Method and system for adaptive multi-link unmanned aerial vehicle networking and resource allocation

Abstract

The application provides a multi-link based adaptive unmanned aerial vehicle networking and resource allocation method and system, belongs to the technical field of unmanned aerial vehicle cluster networking and resource allocation, and constructs a three-level unmanned aerial vehicle networking method composed of basic connection, cooperative relay and autonomous intelligence to adapt to dynamic networking requirements in different communication environments; based on the three-level unmanned aerial vehicle networking method, the cooperative control and management of the infrastructure control layer and the unmanned aerial vehicle group control layer are realized; in combination with the networking mode switching mechanism of cross-layer and service index, the adaptability and switching efficiency of the system are enhanced; the resource allocation strategy in the unmanned aerial vehicle task is optimized by using deep reinforcement learning, and efficient cooperation among resources is realized. The application can realize dynamic switching of three networking modes in different communication environments, can intelligently trigger smooth conversion of the networking mode when the link state changes, ensures uninterrupted communication and task, and improves the task execution efficiency and communication stability.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) swarm networking and resource allocation technology, specifically to a multi-link-based adaptive UAV networking and resource allocation method and system. Background Technology

[0002] With the continuous development of fifth-generation mobile communication technology (5G), relying solely on terrestrial base stations is no longer sufficient to meet the coverage requirements of services across a wide area and the demands for highly reliable communication. Therefore, an integrated air-space-ground network architecture has emerged, expanding the traditional two-dimensional network coverage into three-dimensional coverage by integrating satellites, drones, and terrestrial base stations. Compared to traditional terrestrial networks, integrated air-space-ground networks have stronger coverage capabilities in remote and complex terrain areas, better meeting the diverse communication needs of users in different environments. However, even under this integrated architecture, the system may still experience communication unavailability or coverage interruptions due to factors such as natural disasters, extreme weather, network attacks, or insufficient temporary resource allocation. In such cases, the autonomous execution and self-organizing network capabilities of drones become a key guarantee for improving system resilience, disaster recovery capabilities, and mission continuity.

[0003] Current drone networking solutions mainly include two modes: One mode involves drones communicating with 5G base stations, with each drone collaborating with a control center to complete tasks. However, when the channel environment is poor, signal blockage is severe, or the ground network is unavailable, drones may lose contact with the control center, leading to decreased task collaboration efficiency or even task failure. The other mode uses point-to-point communication between multiple drones to achieve distributed collaborative task execution. While this method enhances the system's resilience to disconnection to some extent, it still struggles to efficiently complete tasks in scenarios with high task complexity, intensive data processing, or high precision requirements for collaborative control, limited by the drones' computing resources, communication bandwidth, and energy consumption. Furthermore, when the network environment changes, existing systems often cannot perceive link status in real time, lacking efficient, low-latency multi-link switching capabilities, leading to problems such as long interruption times, data loss, or low retransmission efficiency.

[0004] To address the issues mentioned above, current UAV networking systems still have significant shortcomings in areas such as communication stability assurance, efficient multi-link switching capabilities, and intelligent decision-making mechanisms for mission and communication collaboration. Summary of the Invention

[0005] The purpose of this invention is to provide a multi-link-based adaptive UAV networking and resource allocation method and system with network adaptive capability, link status awareness capability, and intelligent task scheduling capability, so as to solve at least one of the technical problems existing in the background art.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides an adaptive UAV networking and resource allocation method based on multiple links, comprising: A three-tiered UAV networking method consisting of basic connectivity, collaborative relay, and autonomous intelligence is constructed to adapt to dynamic networking requirements under different communication environments; Based on a three-tiered UAV networking method, collaborative control and management between the infrastructure control layer and the UAV swarm control layer are achieved. By combining cross-layer and business indicator networking mode switching mechanisms, the system's adaptability and switching efficiency are enhanced; deep reinforcement learning is used to optimize resource allocation strategies in UAV missions, achieving efficient collaboration among resources.

[0007] As a further limitation of the first aspect of the present invention, the three-layer UAV network, under the collaboration of the infrastructure control layer and the UAV swarm control layer, constructs three networking modes according to communication conditions and task requirements: basic connection networking, cooperative relay networking, and autonomous intelligent networking. In the basic connection networking, when the channel quality is good and the infrastructure capability is strong, the UAV communicates directly with the ground base station through a single hop, and the task is led and directly issued by the infrastructure control layer. In the cooperative relay networking, when the signal is attenuated, there are many neighboring UAVs, and the channel quality is moderate, the infrastructure control layer assists in leading and relaying through point-to-point communication, and the task is indirectly issued. In the autonomous intelligent networking, when the link is unstable, neighboring UAVs are sparse, and the channel quality is poor, relying on the self-organizing network capability of the UAVs, the swarm control layer leads and makes autonomous decisions, and the task is completed collaboratively through point-to-point communication.

[0008] As a further limitation of the first aspect of the present invention, the two-level UAV control architecture diagram includes an infrastructure control layer and a UAV swarm control layer; the infrastructure control layer, as a global control unit, is used for UAV access management, global wireless resource management, task policy distribution and relay node optimization, and has a global perspective and centralized scheduling capability; the UAV swarm control layer, as a local control unit, is responsible for distributed routing, self-organizing network configuration, local wireless resource management and path and task decision-making.

[0009] As a further limitation of the first aspect of the present invention, the switching mechanism between different modes includes: constructing a switching mechanism that combines cross-layer indicators and task indicators to realize the switching mechanism between different networking modes; and determining whether to switch to the basic connection networking based on key indicators such as channel signal-to-noise ratio, wireless transmission rate, UAV remaining energy and task tolerance latency.

