network. Without a suitable task, it's impossible to discuss task allocation or analyze the specific constraints the task places on the unmanned cluster's power consumption, flight speed, and distribution location. The direct resource allocation of the local resource scheduling module depends on the resource distribution within the nodes, such as the number of CPU cores, clock speed, GPU, memory, CPU-GPU bus, and flight control efficiency.

[0104] Example 4:

[0105] Please refer to Figure 4, which is a diagram of the sensing network architecture of the present invention. This architecture is included in the four-network convergence architecture of Embodiment 1, and the specific technical solution is as follows:

[0106] The unmanned sensing module is deployed on various intelligent carriers within the cluster, establishing a training sample knowledge base to process raw data from the physical and information domains. Based on the input received by the unmanned cluster application and the characteristics of the training sample set read from the training sample library, appropriate machine learning algorithms are selected. The sensing network is characterized by dividing the entire unmanned sensing module into three system sensing modules according to task attributes: scene sensing, computing power sensing, and link sensing. The scene sensing module is the key part of the sensing network to acquire external information and understand the scene. It has the ability to identify and understand elements, structures, and changes in various scenes, and can identify natural geographical features, complex task environments, and other information. This information provides necessary real-time semantic data support for overall decision-making. The computing power sensing module plays a role in the sensing network's state analysis within the cluster, undertaking the important task of monitoring and collecting sensor data, and enabling the sensing network to have better pattern recognition and trend prediction capabilities. This intelligent data processing method helps the decision-making network make timely decisions in complex situations. The link sensing module is responsible for ensuring the smoothness of data transmission and the stability of communication. By monitoring network connection status and optimizing data transmission paths, it ensures that sensing information can be transmitted to the decision-making network in a timely and reliable manner. This mechanism helps ensure that the sensing network can maintain efficient operation when facing complex environments and changes.

[0107] Example 5:

[0108] Please refer to Figure 5, which is a diagram of the decision network architecture of this invention. This architecture is included in the four-network convergence architecture of Embodiment 1, and the specific technical solution is as follows:

[0109] The decision network's input originates from the results of unmanned swarm perception, enabling it to extract key data from a broad range of information. Through in-depth analysis of swarm perception data, the decision network can gain insights into the current context of task execution and capture underlying patterns and trends. Intelligent input based on swarm perception provides the system with a more comprehensive and real-time information foundation, supporting the subsequent generation of task-optimized solutions. The task-optimized solutions utilize two key technologies: dynamic task migration and on-demand function reconfiguration. Dynamic task migration empowers the system with the ability to respond dynamically in different task scenarios. This means the system can adjust its decision-making strategies in real time based on the characteristics and requirements of the current task to best adapt to different working environments. This flexibility allows the decision network to maintain efficient operation in complex and ever-changing situations, providing a more intelligent and practical solution for task execution. On-demand function reconfiguration adds another layer of adaptability to the decision network. The system can dynamically reorganize and configure its functional modules according to the nature and requirements of the task to provide optimal execution efficiency. This capability allows the decision network to meet the needs of different tasks in a more refined and customized manner, thereby achieving a more optimized decision-making and execution process in various complex situations.

[0110] Example 6:

[0111] Please refer to Figure 6, which is a diagram of the communication network architecture of the present invention. This architecture is included in the four-network converged architecture of Embodiment 1, and the specific technical solution is as follows:

[0112] The UAV communication module is deployed on various intelligent carriers within the cluster. Data transmission is achieved through an inter-node self-organizing network, integrating the business units (autonomous task coordination and real-time planning), software-defined radio operating framework (resource scheduling and internal communication), scene perception, computing (intelligent reasoning and decision-making), network communication (communication parameters and schemes), and the cluster-side perception, networking, and scheduling within the UAV swarm nodes. The communication network is characterized by dividing the entire communication module into two functional modules based on their functional attributes: task interaction and heterogeneous interconnection. The task interaction module is responsible for generating UAV swarm mission planning, handling battlefield situations, task coordination, and ensuring more autonomous, efficient, and accurate mission implementation. This makes decision-making and control less or noly dependent on communication; through local information interaction and long-range correlation, it enables distributed self-organization of the UAV swarm under communication constraints, forming a systematic interoperability capability for the UAV swarm system. The heterogeneous interconnection module is responsible for the interconnection of information between different nodes in the UAV self-organizing network. UAV nodes establish an inter-node self-organizing network through a unified communication module.

