Resource orchestration method and device for collaborative network, equipment, storage medium and program product

Abstract

The invention relates to a resource orchestration method and device for a collaborative network, equipment, a storage medium and a program product. Comprising the following steps: constructing an optimization model taking maximization of total satisfaction, minimization of service duration and minimization of energy consumption of an unmanned aerial vehicle as targets; the total satisfaction is the sum of the satisfaction of each target area, the service duration is the sum of the sensing duration, the transmission duration, the calculation time and the flight time corresponding to each target area, and the energy consumption of the unmanned aerial vehicle is the sum of the sensing energy consumption, the transmission energy consumption, the flight energy consumption and the hovering energy consumption corresponding to each target area; solving the optimization model under the constraint of a preset constraint condition to obtain a computing resource allocation strategy and a sensing communication strategy; comprise computing time, sensing duration, service sequences, flight heights and transmission power of various computing resources. By adopting the method, multi-target collaborative optimization is realized, namely, the service duration and the energy consumption of the unmanned aerial vehicle are minimized while the total satisfaction degree of the system is maximized.

Description

Technical Field

[0001] This application relates to the field of wireless communication technology, and in particular to a resource orchestration method, apparatus, computer device, computer-readable storage medium, and computer program product for a cooperative network. Background Technology

[0002] With the integration of next-generation mobile wireless communication systems and artificial intelligence technologies, emerging applications such as real-time monitoring, smart agriculture, and intelligent transportation are rapidly appearing. These applications generally require the processing of massive amounts of environmental data and demand that the system possess low-latency, high-precision real-time analysis and decision-making capabilities.

[0003] To meet these needs, the industry has proposed the Integrated Sensing, Communication, and Computing (ISCC) technical framework, which aims to provide high-precision sensing and wireless information transmission services by optimizing resource allocation strategies. However, existing ISCC technical solutions mainly focus on terrestrial systems, which face two major challenges in practical applications: First, ground obstacles may block sensing and communication signals, affecting service quality; second, in remote areas or complex terrain, the coverage of ground infrastructure is limited, making it difficult to provide stable and reliable services.

[0004] To overcome the limitations of ground-based systems, unmanned aerial vehicle (UAV)-enabled aerospace networks are considered a promising solution. In this architecture, UAVs serve as the airborne access layer, providing flexible wireless access and edge intelligence services, while aerostated platforms serve as the space access layer, offering ubiquitous communication services and powerful edge computing capabilities.

[0005] Although UAV aerospace networks bring new possibilities to ISCC services, traditional technologies still have the following significant shortcomings in resource management and task scheduling: (1) Single optimization objective: Current ISCC systems in UAV networks mostly focus on optimizing a single performance indicator, such as system throughput, energy efficiency, or sensing accuracy. This single-objective optimization strategy is difficult to meet the comprehensive requirements of system performance in diverse and dynamic application scenarios. (2) Lack of differentiated task scheduling: Traditional resource scheduling methods fail to fully consider the inherent differences between different tasks, such as different priorities and heterogeneous resource requirements. This leads to the inability of the system to achieve efficient and fair task execution in a multi-task environment. (3) Inefficient optimization algorithms: Traditional optimization methods (such as heuristic or convex optimization methods) usually have problems such as high computational complexity, sensitivity to initial values, and difficulty in obtaining the global optimal solution. These limitations make it difficult to adapt to the highly dynamic and uncertain environment of UAV aerospace networks, because such environments require fast and efficient global resource optimization and task scheduling. Summary of the Invention

[0006] Based on this, it is necessary to provide a resource orchestration method, apparatus, computer equipment, computer-readable storage medium, and computer program product for cooperative networks to address the aforementioned technical problems. This method can achieve multi-objective collaborative optimization, namely, maximizing overall system satisfaction while minimizing service duration and drone energy consumption.

[0007] Firstly, this application provides a resource orchestration method for cooperative networks, comprising:

[0008] Based on the continuous perception model, the satisfaction level of each target area is determined according to the perception duration of each target area, and the perception energy consumption of the UAV is determined according to the satisfaction level of each target area.