[0010] As a further limitation of the first aspect of the present invention, if the current communication quality cannot meet the requirements, it is further determined whether to switch to cooperative relay networking based on the number of neighboring nodes, packet loss rate, link load balancing degree and remaining energy; when the nodes are sparse or the relay cannot guarantee communication, it is then switched to autonomous intelligent networking based on the indicators of the number of neighboring nodes, packet loss rate and remaining energy; according to the priority of different tasks, the threshold of the above indicators is dynamically adjusted, thereby realizing the adaptive switching of cross-network mode under task drive.

[0011] As a further limitation of the first aspect of the present invention, the latency and energy consumption optimization based on deep reinforcement learning includes: establishing a transmission and computation model of tasks in a UAV network, and formally expressing the joint optimization problem of task latency and energy consumption; mapping the optimization problem into a deep reinforcement learning framework: the state space consists of environmental information collected in real time by the UAV network, the action space covers the joint scheduling decision of communication, computation and routing resources, and the reward function is defined by the optimization objective and the penalty term for violating the constraints; through the continuous interaction optimization strategy between the agent and the environment, the synergistic improvement of task execution efficiency and resource utilization efficiency is achieved.

[0012] Secondly, the present invention provides a multi-link-based adaptive UAV networking and resource allocation system, comprising: The module is used to build a three-tiered UAV networking method consisting of basic connectivity, cooperative relay, and autonomous intelligence, to adapt to the dynamic networking requirements under different communication environments; The control module is used to achieve collaborative control and management between the infrastructure control layer and the drone swarm control layer based on the three-level drone networking method. The allocation module is used to enhance the system's adaptability and switching efficiency by combining cross-layer and business indicator networking mode switching mechanisms; it also uses deep reinforcement learning to optimize resource allocation strategies in UAV missions to achieve efficient collaboration between resources.

[0013] Thirdly, the present invention provides a non-transitory computer-readable storage medium for storing computer instructions, which, when executed by a processor, implement the multi-link-based adaptive UAV networking and resource allocation method as described in the first aspect.

[0014] Fourthly, the present invention provides a computer device including a memory and a processor, wherein the processor and the memory communicate with each other, the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute the multi-link-based adaptive UAV networking and resource allocation method as described in the first aspect.

[0015] Fifthly, the present invention provides an electronic device, comprising: a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to cause the electronic device to execute instructions for implementing the multi-link-based adaptive UAV networking and resource allocation method as described in the first aspect.

[0016] The beneficial effects of this invention are as follows: By integrating multiple communication modes such as ground base stations and satellites, a multi-link-based adaptive UAV networking and resource allocation method is proposed. This method can dynamically switch between three networking modes under different communication environments and achieve global and local collaborative optimization by combining a two-layer control architecture. It can intelligently trigger smooth transitions in networking modes when link states change, ensuring uninterrupted communication and missions. At the same time, the resource allocation model constructed using deep reinforcement learning algorithms jointly optimizes communication resources, computing resources, and routing resources based on mission characteristics and network states, significantly improving mission execution efficiency and communication stability.

[0017] The advantages of additional aspects of the invention will be set forth more clearly in the following description or will be learned by practice of the invention. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating the implementation of the multi-link unmanned aerial vehicle (UAV) network system according to an embodiment of the present invention.

[0020] Figure 2 This is a diagram of a three-layer unmanned aerial vehicle (UAV) networking architecture as described in an embodiment of the present invention.

[0021] Figure 3 This is a diagram of a two-level UAV control architecture as described in an embodiment of the present invention.

[0022] Figure 4 This is a flowchart illustrating the switching mechanism between different modes as described in an embodiment of the present invention.

[0023] Figure 5 This is a flowchart illustrating the latency and energy consumption optimization based on deep reinforcement learning as described in an embodiment of the present invention. Detailed Implementation

[0024] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0025] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0026] It should also be understood that terms such as those defined in general dictionaries should be understood to have meanings consistent with their meanings in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless defined as here.

[0027] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this specification means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, and / or groups thereof.

[0028] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0029] To facilitate understanding of the present invention, the present invention will be further explained and described below with reference to the accompanying drawings and specific embodiments. However, the specific embodiments do not constitute a limitation on the embodiments of the present invention.

[0030] Those skilled in the art should understand that the accompanying drawings are merely schematic diagrams of embodiments, and the components in the drawings are not necessarily essential for implementing the present invention.

[0031] Example 1

[0032] In this embodiment 1, an adaptive UAV networking and resource allocation method based on multi-link is provided, which solves the problems of UAV communication interruption, low efficiency of multi-link switching, and lack of coordination between task and network scheduling in complex environments.

[0033] To address the issue of unstable communication, a networking framework integrating three communication modes—terrestrial base stations and satellites—is constructed to enhance network coverage and system robustness. To address the issue of low link switching efficiency, a cross-layer state awareness and rapid switching mechanism is designed to improve communication reliability. Furthermore, by combining reinforcement learning algorithms, collaborative scheduling of task execution and network resources is achieved, ensuring efficient task completion and stable communication operation in complex and dynamic environments.

[0034] This embodiment describes a three-tiered UAV networking method consisting of basic connectivity networking, cooperative relay networking, and autonomous intelligent networking. It includes the functionality of three networking modes. Based on dynamic changes in the network environment, the three networking modes can coordinate and dynamically switch between each other.