[0113] Example 7:

[0114] Please refer to Figure 7, which is a flowchart of the distributed heterogeneous resource scheduling module method for computing power networks according to the present invention. This method is implemented in Embodiment 3 and includes the following steps:

[0115] Step A1: Problem Modeling. The distributed heterogeneous resource scheduling problem is abstracted into a mathematical model. It is assumed that all individuals in the cluster can obtain information about neighboring drones in real time using the communication network, i.e., the resource allocation vector x of neighboring drone i at a certain moment. i For each UAV platform i, its domain is defined as the set of individuals Ω linked by a communication link. i ,Right now Ω i ={k∈N|(i,k)∈E}, where E represents the connecting edge supporting bidirectional communication between UAVs. The UAV swarm communication network considered in this invention is a connected graph. Assuming the UAV swarm consists of N = {1, 2, 3, ..., n} UAVs, in order to complete "meta-tasks" such as formation flight, assembly and division of labor, search and detection, and target tracking, the UAVs need to provide M = {1, 2, 3, ..., m} types of task resources. The task requirement vector for different tasks can be represented as x. d =[x d，1 x d,2 ,…,x d，m ] T x d,j The j-type resources required to complete the task. The task requirement vector x at a given moment. d For resources to be allocated, the drone platform can be viewed as an entity that receives the corresponding resources. Let the set of resource vectors that drone i can provide be denoted as . The feasible resource allocation set is Δ f ={x=[x1, x2, ..., x...]} n ] T Due to the inherent differences among drones, each drone will generate different benefits when receiving the same mission resources.

[0116] Step A2: Define the objective function. Determine the objective function to be optimized. Assume the resource allocation combination of the UAV is x = [x1, x2, ..., x...]. n ], and let the global utility function U: Δ f →[0, +∞) is In the formula, u i Let represent the utility function of the i-th drone. Then the optimal resource allocation problem can be expressed as the following formula:

[0117] In the formula, u i (xi Let ) represent the utility function for resource allocation of the i-th UAV. This represents the maximum value of type j resources required for the i-th drone to complete its mission.

[0118] Step A3: Select an appropriate optimization algorithm to find the global optimum using a distributed algorithm based on local resource exchange. This algorithm first initializes the resource allocation for each individual drone based on the total resources required for the task. Then, at each time step, each drone calculates its adaptivity value using its own state and information from neighboring drones, and identifies the drone with the largest difference in reward among its neighbors. Next, each drone asynchronously adjusts its allocated resources based on its own resource state, its own reward, and the average reward of its neighboring drones. If the resource increment for all individuals in the drone cluster is 0, the algorithm converges, achieving a globally or locally optimal resource allocation effect; otherwise, it continues to calculate the adaptivity value and the neighborhood adaptivity value.

[0119] Example 8:

[0120] Please refer to Figure 8, which is a flowchart of the local resource scheduling module method for the computing power network of the present invention. This method is implemented in Embodiment 3 and includes the following steps:

[0121] Step B1: Merging Multiple Directed Acyclic Graphs. First, merge multiple directed acyclic graph task graphs by adding a virtual entry task node and an exit task node. The entry task node simultaneously serves as the parent node of the entry task node in each sub-directed acyclic graph, and the exit node simultaneously serves as the child task node of the exit task node in each sub-directed acyclic graph. Then, update the communication overhead and computational cost between the virtual entry and exit task nodes and the relevant task nodes.

[0122] Step B2: Task Ranking. A fair ranking method will be used, employing the time lag value of the cost calculated on internal resources as a parameter to rank the tasks in the directed acyclic graph. A larger time lag value assigns a higher priority to the task, and vice versa.

[0123] Step B3: Internal Resource Selection and Task Scheduling. Based on the sorted task set, the tasks are scheduled sequentially from highest to lowest priority, and the tasks are scheduled to be executed on the internal resource hardware with the earliest completion time according to the task scheduling strategy in the HEFT algorithm.

[0124] Example 9:

[0125] Please refer to Figure 9, which is a flowchart of the unmanned cluster task interaction mechanism and protocol method of the present invention. This method is implemented in Embodiment 6 and includes the following steps:

[0126] Step C1: Planning and Generation of Synergistic Information. Synergistic information is pre-designed and generated by the unmanned swarm task planning system based on the task objectives and requirements, and transformed into task action instructions for execution by the unmanned swarm to advance the task. Synergistic information interaction refers to the process of sharing and exchanging strategic policies within the unmanned swarm system to achieve collaborative task allocation and coordination.