[0009] Based on the communication model, the signal-to-interference-plus-noise ratio (SIR / NNR) of each transmission stage is determined according to the UAV's flight altitude and transmission power, and the transmission duration and UAV's transmission energy consumption are determined according to the SIR / NNR.

[0010] Based on the collaborative computing model, the computation time of various computing resources is determined;

[0011] Based on the drone trajectory model, the flight time of the drone is determined according to the service sequence and the drone's flight altitude. The hovering time of the drone is determined according to the sensing duration and transmission duration. The flight energy consumption of the drone is determined according to the flight time. The hovering energy consumption of the drone is determined according to the hovering time.

[0012] An optimization model is constructed with the objectives of maximizing total satisfaction, minimizing service duration, and minimizing drone energy consumption. Total satisfaction is the sum of satisfaction in each target area, service duration is the sum of perception duration, transmission duration, computation time, and flight time for each target area, and drone energy consumption is the sum of perception energy consumption, transmission energy consumption, flight energy consumption, and hovering energy consumption for each target area.

[0013] Under the constraints of the preset conditions, the optimization model is solved to obtain the computing resource allocation strategy and the sensing communication strategy. The computing resource allocation strategy includes the computing time of various computing resources, and the sensing communication strategy includes the sensing duration, service sequence, flight altitude and transmission power.

[0014] In one embodiment, under preset constraints, the optimization model is solved to obtain a computing resource allocation strategy and a sensing communication strategy, including:

[0015] The optimization model is broken down into a computational resource allocation sub-problem and a perception communication sub-problem;

[0016] Under the constraints of the preset conditions, the subproblem of computing resource allocation is solved to obtain the computing resource allocation strategy;

[0017] Under the constraints of the preset conditions, the perceptual communication subproblem is solved to obtain the perceptual communication strategy.

[0018] In one embodiment, under preset constraints, the computational resource allocation subproblem is solved to obtain a computational resource allocation strategy, including:

[0019] Based on the coordinate descent iterative algorithm, each variable in the computational resource allocation subproblem is optimized sequentially according to the load balancing strategy to obtain the computational resource allocation strategy.

[0020] In one embodiment, under preset constraints, the sensing communication subproblem is solved to obtain a sensing communication strategy, including:

[0021] Acquire the state space data of the intelligent agent; the state space data includes the amount of sensing data, channel gain, UAV location information, and computing resource allocation strategy.

[0022] Using the policy network corresponding to the perception and communication subproblem, the action space data of the agent is determined based on the state space data; the action space data includes the perception and communication policy.

[0023] In one embodiment, after determining the agent's action space data based on the state space data using the policy network corresponding to the perception communication subproblem, the method further includes:

[0024] The reward function value of the agent is determined based on the computing resource allocation strategy and the perception and communication strategy;

[0025] State space data, action space data, and reward function values ​​are stored as empirical data.

[0026] Based on empirical data, the parameters of the agent's policy network and evaluation network are updated using an LSTM-enhanced proximal policy optimization algorithm.

[0027] In one embodiment, before constructing an optimization model aimed at maximizing overall satisfaction, minimizing service duration, and minimizing drone energy consumption, the method further includes:

[0028] The sensing duration, transmission duration, computation time, and flight time corresponding to each target area are normalized, and the sensing energy consumption, transmission energy consumption, flight energy consumption, and hovering energy consumption corresponding to each target area are also normalized.

[0029] Secondly, this application also provides a resource orchestration apparatus for a cooperative network, comprising:

[0030] The perception module is used to determine the satisfaction level of each target area based on the perception duration of each target area, and to determine the perception energy consumption of the UAV based on the satisfaction level of each target area, based on the continuous perception model.

[0031] The communication module is used to determine the signal-to-interference-plus-noise ratio (SIR / NNR) for each transmission stage based on the communication model, the UAV's flight altitude and transmission power, and the transmission duration and UAV's transmission energy consumption based on the SIR / NNR.