[0035] The three networking modes are basic connectivity networking, cooperative relay networking, and autonomous intelligent networking, each with different communication capabilities and mission support characteristics. Basic connectivity networking, under good communication conditions, achieves stable low-latency, high-bandwidth connections by directly accessing ground base stations or satellites via a single hop. Cooperative relay networking, when the quality of direct channels deteriorates, relies on multi-hop relays between UAVs to extend coverage and ensure communication reliability. Autonomous intelligent networking, in harsh or complex environments, supports UAVs in independently completing tasks based on autonomous perception and decision-making capabilities, and synchronizes critical data after communication is restored.

[0036] To facilitate coordination and switching between different networking modes, we consider a dynamic conversion mechanism based on real-time perception of multiple dimensions, including communication link quality, neighbor node reachability, and task requirements. This mechanism can intelligently identify and smoothly switch between basic connection networking, collaborative relay networking, and autonomous intelligent networking. Specifically, when link quality degrades and available drones are available nearby, it automatically switches to collaborative relay networking to ensure communication continuity. If the relay link is unavailable or there are insufficient neighbor nodes, it switches to autonomous intelligent networking, relying on the drone's self-organizing networking capabilities and embodied intelligence to achieve autonomous task execution in network outage environments. This effectively avoids communication interruptions, data loss, and task termination, improving the stability and robustness of drone operations.

[0037] This embodiment describes a management method for collaborative control between the infrastructure control layer and the UAV swarm control layer under different networking modes. The responsibilities of the two-layer control architecture are defined, and interaction and collaboration mechanisms between the two layers are designed for different networking modes.

[0038] The two-layer control architecture consists of an infrastructure control layer and a drone swarm control layer. The infrastructure control layer, deployed at ground base stations or satellites, possesses global awareness and centralized scheduling capabilities, primarily responsible for drone access management, global wireless resource allocation, task policy distribution, and relay node optimization. The drone swarm control layer, deployed on the drone itself, possesses distributed routing, self-organizing network configuration, local resource management, and task decision-making capabilities. It can autonomously construct relay paths based on local network conditions and independently complete task planning and execution, achieving reliable operation and local optimization even in network outage environments.

[0039] The interaction and collaboration mechanism between the two layers supports flexible operation under different networking modes. The two-layer control architecture can operate independently and has strong functional independence. Under good channel conditions, the two parties can achieve global and local collaborative optimization through data interaction and policy synchronization. During the dynamic switching of networking modes, control can be smoothly transferred between global and local modes to ensure uninterrupted communication and tasks, thereby improving the stability and adaptability of the system.

[0040] This embodiment considers a switching method that combines cross-layer metrics and service metrics. This includes the working modes of basic connectivity networking, cooperative relay networking, and autonomous intelligent networking. This method, by combining metrics from the physical layer, link layer, network layer, and the UAV's status and mission, can smoothly switch between the three modes.

[0041] The basic connectivity networking operating mode is determined by comprehensively considering factors such as the signal-to-noise ratio in the channel, the wireless transmission rate, the remaining energy of the UAV, and the mission's tolerable latency to determine whether the UAV is currently in basic connectivity networking mode. This mode focuses on whether the communication quality meets the standards and whether it can support the mission's latency and energy consumption constraints; therefore, it primarily selects basic indicators from the physical layer and the data link layer.

[0042] The described collaborative relay networking mode comprehensively determines whether a drone is currently in collaborative relay networking mode by considering indicators such as the number of neighboring nodes, packet loss rate, link load balancing value, and the drone's remaining energy. This mode focuses on drone collaboration capabilities, network structure stability, and energy consumption balance, with indicators covering the link layer, network layer, and drone status.

[0043] The described autonomous intelligent networking mode comprehensively determines whether a drone is currently in autonomous intelligent networking mode by considering indicators such as the number of neighboring nodes, the drone's remaining energy consumption, and packet loss rate. This mode mainly focuses on the survival and mission execution capabilities of individual drones under extreme conditions, ensuring that even in environments where communication is lost, they can still autonomously complete tasks and guarantee that data is not lost.

[0044] In this embodiment, a method for resource allocation in UAV mission execution using deep reinforcement learning is proposed, including: describing the transmission delay, computation delay, transmission energy consumption, and computation energy consumption models of the task during a single hop of the UAV, and using spectrum resources, UAV computational resources, UAV power, and the selection of the next-hop UAV as decision variables; transforming the task delay and UAV energy consumption optimization problem into a Markov decision process, and describing the design of the state space, action space, and reward in the Markov decision process. The task delay and UAV energy consumption optimization problem first considers the transmission rate of the task in wireless fading and interference channels, thus obtaining the transmission delay and transmission energy consumption; then, the computation delay and computation energy consumption are obtained based on the UAV's computational resources; furthermore, since propagation delay and queuing delay are short, they can be ignored; finally, the optimization objective is proposed as a weighted sum of the task delay and UAV energy consumption, with constraints including the range of values ​​for delay, energy consumption, load balancing, and all decision variables. The state space considers the state information collected from the UAV network, and treats each hop of the task to the UAV as a state. Since the state space is related to the optimization objective, we need to consider direct or indirect factors affecting task latency and UAV energy consumption. These include the signal-to-noise ratio of links between UAVs, the set of available wireless resource blocks, the remaining energy consumption of the UAVs, the set of neighboring nodes of the UAVs, and the remaining latency of the task.

[0045] The aforementioned action space requires resource allocation decisions based on current network status information. Due to the high dynamism and resource constraints of UAV networks, resource management needs to consider multi-dimensional resources: communication resources, computing resources, and routing resources. Specifically, we define the action space to include the selection of relay UAVs, the allocation of wireless resource blocks and UAV computing resources, and power allocation.