[0127] Step C2: Generation and Processing of Contingency Messages. Contingency information refers to the standardized and structured representation of information or algorithms required by the unmanned swarm to handle unexpected situations based on changes in the battlefield situation during mission execution. Contingency information interaction refers to the unmanned swarm system selecting appropriate task methods and rules based on tacit coordination information and the battlefield situation to form a distributed action plan for the group, driving the unmanned swarm to complete a series of coherent task actions.

[0128] Step C3: Interoperability Message Generation and Processing. Interoperability messages are formatted messages that enable network interconnection, information exchange, and system interoperability between platforms within an unmanned swarm and between unmanned systems and other heterogeneous platforms. This allows for basic functions such as situational awareness sharing, command and control, task coordination, and intelligence distribution during unmanned system mission execution. Interoperability message interaction refers to the unmanned swarm system developing distributed action plans based on ad-hoc information exchange. This involves defining interoperability instructions for the unmanned system platform's control units, sensor units, and equipment units, and then transmitting interoperability messages through the unmanned swarm network to achieve coordination among formation, sensors, and equipment elements.

[0129] Example 10:

[0130] Please refer to Figure 10, which is a flowchart of the tacit collaborative information interaction method of the present invention. This method is implemented in step C1 of embodiment 9, and includes the following steps:

[0131] Step C1.1: Unmanned Swarm Mission Information Planning. Based on the analysis and decomposition of unmanned swarm mission objectives in typical mission concepts, strategic policies such as unmanned swarm mission objectives, payloads, and formation are planned.

[0132] Step C1.2: Unmanned Swarm Task Process Planning. This involves planning the flow of planned and executed tasks for the unmanned swarm. To explore the generation and evolution of unmanned swarm control, task transition relationship planning involves planning the conditions and methods for transitions between tasks. Examples include unmanned swarm branching, succession, self-looping, and dynamic task planning.

[0133] Step C1.3: Output of Unmanned Cluster Task Plan. To form a loop from task planning to plan-based messages, and then to hardware-in-the-loop simulation testing, a standard-format unmanned cluster task plan is generated. During the task plan planning process, each element node of the system automatically associates with the plan-based message data in the unmanned cluster data element dictionary, and automatically stores the planned relevant data in the plan-based message case set database.

[0134] Example 11:

[0135] Please refer to Figure 11, which is a flowchart of the emergency response information generation and processing method of the present invention. This method is implemented in step C2 of Embodiment 9, and includes the following steps:

[0136] Step C2.1: Rule construction, including the construction of unmanned swarm rules and individual rules. Unmanned swarm rule construction involves building task knowledge such as formation rules and algorithms, formation change rules and algorithms, leader succession rules and algorithms, and detection rules and algorithms for the unmanned swarm. Unmanned individual rule construction involves building rule knowledge such as safe flight rules and algorithms, search payload usage rules and algorithms, and execution rules for individual unmanned personnel.

[0137] Step C2.2: Execution rule management, which involves visually editing and managing execution rules and algorithms, and customizing them according to the task requirements of the unmanned cluster. It supports editing, modifying, saving, and customizing execution rules and algorithms for both unmanned clusters and individual unmanned units.

[0138] Step C2.3: Behavior Flow Editing. This step adds atomic behaviors to the individual and group models of the unmanned swarm and constructs the behavior state flow. Behavior editing includes atomic behaviors, a method list, and behavior state flow editing. Atomic behaviors are common, indivisible actions, such as maneuvering to a designated point or turning the radar on / off. The method list combines multiple atomic behaviors to form an atomic behavior flow. Behavior state flow editing allows for the chaining of atomic behaviors or combined behaviors and the setting of trigger condition attributes.

[0139] Example 12:

[0140] Please refer to Figure 12, which is a flowchart of the interoperability message generation and processing method of the present invention. This method is implemented in step C3 of embodiment 9, and includes the following steps:

[0141] Step C3.1: Interoperability Message Editing and Specification. This step involves editing and specifying the message commands that the unmanned system control unit, sensor unit, and equipment unit may transmit during the simulation. When editing interoperability messages, the system should automatically associate them with the interoperability message data in the data element dictionary to ensure that information such as message name, data item name, bit position, and number of bits is consistent with the data element dictionary.