[0032] The computing module is used to determine the computing time of various computing resources based on the collaborative computing model;

[0033] The trajectory planning module is used to determine the flight time of the UAV based on the UAV trajectory model, the service sequence, and the UAV's flight altitude; to determine the hovering time of the UAV based on the sensing duration and transmission duration; to determine the flight energy consumption of the UAV based on the flight time; and to determine the hovering energy consumption of the UAV based on the hovering time.

[0034] The optimization module is used to construct an optimization model with the objectives of maximizing overall satisfaction, minimizing service duration, and minimizing UAV energy consumption. Overall satisfaction is the sum of satisfaction levels for each target area; service duration is the sum of perception duration, transmission duration, computation time, and flight time for each target area; and UAV energy consumption is the sum of perception energy consumption, transmission energy consumption, flight energy consumption, and hovering energy consumption for each target area. Under preset constraints, the optimization model is solved to obtain a computational resource allocation strategy and a perception communication strategy. The computational resource allocation strategy includes the computation time of various computational resources, and the perception communication strategy includes perception duration, service sequence, flight altitude, and transmission power.

[0035] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described in the first aspect above.

[0036] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in the first aspect above.

[0037] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the method described in the first aspect above.

[0038] The resource orchestration method, apparatus, computer equipment, computer-readable storage medium, and computer program product of the aforementioned cooperative network, based on a continuous sensing model, determine the satisfaction level of each target area according to the sensing duration of each target area, and determine the sensing energy consumption of the UAV based on the satisfaction level of each target area; based on a communication model, determine the signal-to-interference-plus-noise ratio (SIR / NNR) of each transmission stage according to the UAV's flight altitude and transmission power, and determine the transmission duration and UAV's transmission energy consumption based on the SIR / NNR; based on a cooperative computing model, determine the computing time of various computing resources; based on a UAV trajectory model, determine the UAV's flight time according to the service sequence and the UAV's flight altitude, determine the UAV's hovering time according to the sensing duration and transmission duration, and determine the UAV's flight time based on the flight time. The flight energy consumption of the UAV is determined based on its hovering time. An optimization model is constructed with the objectives of maximizing overall satisfaction, minimizing service duration, and minimizing UAV energy consumption. Overall satisfaction is the sum of satisfaction levels across all target areas; service duration is the sum of perception duration, transmission duration, computation time, and flight time for each target area; and UAV energy consumption is the sum of perception energy consumption, transmission energy consumption, flight energy consumption, and hovering energy consumption for each target area. Under preset constraints, the optimization model is solved to obtain a computational resource allocation strategy and a perception-communication strategy. The computational resource allocation strategy includes the computation time of various computational resources, and the perception-communication strategy includes perception duration, service sequence, flight altitude, and transmission power. Through this approach, multi-objective collaborative optimization is achieved, that is, maximizing overall system satisfaction while minimizing service duration and UAV energy consumption. To achieve optimal global performance, multi-dimensional and heterogeneous resources spanning multiple stages such as perception, communication, computation, and flight are jointly optimized. These resources include the UAV's perception duration, transmission power, flight altitude, service sequence, and the computational resource allocation strategy of the aerostat platform. It provides an intelligent optimization framework with real-time decision-making capabilities to improve the adaptability and overall operational efficiency of the ISCC system in complex real-world scenarios. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is a diagram illustrating the application environment of a resource orchestration method for a cooperative network in one embodiment.

[0041] Figure 2 This is a flowchart illustrating a resource orchestration method for a cooperative network in one embodiment.

[0042] Figure 3 This is a schematic diagram of the algorithm framework in one embodiment;

[0043] Figure 4 This is a structural block diagram of a resource orchestration device for a cooperative network in one embodiment;

[0044] Figure 5 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0046] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0047] To facilitate understanding of the embodiments provided in this application, some terms that appear in the detailed description are introduced below:

[0048] Collaborative Sensing, Communication, and Computing (CSCC): This is a core framework referring to a technological paradigm that deeply integrates and collaboratively optimizes the three elements of sensing, communication, and computing. In the embodiments of this application, it specifically refers to an aerospace network system composed of ground-based sensing devices, unmanned aerial vehicles (UAVs), and hovering platforms (HAPs). The UAVs collaborate with ground-based devices to perform sensing tasks, while the UAVs act as communication relays, transmitting data to the hovering platform for computational processing.