[0046] The reward space is mainly divided into two parts: positive reward values ​​and penalty terms. This is because appropriate penalties can help the agent learn the optimal strategy more quickly. Specifically, the positive reward is the weighted sum of task latency and drone energy consumption, while the penalty terms are mainly constraints of the optimization problem, namely latency penalties, energy consumption penalties, and load balancing penalties.

[0047] Example 2

[0048] This embodiment first provides a multi-link-based adaptive UAV networking and resource allocation system, including: a construction module for constructing a three-layer UAV networking method consisting of basic connectivity, cooperative relay, and autonomous intelligence to adapt to dynamic networking requirements under different communication environments; a control module for realizing collaborative control and management of the infrastructure control layer and the UAV swarm control layer based on the three-layer UAV networking method; an allocation module for enhancing the system's adaptability and switching efficiency by combining cross-layer and business indicator networking mode switching mechanisms; and utilizing deep reinforcement learning to optimize resource allocation strategies in UAV missions to achieve efficient collaboration among resources.

[0049] Based on the above system, a multi-link-based adaptive UAV networking and resource allocation method was implemented, including: constructing a three-layer UAV networking method consisting of basic connectivity, cooperative relay, and autonomous intelligence to adapt to dynamic networking requirements under different communication environments; realizing collaborative control and management between the infrastructure control layer and the UAV swarm control layer based on the three-layer UAV networking method; enhancing the system's adaptability and switching efficiency by combining cross-layer and business indicator networking mode switching mechanisms; and optimizing resource allocation strategies in UAV missions using deep reinforcement learning to achieve efficient collaboration among resources.

[0050] like Figure 1 As shown, the multi-link UAV network system is specifically designed as follows: First, a three-tiered UAV networking method consisting of basic connectivity, cooperative relay, and autonomous intelligence is constructed to adapt to dynamic networking requirements under different communication environments. Second, collaborative control and management of the infrastructure control layer and the UAV swarm control layer are implemented to improve global and local scheduling capabilities. Third, a networking mode switching mechanism combining cross-layer and business indicators is designed to enhance the system's adaptability and switching efficiency. Finally, deep reinforcement learning is used to optimize resource allocation strategies in UAV missions to achieve efficient collaboration among resources.

[0051] like Figure 2As shown, the three-tiered UAV networking architecture demonstrates three networking modes built according to communication conditions and mission requirements, under the collaboration of the infrastructure control layer and the UAV swarm control layer: basic connectivity networking, cooperative relay networking, and autonomous intelligent networking. In basic connectivity networking, when channel quality is excellent and infrastructure capabilities are strong, UAVs communicate directly with ground base stations via single-hop communication, with missions directly issued and controlled by the infrastructure control layer. In cooperative relay networking, when signal attenuation occurs, there are many neighboring UAVs, and channel quality is moderate, the infrastructure control layer assists in relaying and forwarding via point-to-point communication, achieving indirect mission issuance. In autonomous intelligent networking, when the link is unstable, neighboring UAVs are sparse, and channel quality is poor, relying on the self-organizing network capabilities of UAVs, the swarm control layer leads and makes autonomous decisions, collaboratively completing missions through point-to-point communication. The diagram also clearly indicates the switching conditions and master-slave control relationships between different networking modes.

[0052] like Figure 3 As shown, the two-tiered UAV control architecture comprises an infrastructure control layer and a UAV swarm control layer. Specifically, the infrastructure control layer, as the global control unit, is responsible for UAV access management, global wireless resource management, task policy distribution, and relay node optimization, possessing a global perspective and centralized scheduling capabilities. The UAV swarm control layer, as a local control unit, undertakes responsibilities such as distributed routing, self-organizing network configuration, local wireless resource management, and path and task decision-making, exhibiting strong adaptability and autonomy. The two control layers are functionally relatively independent, yet achieve information sharing and task coordination through collaborative interaction.

[0053] like Figure 4 As shown, the switching mechanism between different modes is as follows: First, a switching mechanism combining cross-layer indicators and task indicators is constructed to achieve switching between different network modes. Then, based on key indicators such as channel signal-to-noise ratio, wireless transmission rate, remaining UAV energy, and task-tolerable latency, it is determined whether to switch to the basic connectivity network. If the current communication quality cannot meet the requirements, it is further determined whether to switch to the cooperative relay network based on the number of neighboring nodes, packet loss rate, link load balancing, and remaining energy. When nodes are sparse or relays cannot guarantee communication, it switches to the autonomous intelligent network based on the number of neighboring nodes, packet loss rate, and remaining energy. Finally, according to the priority of different tasks, the thresholds of the above indicators are dynamically adjusted to achieve task-driven adaptive switching between cross-network modes.

[0054] like Figure 5As shown, this process for latency and energy consumption optimization based on deep reinforcement learning first establishes a transmission and computation model for tasks in an unmanned aerial vehicle (UAV) network, and formalizes the joint optimization problem of task latency and energy consumption. Then, this optimization problem is mapped onto a deep reinforcement learning framework: the state space consists of environmental information collected in real-time by the UAV network, the action space encompasses the joint scheduling decisions of communication, computation, and routing resources, and the reward function is defined by the optimization objective (a weighted sum of latency and energy consumption) and a penalty term for violating constraints. Finally, through continuous interaction between the agent and the environment, the strategy is gradually optimized to achieve a synergistic improvement in task execution efficiency and resource utilization efficiency.