[0142] Step C3.2: Message Compilation. The interoperability message compilation plugin is invoked to digitize the information transmitted by the unmanned system control unit, sensor unit, and equipment unit, and to demonstrate this in a simulation test environment. The interoperability message editing module edits and defines the interoperability message case content, storing it in the interoperability message case set database. Then, the interoperability message compilation plugin is invoked to convert the interoperability message content into a bit data stream, which is then demonstrated through simulation.

[0143] Step C3.3: Message command simulation execution demonstrates and verifies the interoperability message transmission process using the data link model of the task simulation unit. In the simulation system, interoperability messages are transmitted by calling the data link models of the unmanned system control unit, sensor unit, and equipment unit, and the message transmission content and process are displayed in the visualization subsystem.

Claims

1. A four-network fusion architecture for unmanned swarm system, characterized in that, The individual sub-modules are abstractly fused to form a mesh topology of a cluster, and the four-network fusion architecture is based on sensors, a physical world, actuators, and controllers, and the computing power network, the perception network, the decision network, and the communication network are modules, and each module is connected to each other within and between the modules; The modules of the perception network, the communication network, and the decision network consume the system computing power provided by the computing power network, and the heterogeneous resource reorganization and pooling within the computing power network module is used to establish local resource scheduling based on a directed acyclic graph, distributed heterogeneous scheduling, hybrid scheduling and optimization, and computing power data slicing; The computing power network is packaged as a black box to provide a unified scheduling model for the entire unmanned cluster operating system; The perception network forms the understanding of the scene, the perception of the communication link, and the perception of the computing power through three system perception modules of environment perception, computing power perception, and link perception, wherein a unified data format designed by an artificial intelligence is used as the scene understanding information output to the decision network for subsequent autonomous planning; The decision network takes the group perception results as input and is based on the heterogeneous node interconnection to generate a task-oriented optimal guidance scheme and send it to the controller, which sequentially executes autonomous planning, strategy generation, task allocation, task dynamic migration, function on-demand reorganization, and physical topology reconstruction to achieve dynamic response to the task scene and improve the execution efficiency; The communication network ensures that the computing power network, the perception network, and the decision network form a data closed loop, and through a unified interaction protocol, the interconnection and task interaction protocol between heterogeneous nodes are realized, the network topology is reconstructed based on physical topology reconstruction, the physical topology and the network topology are unified, the networking protocol and the transmission mode switching are finally completed, and a complete unmanned cluster network sharing system is formed; The unmanned cluster bionic operating system effectively combines the computing power network, the perception network, the decision network, and the communication network, and deploys the unmanned cluster bionic operating system on the unmanned cluster to call different functional module interfaces and unify the scheduling model of the cluster computing power network, the data format of the perception network, the functional interface of the decision network, and the interaction protocol of the communication network.

2. The four-network converged architecture for unmanned swarm system of claim 1, wherein, The unmanned cluster bionic operating system decouples resources, functions, and platforms according to the design idea of resource standardization, functional software, application intelligence, and system reconfigurability, identifies the environmental situation, processes resources and task requirements, drives the dynamic aggregation of cluster system resources and the reconstruction of collaborative computing environment, forms the joint perception, collaborative decision-making, coordinated control, and consistent action ability of the unmanned cluster, supports the global maintenance and flexible reorganization of the cluster system, shortens the task response time, ensures the efficient completion of the task, and the specific operation process is as follows: The unmanned aerial vehicles 1, 2, and n realize the interaction of sensing data, decision data, and control instruction information through the cluster network; The cluster operating system identifies the environmental situation, processes resources and task requirements, uses the group intelligence algorithm interface and the group intelligence function library interface to realize the collaborative scheduling between the cluster perception computing, cluster decision-making, and distributed control system; The system drives the dynamic aggregation of the cluster system resources and the reconstruction of the collaborative computing environment, forms the joint perception, collaborative decision-making, coordinated control, and consistent action ability of the unmanned cluster, In the process of performing preset tasks, the unmanned cluster continuously senses the surrounding environment and its own state to complete autonomous obstacle avoidance and control.

3. The four-network converged architecture for unmanned swarm system of claim 1, wherein, The computing power network is divided into a distributed heterogeneous resource scheduling module for indirect resource allocation and a local resource scheduling module for direct resource allocation, and two systematic scheduling are performed, one of which is indirect computing power resource allocation for task and role allocation from the unmanned cluster to the unmanned node, and the other of which is direct computing power resource allocation from the unmanned node to the internal resource of the node, wherein the indirect resource allocation of the distributed heterogeneous resource scheduling module depends on the specific task and the communication network, and the direct resource allocation of the local resource scheduling module depends on the resource distribution in the node.