[0049] Unmanned Aerial Vehicle (UAV): Refers to an unmanned aerial vehicle. In the embodiments of this application, the UAV plays a dual role: as an aerial entity, it collaborates with ground equipment to perform perception tasks; and as an aerial access layer and data relay, it is responsible for connecting ground equipment and the airborne platform.

[0050] High altitude platform (HAP): Operating in the stratosphere, it acts as a space access layer and space edge server, providing ubiquitous communication services and edge computing capabilities for data collected by drones and ground equipment.

[0051] Terrestrial Sensing Devices (TSDs): Sensor devices deployed in target areas on the ground. They are responsible for performing environmental perception tasks, collecting detailed ground information, and collaborating with UAVs to achieve overall target perception.

[0052] Space-Air Networks: These refer to a comprehensive network architecture consisting of a space layer (such as aerostated platforms) and an air layer (such as unmanned aerial vehicle networks). This type of network offers advantages such as flexible deployment, seamless coverage, and intelligent computing services, making it an ideal solution for supporting CSCC applications.

[0053] Multi-objective optimization: A mathematical optimization problem that requires simultaneously optimizing multiple, potentially conflicting, objectives. The objective of this application's embodiments is to simultaneously achieve three objectives while satisfying a series of constraints: maximizing overall perception satisfaction, minimizing total service duration, and minimizing the energy consumption of the UAV.

[0054] Deep Reinforcement Learning (DRL) is an artificial intelligence technique that combines deep learning and reinforcement learning. It is used as the core intelligent engine that drives systems to perceive, make decisions, and execute, in order to solve complex resource scheduling and joint optimization problems.

[0055] Proximal Policy Optimization (PPO): An advanced deep reinforcement learning algorithm. This application's embodiments employ a PPO-based DRL algorithm combined with an LSTM module to effectively solve the sub-problem of joint perception-communication optimization.

[0056] Coordinate Descent (CD): An iterative optimization algorithm that solves a problem by fixing other variables and optimizing only one variable in each step. This application utilizes this algorithm to solve the subproblem of HAP edge computing resource allocation, minimizing computation time.

[0057] Sensing Satisfaction: A performance metric introduced in this application embodiment to evaluate the quality of the sensing task. It maps the amount of collected data to a non-linear index using a logistic function to reflect the phenomenon of diminishing marginal returns, thereby avoiding waste of system resources.

[0058] The resource orchestration method for cooperative networks provided in this application can be applied to, for example... Figure 1 The application environment shown is as follows. The cooperative sensing communication computing system includes multiple ground sensing devices (TSDs), a UAV, and a hovering platform (HAP). The transmission links include ground-to-air links and air-to-space links. The resource orchestration method for cooperative networks provided in this application embodiment is applied to the hovering platform. The hovering platform is denoted as... As a space server, it can provide edge computing services for UAVs and TSDs. (The term "drone" is a colloquial expression meaning "unmanned aerial vehicle" or "tethered drone"). In addition to performing sensing tasks, the ground sensing devices also need to provide data relay services for the TSD. The set of ground sensing devices is denoted as... Each TSD is responsible for performing environmental perception and acquiring perception data. The area where the TSD is located serves as the target area for UAV perception. The set of target areas is denoted as . Introducing binary variables This indicates the priority of the target area. and These represent high-priority and low-priority regions, respectively. Consider a three-dimensional Cartesian coordinate system, where the UAV... and ground sensing equipment The positions are described as follows: and The drone was located at The system departs from the mission center, executes the scheduled operation, and returns to the deployment origin.

[0059] In one exemplary embodiment, such as Figure 2 As shown, a resource orchestration method for cooperative networks is provided, which can be applied to... Figure 1 Taking the floating platform in the middle as an example, the following is an explanation:

[0060] Step 202: Based on the continuous perception model, determine the satisfaction level of each target area according to the perception duration of each target area, and determine the perception energy consumption of the UAV according to the satisfaction level of each target area.