[0055] Example 3

[0056] The UAV networking and control system provided in this embodiment can adapt to dynamic communication environments and diverse mission requirements. It can be applied to scenarios where UAVs perform missions in complex and harsh environments, such as... Figure 2 As shown, the UAV accesses the networking system according to mission requirements and communication environment conditions. Then, through the cross-layer and mission indicator combined networking mode switching mechanism proposed in this embodiment, and with the help of a pre-trained deep reinforcement learning model, it can quickly determine and switch to the optimal networking mode based on the latency requirements, remaining energy, channel quality, and other characteristics of the current mission. This achieves optimal synergy between communication performance and energy efficiency, and improves the overall mission completion efficiency and system resource utilization.

[0057] Specifically, by designing a three-tiered networking architecture that includes basic connectivity, collaborative relay, and autonomous intelligence, the system possesses flexible network organization capabilities; by implementing collaborative management of the infrastructure control layer and the UAV swarm control layer, global scheduling and local response are achieved; at the same time, an intelligent switching mechanism combining communication status and task requirements is introduced; finally, by using deep reinforcement learning methods to jointly optimize resource scheduling, efficient collaboration between communication and computing is realized.

[0058] To address the challenges of communication and mission execution for unmanned aerial vehicles (UAVs) in complex, dynamic, and extreme environments, this implementation proposes a multi-level adaptive network framework, consisting of three networking modes, a dynamic transition mechanism between modes, and a two-layer control architecture. For complex communication environments, the framework designs three hierarchical networking modes: when the communication environment is good, a basic connection network is used, directly connecting to ground base stations or satellites via a single hop to achieve stable communication with low latency and high bandwidth; when the direct connection channel weakens and other UAVs are nearby, it switches to a cooperative relay network, utilizing multi-hop relay links to extend coverage and ensure reliable data and command transmission; in extremely harsh environments, when both direct connections and relay paths are unavailable, it switches to an autonomous intelligent network, relying on the UAV's local perception and decision-making capabilities to achieve autonomous mission execution and data synchronization, ensuring the continuity and integrity of missions under network outage conditions.

[0059] To ensure flexible switching between the three networking modes in dynamic environments, the system further introduces a dynamic conversion mechanism between modes. This mechanism can intelligently identify and adaptively switch to the optimal networking mode based on real-time perception and comprehensive evaluation of multi-dimensional factors such as communication link quality, neighbor node reachability, and task requirements. When link quality degrades or communication signal attenuates, and there are available UAV nodes nearby, the system automatically switches to cooperative relay networking, achieving indirect connection with ground base stations or satellites through multi-hop relay links to ensure communication continuity and stability. If the relay link is unavailable, or the number of nearby UAV nodes is insufficient or the link is unstable, it further converts to autonomous intelligent networking mode. At this time, the system relies on the self-organizing networking capability of the UAV itself to build a local point-to-point communication link, and combines embodied intelligence to realize autonomous planning and execution of tasks, ensuring that tasks can still be independently advanced and key data can be retained even in a completely disconnected environment, providing support for data synchronization and task collaboration after subsequent communication is restored.

[0060] To effectively support the aforementioned networking modes, the designed network control architecture consists of an infrastructure control layer and a drone swarm control layer. Each layer possesses independent control and decision-making capabilities, and they collaborate under different networking modes to ensure the continuity and optimization of the drone network and missions. The infrastructure control layer, deployed at ground base stations or satellites, possesses global awareness and centralized scheduling capabilities. It is primarily responsible for drone access management, global wireless resource allocation, dynamically issuing mission strategies based on drone location information and network status, and assisting in selecting the optimal relay node and path in the cooperative relay networking mode. This control layer plays a leading role in the basic connectivity networking mode and provides global assistance in cooperative relay networking. The drone swarm control layer, deployed on the drone itself, plays a leading role, particularly in cooperative relay networking and autonomous intelligent networking modes. Its core functions include dynamically constructing multi-hop relay links and self-organizing network configurations based on neighboring nodes and network conditions, flexibly managing local wireless resources to reduce interference and improve local network efficiency, and autonomously completing path planning and mission decisions. Especially in the autonomous intelligent networking mode, it completely relies on local control to independently execute missions. The two-layer control architecture can independently perform control and management in their respective network modes, without relying on each other. When channel conditions are favorable, global and local collaborative optimization is achieved through data interaction and policy synchronization. The infrastructure control layer is responsible for global scheduling and policy formulation, while the UAV swarm control layer executes specific operations according to the global policy and feeds back local status and execution results, achieving two-layer coordination. More importantly, during the dynamic transformation of network modes, control can be smoothly transferred between the global (infrastructure control layer) and local (UAV swarm control layer) modes, ensuring that UAVs can seamlessly take over and switch control when link quality degrades, fails, or recovers, avoiding communication interruptions, mission abortions, or UAV loss of control.

[0061] In the detailed UAV network architecture described above, a detailed handover process needs to be formulated. To achieve more intelligent, task-aware, and resource-efficient network handover, the system integrates cross-layer communication indicators and service requirements to comprehensively perceive environmental conditions and task characteristics, thereby improving the accuracy and adaptability of handover decisions. Furthermore, a handover mechanism based on joint cross-layer and service perception is constructed. Specifically, Phase 1 is the basic connection network setup. This phase focuses on whether the communication quality meets standards and whether it can support task latency and energy consumption constraints. Therefore, basic indicators at the physical and link layers are selected to ensure that direct communication is the preferred solution. The signal-to-noise ratio of the wireless channels between UAVs is selected. wireless transmission rate The remaining energy of the drone and the tolerance latency of the task Therefore, we can obtain the judgment criteria as shown in the following formula:

[0062] in It is the threshold of the signal-to-noise ratio. It is the minimum wireless transmission rate required by the task. This is the energy consumption threshold for stage 1. This represents the maximum latency required by the task. It's important to note that the above metric values ​​are determined by the specific task. Different priorities of tasks will have different metric values.