4. The four-network converged architecture for unmanned swarm system of claim 3, wherein, The distributed heterogeneous resource scheduling module adopts a continuous resource distributed optimization method, which specifically includes the following steps: Step A1: problem modeling, abstracting the distributed heterogeneous resource scheduling problem into a mathematical model, assuming that each individual in the cluster can use the communication network to obtain real-time information of adjacent UAVs, i.e. the resource allocation vector x of adjacent UAV i at a certain moment i ; for each UAV platform i, define its field as the set of individuals linked by the communication link Ω i , i.e. In the formula, E represents a connection edge supporting two-way communication between unmanned aerial vehicles. It is assumed that the unmanned aerial vehicle cluster is composed of N = {1, 2, 3,..., n} unmanned aerial vehicles, and the unmanned aerial vehicles provide M = {1, 2, 3,..., m} types of task resources. The task demand vector of the unmanned aerial vehicles facing different tasks is represented as x d = [x d，1 , x d，2 , …, x d，m ] T , x d,j is the jth type of resource required to complete the task. The task demand vector x d at a certain moment is the resource to be allocated, and the unmanned aerial vehicle platform is regarded as an entity that accepts the corresponding resources provided. Let the resource vector set that the unmanned aerial vehicle i can provide be Then the feasible resource allocation set is Δ f = {x = [x1, x2,..., x n N] | x1, x2,..., x T N ∈ {0, 1, 2,..., 7}} ; Step A2: Objective function definition, determine the objective function that needs to be optimized, assuming that the resource allocation combination of the UAV is x = [x1, x2,..., xN], and let the global utility function U: Δ n → [0, +∞) be f ​ In the formula, u i Let u In the formula, u i (x i ) represents the utility function of the i-th UAV resource allocation, The maximum value of the jth resource required by the ith unmanned aerial vehicle to complete the task is represented. Step A3: Select an optimization algorithm to solve the global optimal solution based on a distributed algorithm of local resource exchange, which first initializes the allocation value of each unmanned aerial vehicle individual according to the total resource required by the task, then at each time, the unmanned aerial vehicle individual calculates the fitness value using the information of its own state and adjacent unmanned aerial vehicles, and calculates the individual with the largest yield difference in the neighborhood, then each unmanned aerial vehicle adjusts the allocated resource asynchronously according to its own resource state, its own yield and the average yield of adjacent unmanned aerial vehicles, if the resource increment of all individuals in the unmanned aerial vehicle cluster is 0, the algorithm converges, and the global or local optimal resource allocation effect is achieved, otherwise the fitness value and the neighborhood fitness value are continuously calculated.

5. The four-network converged architecture for unmanned swarm system of claim 3, wherein, The local resource scheduling module of the computing power network adopts a directed acyclic graph model method, which specifically includes the following steps: Step B1: Multiple directed acyclic graph merging: First, merge multiple directed acyclic graph task graphs by adding a virtual entry task node and an exit task node, the entry task node simultaneously serves as the parent node of each sub-directed acyclic graph entry task node, and the exit node simultaneously serves as the child task node of each sub-directed acyclic graph exit task node; then update the communication overhead and calculation cost of the virtual entry task node and the exit task node and related task nodes; Step B2: Task sorting: A fair sorting method is used to calculate the time lag value of the internal resource as a parameter, and the directed acyclic graph tasks are sorted, the larger the time lag value, the higher the priority given to the task, and vice versa, the smaller the time lag value, the smaller the priority given to the task; Step B3: Internal resource selection and task scheduling: According to the sorted task set, the task set is scheduled in order from high to low, and the task scheduling strategy in the HEFT algorithm is used to schedule the task to the internal resource hardware that makes the task completion time earliest.