[0061] In this process, the perception processes of ground-based sensing devices and drones are pre-modeled to obtain a continuous perception model. In specific implementations, drones can perceive targets over a wider range, while ground-based sensing devices can capture more detailed information about the perceived targets. Based on this, this application's embodiments employ a dual-sensing strategy based on both air and ground. Sensing time is shared between the ground-based sensing devices and drones to acquire target areas. The sensing data. From ground-based sensing devices. or drone The obtained The raw sensor data for each target region can be calculated as follows: ,in ; These represent ground sensing devices. or drone The perceived data rate, measured in bits per second, assumes that the drone has superior perception capabilities, i.e. ; This refers to the duration of perception. (Drone) The perceived energy consumption can be obtained as follows: ,in It is the energy required to sense one bit of data, measured in joules.

[0062] To ensure perceived quality while simultaneously improving energy and resource efficiency, a satisfaction level based on the logistic function is introduced, which can be described as:

[0063] ;

[0064] Where X and z are the satisfaction saturation coefficient and steepness coefficient, respectively. It is the perception critical point, representing the moment when the rate of increase in satisfaction shifts from acceleration to deceleration. , These represent the minimum and maximum amount of perceived data, respectively. The overall satisfaction across all target regions can be expressed as: Based on the above model, the amount of perceived data can be transformed from a linear to a nonlinear model.

[0065] This approach captures the phenomenon of diminishing marginal returns in the real world, avoiding waste of system resources. Furthermore, by adjusting parameters related to satisfaction, the adaptability of the CSCC service in complex, heterogeneous, multi-task scenarios can be significantly improved. The boundedness of the satisfaction function provides a stable and predictable reward signal for the subsequently proposed DRL-based algorithm, effectively mitigating training instability caused by unbounded reward growth.

[0066] Step 204: Based on the communication model, determine the signal-to-interference-plus-noise ratio (SIR / NDR) for each transmission stage according to the UAV's flight altitude and transmission power, and determine the transmission duration and UAV's transmission energy consumption based on the SIR / NDR.

[0067] Among them, reference Figure 1 The ground-to-air and air-to-space links are pre-built with ground-to-air and air-to-space transmission models. In the specific implementation, the ground-to-air transmission model can be described as follows:

[0068] To improve spectral efficiency, the UAV operates in full-duplex (FD) mode, simultaneously receiving data from ground-based sensing devices and transmitting the acquired sensing data to an aerostat platform for further processing. Assuming there are no obstacles between the UAV and the ground-based sensing devices, the channel is primarily dominated by line-of-sight (LoS) signals. The channel power gain can be expressed as: ,in It is the channel power gain per unit distance, and Ground sensing equipment With drones The distance between them. Ground sensing equipment. To drones The signal-to-interference-plus-noise ratio (SINR) can be expressed as:

[0069] ;

[0070] in Ground sensing equipment The transmit power (i.e., the transmission power). It is the ratio of residual self-interference to power. Indicates drone The transmit power (i.e., the transmission power). Ground sensing equipment With drones The noise power between them. Then the ground sensing equipment. To drones The reachable uplink data rate is: ,in This refers to system bandwidth. Ground sensing equipment. To drones The transmission delay is:

[0071] .

[0072] In practical implementation, the air-to-space transmission model can be described as follows:

[0073] drones With floating platform The channel gain between them can be expressed as:

[0074] ;

[0075] in and They represent drones Transmit antenna gain and floating platform The receiving antenna gain, For wavelength, and Indicates drone With floating platform The distance between them.

[0076] Phase 1: In-band full-duplex relay phase. (UAV) To the floating platform SINR can be expressed as:

[0077] ;

[0078] in It is a drone With floating platform Noise power between. Achievable uplink data rate: .

[0079] In Phase 1, the ground-based sensing device-UAV link and the UAV-aerial platform link share time and frequency resources. Transmission to the aerial platform... The amount of data is .

[0080] At the end of Phase 1, the drone The total amount of remaining data is: .

[0081] Phase 2: Exclusive Spectrum Simplex Phase. In this phase, self-interference from the drone is eliminated. Drone To the floating platform SINR is represented as:

[0082] .