[0063] Phase 2 involves collaborative relay networking. This phase focuses on the drone's collaborative capabilities, network stability, and energy balance. Therefore, covering the link layer, network layer, and drone status is the preferred approach. We select the number of neighbor nodes for each drone. Link load balancing value Packet loss rate And the remaining energy of the drone. Therefore, we can obtain the determination criteria for Phase 2, as shown in the following formula:

[0064] in It is the minimum number of neighbors required. It is the threshold for load balancing of the link. This represents the highest packet loss rate during this phase of the task. It is the energy consumption threshold for stage 2.

[0065] Phase 3 is autonomous intelligent networking. The main focus of this phase is on the survivability and operational capabilities of individual drones, ensuring they can independently complete tasks and prevent data loss even under extreme conditions. Therefore, selecting network layer, drone status, and task indicators is the preferred approach. The selected indicators are the number of neighboring nodes of the drone, the drone's remaining energy, and the packet loss rate. Thus, we can obtain the criteria for Phase 3, as shown in the following formula:

[0066] in This is the energy consumption threshold for stage 3. This represents the maximum packet loss rate for this phase of the mission. When performing a mission, the drone first attempts to establish a basic network connection. If the above criteria are not met, it switches to a system relay network. If the indicator value continues to decrease, it will proactively switch to an autonomous network phase. Simultaneously, continuous sensing and dynamic reverse switching occur at each phase; if communication recovers, it gradually returns to cooperative relay or direct connection.

[0067] To ensure the stability, real-time performance, and resource utilization efficiency of UAVs during mission execution, it is necessary to rationally allocate critical resources such as communication, computing, and routing. UAVs often face challenges such as limited computing power, limited energy, and unstable communication links. Without an effective resource scheduling mechanism, mission interruptions, performance degradation, or excessive energy consumption can easily occur. Furthermore, different missions have different resource requirements; some missions emphasize low-latency transmission, while others require high computational load processing. Dynamically allocating resources by combining mission type, environmental conditions, and resource availability can effectively improve system adaptability and overall operational efficiency. We consider a set of UAVs as... The set of tasks is and the set of wireless resource blocks as According to Shannon's theorem, when the task... From drones Transmit to drone At that time, its transmission rate can be expressed as:

[0068] in, It is an allocated wireless resource block. It is the bandwidth of a single radio resource block. It is a drone The transmission power, It is a drone To drones Channel gain between It is the power spectral density of the noise. This is interference power. If the task... The data packet size is Then it comes from Transmit to The required transmission delay is:

[0069] In drones Execute the task The computation delay is:

[0070] in This represents the number of CPU cycles required per bit of data. For drones Allocated computing resources (CPU frequency). Accordingly, the task... from Transmit to The transmission energy consumption is:

[0071] And in drones The calculated energy consumption is:

[0072] in It is the effective capacitance coefficient, which depends on the drone's CPU chip architecture. Therefore, the mission... The one-hop latency and energy consumption after one communication and computation process are as follows:

[0073] Since propagation delay and queuing delay are short, they can be ignored. Furthermore, considering that drone hovering delay is typically short, this method ignores this type of delay. To improve the performance efficiency and energy sustainability of the drone network during task execution, it is necessary to jointly optimize the total task delay and drone energy consumption. This is because task completion delay directly relates to the degree to which its timeliness requirements are met, while energy consumption determines the drone's endurance and task continuity. Based on the established transmission and computation model, the delay and energy consumption generated during task execution are comprehensively affected by multiple factors, including spectrum resource allocation, transmit power control, computational resource scheduling, and routing selection (i.e., determining the next-hop drone). In scenarios with limited resources and dynamically changing communication environments, jointly optimizing the above key decision variables helps to achieve efficient collaborative scheduling of communication and computational resources, thereby ensuring task timeliness while reducing energy consumption and improving the overall system's resource utilization and task processing performance. Therefore, the optimization problem can be defined as: , The resource allocation decision value constraints are as follows: , , , , , , , in, It is an energy consumption weighting factor. It is a drone Current load balancing index, It is the threshold of the load balancing index. It is a drone Maximum transmission power It is a drone The maximum allocated CPU frequency, It is a task The selected next-hop drone It is a drone The set of neighbors.

[0074] Because the multiple decision variables in the optimization problem (such as spectrum allocation, power control, computational resource scheduling, and routing) are complexly coupled with each other, and because the system needs to make decisions in real time in a highly dynamic and unpredictable environment (such as channel state, task arrival, and neighbor changes), traditional deterministic modeling and static optimization methods are difficult to effectively handle. Therefore, reinforcement learning, as a data-driven, long-term reward-oriented sequential decision-making method, possesses strong adaptability and generalization capabilities, making it very suitable for solving such complex dynamic optimization problems. To this end, the resource scheduling and task distribution problem is modeled as a Markov decision process. By defining reasonable state space, action space, state transition probabilities, and reward functions, the agent can continuously learn the optimal policy through interaction with the environment, thereby achieving efficient resource management and continuous optimization of task performance.