6. The four-network converged architecture for unmanned swarm system of claim 1, wherein, In the perception network, the entire unmanned perception module is divided into three system perception modules of scene perception, computing power perception and link perception according to the task attributes, wherein: The perception network obtains external information and understands the key part of the scene through the perception module, has the ability to identify and understand elements, structures and changes in various scenes, and can identify natural geographical features and complex task environment information; The perception network analyzes the state of the cluster through the computing power perception module, monitors and collects sensor data, and enables the perception network to have better pattern recognition and trend prediction capabilities; The link perception module is responsible for ensuring the smoothness of data transmission and the stability of communication, and ensures that the perception information can be transmitted to the decision network in a timely and reliable manner by monitoring the network connection state and optimizing the data transmission path; The operation process of the perception network in the cluster system is as follows: The input is the data and information domain of the sensor returned by the physical domain and the data transmitted by the communication network of the other nodes in the cluster; the data reaches the cognitive domain, is distributed to the three system perception modules by the classifier for distributed intelligent algorithm reasoning; according to the current perception device data, the corresponding perception algorithm is deployed to analyze the natural geographical features, physical obstacle features and target detection features contained in the current task environment, and the information to be interacted is output to the communication network for data exchange, and the perception information is used as part of the input of the decision network for subsequent task planning; According to the hardware resource occupation of the operating system, the computing power perception module monitors the load state of the currently running program in real time, predicts and reserves a part of cache resources, and counts the idle resources and gives them to the computing power network for pooling and redistribution; Combined with the current local perception information and the state of the cluster network, the link perception module analyzes the overall state of the current cluster to form the overall situation information of the task scene. The channel formed by the cluster operating system inputs the situation information as another part of the decision network.

7. The four-network converged architecture for unmanned swarm system of claim 1, wherein, The input of the decision network comes from the group perception result, and through in-depth analysis of the group perception data, the decision network understands the current situation of task execution and captures the potential patterns and trends behind the task; The intelligent input based on group perception provides a more comprehensive and real-time information basis for the system, and supports the subsequent task-oriented optimal oriented scheme generation; the task-oriented optimal oriented scheme uses two key technologies of task dynamic migration and function on-demand reorganization, wherein the task dynamic migration gives the system the ability to dynamically respond in different task scenes; the function on-demand reorganization technology adds another level of adaptability to the decision network, and the system can dynamically reorganize and configure its function modules according to the nature and requirements of the task to provide the best execution efficiency; The operation process of the decision network in the cluster system is as follows: Group perception data collection: first, the perception nodes in the cluster system collect various task-related data, which are sent to the central processing unit of the decision network as input information; Analysis of group perception results: the decision network deeply analyzes the collected group perception data; Task dynamic migration: According to the group perception results, the decision network evaluates the characteristics and requirements of the current task scene, and based on these evaluations, the system decides whether to perform task dynamic migration: if so, the system will adjust resource allocation, optimize network topology or reconfigure task allocation strategy to adapt to the new environment. Function on-demand reorganization: At the same time, the decision network reorganizes the functions on demand according to the nature and requirements of the current task; Optimal solution generation: After dynamic migration and function reorganization, the decision network integrates the results of group perception information, task dynamic migration and function on-demand reorganization to generate a task-oriented optimal decision scheme; Execution and feedback: The generated decision scheme is conveyed to the execution unit of the swarm system, and the execution unit executes the corresponding task according to the scheme, while the system continuously perceives the environment and collects data during the execution process to provide feedback information for the next round of decision-making, forming a closed loop.