[0083] The corresponding achievable speed and transmission delay are as follows:

[0084] ;

[0085] ;

[0086] drones The transmission energy consumption is: ,in This represents the total transmission time.

[0087] Step 206: Based on the collaborative computing model, determine the computing time for various computing resources.

[0088] It is understandable that various computing resources can include CPUs (Central Processing Units) and GPUs (Graphics Processing Units). Once the floating platform... Upon receiving sensor data from the drone, it will provide computing services based on CPU-GPU collaborative parallel computing. Considering heterogeneous computing tasks, tuples are introduced. Describes heterogeneous tasks from ground-based sensing devices and drones, in which This indicates the number of floating-point operations (in FLOPs) required to compute the task. This indicates the proportion of data offloaded to the GPU. and These represent the resource allocation ratios for the GPU and CPU, respectively. and The processing efficiency of the GPU and CPU reflects the inherent preferences of different hardware for various tasks. In the embodiments of this application, the CPU is better suited for handling tasks related to ground sensing devices, while the GPU performs exceptionally well in handling tasks related to drones. and Ground sensing devices processed by GPUs and drones The task computation time can be given by the following formulas:

[0089] ;

[0090] .

[0091] Similarly, ground sensing devices processed by CPUs and drones The task computation time can be calculated separately as follows:

[0092] ;

[0093] .

[0094] in and These represent the computing power of the GPU and CPU (unit: FLOPS), respectively. The final computation time can be expressed as: .

[0095] Step 208: Based on the UAV trajectory model, determine the UAV's flight time according to the service sequence and the UAV's flight altitude, determine the UAV's hovering time according to the sensing duration and transmission duration, determine the UAV's flight energy consumption according to the flight time, and determine the UAV's hovering energy consumption according to the hovering time.

[0096] Information about high-priority targets needs to be detected and transmitted in a timely manner. Therefore, drones need to prioritize access to these targets. The service sequence is defined as follows:

[0097] ;

[0098] Since each target region is accessed exactly once, the length of the service sequence is equal to the number of target regions. Under the premise of prioritizing access to high-priority areas, the following constraints must be met:

[0099] ;

[0100] This application supports differentiated services based on task priority, enabling differentiated service scheduling to address the different time-sensitivity requirements of various tasks. By prioritizing service areas and constraining UAVs to access high-priority targets first, it ensures that critical tasks can be processed in a timely manner.

[0101] For any given target area of ​​drone service Its operation is divided into two phases: flight and hovering. Assuming that when a drone... When its hovering position projection coincides with the position of the ground sensing device, it begins data sensing and interacts with the ground sensing device. and floating platform A transmission link is established. Therefore, in the proposed application scenario, the trajectory optimization of the UAV can be modeled as a constrained traveling salesman problem. The UAV's trajectory is then uniquely determined by two key parameters: the sequence of visits to the target area. and the flight altitude of each time slot This modeling approach can decouple spatiotemporal optimization variables into discrete sequential decisions and continuous height control parameters.

[0102] Specifically, drones The flight time can be calculated as follows:

[0103] ;

[0104] in This is the average flight speed of the drone. The drone's hovering time equals the sum of its sensing time and transmission time, that is: Drones The energy consumption for flight and hovering can be: and ,in and These represent the drone's flight power and hovering power, respectively.

[0105] Step 210: Construct an optimization model with the objectives of maximizing total satisfaction, minimizing service duration, and minimizing UAV energy consumption. Total satisfaction is the sum of satisfaction in each target area. Service duration is the sum of perception duration, transmission duration, computation time, and flight time for each target area. UAV energy consumption is the sum of perception energy consumption, transmission energy consumption, flight energy consumption, and hovering energy consumption for each target area.