[0075] In the state space, to accurately characterize the dynamic features of the UAV network and provide sufficient information input for optimizing resource allocation strategies, key factors that directly or indirectly affect task latency and UAV energy consumption need to be considered. Specifically, the state space should include the following: First, the signal-to-noise ratio of the link. This reflects the quality of the current communication link, directly determining the mission's transmission rate and energy consumption; secondly, it reflects the set of wireless resource blocks currently available to the UAV. This determines the amount of spectrum resources it can allocate; secondly, the drone's remaining energy consumption This indicates its power status, limiting the scale of computational and forwarding tasks it can handle; furthermore, the set of neighboring nodes... This reflects the local network topology of the UAV, affecting its routing decisions and relay capabilities; finally, the remaining latency of the mission. This describes the real-time requirements for completion time, which can be used to dynamically adjust resource allocation priorities. Through joint modeling of the above multi-dimensional information, the state space can comprehensively reflect the current resource status and task urgency of the system, thereby providing a decision-making basis for reinforcement learning agents and improving the timeliness and resource adaptability of strategies. The description of the state space is shown in the following formula:

[0076] In the action space, to achieve fine-grained control over the task transmission and computation process, key resource management decisions during each hop of task forwarding must be incorporated. Specifically, the action space should cover the following decisions: First, the selection of relay drones. Firstly, the allocation of radio resource blocks directly determines the task transmission path and has a significant impact on link quality, hop count, overall latency, and energy consumption. This determines the available bandwidth resources during each hop of transmission, affecting the instantaneous transmission rate and channel occupancy efficiency of the task; thirdly, the allocation of computing resources This involves allocating appropriate CPU resources to the selected UAV nodes to meet the task's computing needs and control computational latency and energy consumption; fourthly, power allocation. Therefore, under the premise of meeting communication quality and energy consumption constraints, the drone's transmission power needs to be dynamically adjusted to achieve a trade-off between communication efficiency and energy consumption. In summary, the action space, by jointly modeling various key resource decision variables involved in task forwarding and processing, provides a high-dimensional and tightly coupled optimization space for reinforcement learning agents. The action space is described by the following formula:

[0077] In the reward design, the weighted sum of the total task latency and UAV energy consumption is used as the single-step reward to measure the overall performance of the current resource allocation and routing decisions. Considering that relying solely on latency and energy consumption may lead to the strategy converging to a local optimum, failing to fully meet system constraints and long-term performance requirements, a penalty mechanism is introduced into the reward. The reward is described by the following formula:

[0078] in , , , It is a weighting factor. For a binary indicator function, when When the inequality within the parentheses is true, If it is 0, otherwise, The value is -1. Since the range of values ​​for the decision variable can be directly restricted in reinforcement learning, it is excluded from the penalty term.

[0079] Consider using a swarm of drones to collect environmental data such as communication status, energy consumption levels, and mission information, and transmit this data back to the computing server on the infrastructure side via online or offline methods. The reinforcement learning model is then trained offline on the server. After training, the compressed and optimized lightweight inference model is deployed to the drones themselves, enabling efficient online decision-making and thus improving the system's intelligent scheduling capabilities and autonomous operation level.

[0080] Example 4

[0081] This embodiment 4 provides a non-transitory computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, they implement the multi-link-based adaptive UAV networking and resource allocation method described above. The method includes: A three-tiered UAV networking method consisting of basic connectivity, collaborative relay, and autonomous intelligence is constructed to adapt to dynamic networking requirements under different communication environments; Based on a three-tiered UAV networking method, collaborative control and management between the infrastructure control layer and the UAV swarm control layer are achieved. By combining cross-layer and business indicator networking mode switching mechanisms, the system's adaptability and switching efficiency are enhanced; deep reinforcement learning is used to optimize resource allocation strategies in UAV missions, achieving efficient collaboration among resources.

[0082] Example 5

[0083] This embodiment 5 provides a computer device, including a memory and a processor. The processor and the memory communicate with each other. The memory stores program instructions that can be executed by the processor. The processor calls the program instructions to execute the multi-link-based adaptive UAV networking and resource allocation method described above. The method includes: A three-tiered UAV networking method consisting of basic connectivity, collaborative relay, and autonomous intelligence is constructed to adapt to dynamic networking requirements under different communication environments; Based on a three-tiered UAV networking method, collaborative control and management between the infrastructure control layer and the UAV swarm control layer are achieved. By combining cross-layer and business indicator networking mode switching mechanisms, the system's adaptability and switching efficiency are enhanced; deep reinforcement learning is used to optimize resource allocation strategies in UAV missions, achieving efficient collaboration among resources.

[0084] Example 6

[0085] This embodiment 6 provides an electronic device, including: a processor, a memory, and a computer program; wherein, the processor is connected to the memory, and the computer program is stored in the memory. When the electronic device is running, the processor executes the computer program stored in the memory to cause the electronic device to execute instructions implementing the multi-link-based adaptive UAV networking and resource allocation method described above. The method includes: A three-tiered UAV networking method consisting of basic connectivity, collaborative relay, and autonomous intelligence is constructed to adapt to dynamic networking requirements under different communication environments; Based on a three-tiered UAV networking method, collaborative control and management between the infrastructure control layer and the UAV swarm control layer are achieved. By combining cross-layer and business indicator networking mode switching mechanisms, the system's adaptability and switching efficiency are enhanced; deep reinforcement learning is used to optimize resource allocation strategies in UAV missions, achieving efficient collaboration among resources.

[0086] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0087] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0088] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0089] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment, whereby a series of operational steps are performed to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0090] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that, based on the technical solutions disclosed in the present invention, various modifications or variations that can be made by those skilled in the art without creative effort should be included within the scope of protection of the present invention.