8. The four-network converged architecture for unmanned swarm system of claim 1, wherein, In the communication network, the unmanned aerial vehicle communication module is deployed on each intelligent carrier in the cluster, and data transmission is performed through the inter-node ad hoc network, integrating the business units, software radio operation framework, scene perception, calculation, networking communication, and group-side perception, networking and scheduling in the unmanned aerial vehicle group nodes; The entire communication module is divided into two functional modules according to the functional attributes: task interaction and heterogeneous interconnection. The task interaction module is responsible for generating task planning, environmental conditions, task coordination, and task implementation process for unmanned clusters, and through local interaction and long-range association of information, it realizes distributed self-organization of unmanned clusters under communication constraints and forms systematic interoperability of unmanned clusters. The heterogeneous interconnection module is responsible for the interconnection of information between different nodes in the unmanned aerial vehicle ad hoc network. Unmanned aerial vehicles establish an ad hoc network between unmanned clusters through a unified communication module. The operation process of the communication network in the cluster system is as follows: The unmanned cluster starts the routing discovery function of the communication module, adaptively networks according to the link signal-to-noise ratio between different devices, selects nodes with better signal-to-noise ratio as next-hop nodes as much as possible, and outputs the unmanned cluster network topology. The task interaction mechanism and protocol of the unmanned cluster after networking are designed, including the following steps: Step C1: tacit coordination information planning and generation: Tact coordination information is pre-designed and generated by the unmanned cluster task planning system according to the task purpose and requirements, and is converted into task action instructions for the unmanned cluster to execute, promoting the execution of the task. Tacit coordination information interaction refers to the process of sharing and interacting with the strategic guidelines of the unmanned cluster system within the group to achieve coordinated task allocation and coordination. Step C2: generation and processing of contingency handling messages: Contingency handling information refers to the standardized and structured information or algorithms required by the unmanned cluster to handle unexpected situations during task execution according to the situation changes. Contingency handling information interaction refers to the process of forming a group distributed action plan by relying on tacit coordination information and situation selection to drive the unmanned cluster to complete consecutive actions. Step C3: generation and processing of interoperation messages: The interoperability information is a formatted message for realizing network interconnection, information intercommunication and system interoperability between platforms in the unmanned cluster, between unmanned systems and other heterogeneous platforms, thereby realizing the basic functions of situation sharing, command and control, task coordination and intelligence distribution during task execution of the unmanned system; the interactive interoperability message is a distributed action plan formed by the information interaction of the temporary disposition, and the interoperability instruction content of the control unit, sensor unit and equipment unit of the unmanned system platform is formulated, and the interoperability message is transmitted through the unmanned cluster network to realize the coordination of formation, sensing and equipment.

9. The four-network converged architecture for unmanned swarm system of claim 8, wherein, The tacit cooperation information interaction in the step C1 specifically includes the following steps: Step C1.1: unmanned cluster task information planning: On the basis of the analysis of the task target and the decomposition of the task of the typical task, the strategic guideline is planned; Step C1.2: unmanned cluster task process planning: The process of the planned task and the executed task of the unmanned cluster is planned. In order to explore the control generation and evolution of the unmanned cluster, the task jump relationship planning is the planning of the conditions and conversion modes of the conversion between tasks; Step C1.3: unmanned cluster task scheme output: In order to form a loop from the task planning scheme to the schematized message and then to the semi-physical simulation test, the standard format unmanned cluster task scheme is generated; during the planning of the task scheme, the system element nodes automatically associate the schematized message data in the unmanned cluster data element dictionary, and the related data planned is automatically stored in the schematized message case set database; The temporary disposition information generation process in the step C2 specifically includes the following steps: Step C2.1: execution rule construction, including unmanned group rule construction and single body rule construction: The unmanned group rule construction constructs the formation rule and algorithm of the unmanned group, the formation transformation rule and algorithm, the long-lieutenant replacement rule and algorithm, and the detection rule and algorithm; the unmanned single body rule construction constructs the safe flight rule and algorithm of the unmanned single body, the search load use rule and algorithm, and the execution rule; Step C2.2: execution rule management, visual editing management of the execution rule and algorithm, and knowledge customization according to the task demand of the unmanned cluster; Step C2.3: behavior flow editing, adding atomic behaviors to the unmanned cluster single body and group model, and constructing behavior state flow, the behavior editing including atomic behavior, method list and behavior state flow editing, wherein the atomic behavior is an indivisible common behavior action, the method list is a combination of multiple atomic behaviors to form an atomic behavior flow; through the behavior state flow editing, the atomic behavior or combined behavior is set in series, and the trigger condition attribute is set; The interoperability message generation process in the step C3 specifically includes the following steps: Step C3.1: Interoperable message editing, edit the message instructions that unmanned system control unit, sensor unit, device unit may transmit during simulation; when editing the interoperable message, automatically associate the interoperable message data in the data element dictionary to ensure that the message name, data item name, bit position, bit number and other information are consistent with the data element dictionary; Step C3.2: Message compilation, call the interoperable message compilation plug-in to realize the bitization of the information transmitted by the unmanned system control unit, sensor unit and device unit, and perform demonstration in the simulation test environment; the interoperable message editing module edits the interoperable message case content and stores it in the interoperable message case set database. Then, the interoperable message compilation plug-in is called to convert the interoperable message content into bit data stream and demonstrate it; Step C3.3: Message instruction simulation running, demonstrate and verify the interoperable message transmission process through the data link model of the task simulation unit; in the simulation system, the data link model of the unmanned system control unit and sensor unit is called to transmit the interoperable message, and the message transmission content and process are displayed in the visual subsystem.

Images