[0106] Wherein, the service duration of target area m is This includes flight, sensing, transmission, and computation, and is given by the following formula: The total service duration for all target regions can be calculated as follows: Drone service target area The energy consumption is This includes flight, hovering, sensing, and transmission, and is given by the following formula: Since the limited energy of the drones is the primary limiting factor for the entire CSCC service duration, only the energy consumption of the drones will be considered. The total energy consumed by the drones to serve all target areas is... To ensure the quality of perception tasks, enhance the real-time responsiveness of CSCC services, and extend the lifespan of UAVs, thereby improving overall system efficiency and resource utilization, the goal of the optimization model is to maximize total satisfaction. Minimize service duration And the energy consumption of drones Perception duration Access sequence Flight altitude drones Transmission power Calculate the unloading ratio and the resource allocation ratio of GPU and CPU. and They will be jointly optimized. This multi-objective optimization problem can be modeled as follows:

[0107] P0: .

[0108] Step 212: Under the constraints of the preset constraints, solve the optimization model to obtain the computing resource allocation strategy and the sensing communication strategy. The computing resource allocation strategy includes the computing time of various computing resources, and the sensing communication strategy includes the sensing duration, service sequence, flight altitude and transmission power.

[0109] The preset conditions are as follows:

[0110] C1: ;

[0111] C2: ;

[0112] C3: ;

[0113] C4: ;

[0114] C5: ;

[0115] C6: ;

[0116] C7: ;

[0117] C8: .

[0118] Wherein, C1 is the perception duration constraint, C2-C3 represent the transmission power and flight altitude constraints of the UAV respectively, C4-C6 represent the computation ratio and resource constraints, C7 is the energy consumption constraint, and C8 represents the sequence constraint of the UAV's service target area.

[0119] Optionally, a preset solution strategy is used to solve the optimization model, outputting the computation time, sensing duration, service sequence, flight altitude, and transmission power of various computing resources. For example, the optimal combination solution can be obtained directly using a deep reinforcement learning algorithm.

[0120] In the resource orchestration method of the aforementioned cooperative network, based on the continuous sensing model, the satisfaction level of each target area is determined according to the sensing duration of each target area, and the sensing energy consumption of the UAV is determined according to the satisfaction level of each target area; based on the communication model, the signal-to-interference-plus-noise ratio (SIRR) of each transmission stage is determined according to the UAV's flight altitude and transmission power, and the transmission duration and UAV transmission energy consumption are determined according to the SIRR; based on the cooperative computing model, the computing time of various computing resources is determined; based on the UAV trajectory model, the flight time of the UAV is determined according to the service sequence and the UAV's flight altitude, the hovering time of the UAV is determined according to the sensing duration and transmission duration, the flight energy consumption of the UAV is determined according to the flight time, and the hovering time is determined according to the hovering time. The hovering energy consumption of the UAV is determined. An optimization model is constructed with the objectives of maximizing overall satisfaction, minimizing service duration, and minimizing UAV energy consumption. Overall satisfaction is the sum of satisfaction levels across all target areas. Service duration is the sum of perception duration, transmission duration, computation time, and flight time for each target area. UAV energy consumption is the sum of perception energy consumption, transmission energy consumption, flight energy consumption, and hovering energy consumption for each target area. Under preset constraints, the optimization model is solved to obtain the computational resource allocation strategy and the perception communication strategy. The computational resource allocation strategy includes the computation time of various computational resources, and the perception communication strategy includes perception duration, service sequence, flight altitude, and transmission power. Through this approach, multi-objective collaborative optimization is achieved, that is, maximizing overall system satisfaction while minimizing service duration and UAV energy consumption. To achieve optimal global performance, multi-dimensional and heterogeneous resources spanning multiple stages such as perception, communication, computation, and flight are jointly optimized. These resources include the UAV's perception duration, transmission power, flight altitude, service sequence, and the computational resource allocation strategy of the aerostat platform. It provides an intelligent optimization framework with real-time decision-making capabilities to improve the adaptability and overall operational efficiency of the ISCC system in complex real-world scenarios.

[0121] In an exemplary embodiment, before step 210, the method further includes: normalizing the sensing duration, transmission duration, calculation time and flight time corresponding to each target area, and normalizing the sensing energy consumption, transmission energy consumption, flight energy consumption and hovering energy consumption corresponding to each target area.