Claims

1. A multi-link-based adaptive UAV networking and resource allocation method, characterized in that, include: A three-tiered UAV networking method is constructed, consisting of basic connectivity, cooperative relay, and autonomous intelligence, to adapt to dynamic networking requirements under different communication environments. This three-tiered UAV networking, under the collaboration of the infrastructure control layer and the UAV swarm control layer, comprises three networking modes based on communication conditions and mission requirements: basic connectivity networking, cooperative relay networking, and autonomous intelligence networking. In basic connectivity networking, when channel quality is excellent and infrastructure capabilities are strong, UAVs communicate directly with ground base stations via single-hop communication, with missions directly issued and led by the infrastructure control layer. In cooperative relay networking, when signal attenuation occurs, there are many neighboring UAVs, and channel quality is moderate, the infrastructure control layer assists in relaying and forwarding through point-to-point communication, achieving indirect mission issuance. In autonomous intelligence networking, when links are unstable, neighboring UAVs are sparse, and channel quality is poor, relying on the self-organizing capabilities of UAVs, the UAV swarm control layer leads and makes autonomous decisions, collaboratively completing missions through point-to-point communication. Based on a three-tiered UAV networking method, collaborative control and management between the infrastructure control layer and the UAV swarm control layer are achieved. By combining cross-layer and business indicator-based network mode switching mechanisms, the system's adaptability and switching efficiency are enhanced. Deep reinforcement learning is used to optimize resource allocation strategies in UAV missions, achieving efficient collaboration among resources. Specifically, latency and energy consumption optimization based on deep reinforcement learning includes: establishing a transmission and computation model for tasks in the UAV network and formalizing the joint optimization problem of task latency and energy consumption; mapping this optimization problem into a deep reinforcement learning framework: the state space consists of environmental information collected in real-time by the UAV network, the action space encompasses the joint scheduling decisions of communication, computation, and routing resources, and the reward function is defined by the optimization objective and the penalty term for violating constraints; through continuous interaction between the agent and the environment, optimization strategies are optimized to achieve a synergistic improvement in task execution efficiency and resource utilization efficiency.

2. The adaptive UAV networking and resource allocation method based on multi-links according to claim 1, characterized in that, The two-tiered UAV control architecture includes an infrastructure control layer and a UAV swarm control layer. The infrastructure control layer, as a global control unit, is used for UAV access management, global wireless resource management, task policy distribution, and relay node optimization, and has a global perspective and centralized scheduling capabilities. The UAV swarm control layer, as a local control unit, is responsible for distributed routing, self-organizing network configuration, local wireless resource management, and path and task decision-making.

3. The adaptive UAV networking and resource allocation method based on multi-links according to claim 1, characterized in that, The switching mechanism between different modes includes: constructing a switching mechanism that combines cross-layer indicators and task indicators to realize the switching mechanism between different networking modes; determining whether to switch to the basic connection network based on key indicators such as channel signal-to-noise ratio, wireless transmission rate, UAV remaining energy, and task tolerance latency; if the current communication quality cannot meet the requirements, further determining whether to switch to the cooperative relay network based on the number of neighboring nodes, packet loss rate, link load balancing, and remaining energy; when nodes are sparse or relays cannot guarantee communication, switching to the autonomous intelligent network based on the number of neighboring nodes, packet loss rate, and remaining energy indicators; and dynamically adjusting the thresholds of the above indicators according to the priority of different tasks, thereby realizing task-driven adaptive switching between cross-network modes.

4. A multi-link-based adaptive UAV networking and resource allocation system, characterized in that, include: The module is used to construct a three-tiered UAV networking method consisting of basic connectivity, cooperative relay, and autonomous intelligence to adapt to dynamic networking requirements under different communication environments. The three-tiered UAV networking, under the collaboration of the infrastructure control layer and the UAV swarm control layer, constructs three networking modes according to communication conditions and task requirements: basic connectivity networking, cooperative relay networking, and autonomous intelligence networking. In basic connectivity networking, when channel quality is excellent and infrastructure capabilities are strong, UAVs communicate directly with ground base stations via single-hop communication, and tasks are directly issued by the infrastructure control layer. In cooperative relay networking, when signal attenuation occurs, there are many neighboring UAVs, and channel quality is moderate, the infrastructure control layer assists in relaying through point-to-point communication, and tasks are indirectly issued. In autonomous intelligence networking, when the link is unstable, neighboring UAVs are sparse, and channel quality is poor, relying on the self-organizing capabilities of UAVs, the UAV swarm control layer leads and makes autonomous decisions, completing tasks collaboratively through point-to-point communication. The control module is used to achieve collaborative control and management between the infrastructure control layer and the drone swarm control layer based on the three-level drone networking method. The allocation module is used to enhance the system's adaptability and switching efficiency by combining cross-layer and business indicator network mode switching mechanisms. It utilizes deep reinforcement learning to optimize resource allocation strategies in UAV missions, achieving efficient collaboration among resources. Specifically, the latency and energy consumption optimization based on deep reinforcement learning includes: establishing a transmission and computation model for tasks in the UAV network and formalizing the joint optimization problem of task latency and energy consumption; mapping this optimization problem into a deep reinforcement learning framework: the state space consists of environmental information collected in real-time by the UAV network, the action space covers the joint scheduling decisions of communication, computation, and routing resources, and the reward function is defined by the optimization objective and the penalty term for violating constraints; through continuous interaction between the agent and the environment, optimization strategies are optimized to achieve a synergistic improvement in task execution efficiency and resource utilization efficiency.

5. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium is used to store computer instructions, which, when executed by a processor, implement the multi-link-based adaptive UAV networking and resource allocation method as described in any one of claims 1-3.

6. A computer device, characterized in that, The system includes a memory and a processor, which communicate with each other. The memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute the multi-link-based adaptive UAV networking and resource allocation method as described in any one of claims 1-3.

7. An electronic device, characterized in that, include: The device includes a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to cause the electronic device to execute instructions that implement the multi-link-based adaptive UAV networking and resource allocation method as described in any one of claims 1-3.

Images