[0122] Among them, regarding service duration and energy consumption The magnitude differences among these subprocesses can lead to the optimization process being dominated by the larger numerical target when directly summing. This problem is particularly prominent in learning-based optimization methods, which may neglect smaller but critical targets. To mitigate the impact of the scale differences between different task stages on the optimization objective, Z-score normalization is applied to the execution latency and energy consumption of each stage. The normalized form of the service duration is expressed as: .in:

[0123] ;

[0124] , , and These represent the average values ​​of flight time, sensing duration, transmission duration, and computation time, respectively. , , and This represents their corresponding standard deviations. Similarly, the standardized form of energy consumption is given by the following formula: .in:

[0125] ;

[0126] , , and These represent the average values ​​of the drone's flight energy consumption, sensing energy consumption, transmission energy consumption, and hovering energy consumption, respectively. , , and This represents their corresponding standard deviations. After the above normalization process, the total service duration and total energy consumption of the drone can be rewritten as: and The original problem P0 can be reconstructed as:

[0127] P1: .

[0128] In an exemplary embodiment, step 212 includes: decomposing the optimization model into a computational resource allocation subproblem and a perception communication subproblem; solving the computational resource allocation subproblem under preset constraints to obtain a computational resource allocation strategy; and solving the perception communication subproblem under preset constraints to obtain a perception communication strategy.

[0129] Understandably, for the original problem P0 or the reconstruction problem P1, the optimal combined solution can be obtained directly using deep reinforcement learning algorithms. However, the huge action space makes convergence difficult for the DRL algorithm. To address this issue, the multi-objective optimization problem is decomposed into two sub-problems: a computational resource allocation sub-problem and a perceptual communication sub-problem.

[0130] Understandably, the optimization of edge computing on floating platforms aims to minimize computation time by optimizing the offloading ratio and the allocation of GPU and CPU computing resources. The computational resource allocation subproblem is modeled as follows:

[0131] P1.1: It satisfies C4-C6.

[0132] The optimal computation time is obtained by solving the computational resource allocation subproblem P1.1 of the floating platform. The results are then substituted into the overall problem P1. This decomposition simplifies the originally complex problem, allowing subsequent focus to be placed on the remaining variables. The second subproblem P1.2 addresses the optimization of perception and communication strategies, including the perception duration of the UAV. Service sequence Flight altitude and transmission power The perceptual communication subproblem can be formulated as follows:

[0133] P1.2: It satisfies C1-C3, C7-C8.

[0134] In an exemplary embodiment, under the constraints of preset conditions, the computational resource allocation subproblem is solved to obtain a computational resource allocation strategy, including: an iterative algorithm based on coordinate descent, which optimizes each variable in the computational resource allocation subproblem in turn according to a load balancing strategy to obtain the computational resource allocation strategy.

[0135] Problem P1.1 is a high-dimensional nonlinear optimization problem with a max operation in the objective function, making direct solution challenging. Coordinate descent (CD) is an effective iterative method that optimizes only one variable per iteration while keeping others fixed, thus gradually approaching the global optimum. In applying the CD method, each variable is optimized sequentially by fixing the other three variables at each step and adjusting the current variable to minimize the objective function. , , and For the min-max problem, according to the load balancing principle, when The optimal solution is reached at this time. Based on this condition, the update formula can be further derived. The solution process includes the following:

[0136] 1. Optimization First fix , and The one-dimensional optimization problem that needs to be solved is:

[0137] P1.1.1: ;

[0138] Satisfying C4-C6, C9: ;

[0139] Given that only and Depends on ,make ,Right now:

[0140] ;

[0141] optimal It can be obtained in the following ways:

[0142] ;

[0143] optimal It can be easily deduced that:

[0144] .

[0145] 2. Optimization :fixed , and The one-dimensional optimization problem is represented as:

[0146] P1.1.2: ;

[0147] Satisfying C4-C6, C10: ;

[0148] Because only and Depends on ,make ,Right now:

[0149] ;

[0150] Then the optimal It can be represented as:

[0151] ;

[0152] optimal It can be calculated as:

[0153] .

[0154] 3. Optimization :fixed , and This leads to a one-dimensional optimization problem:

[0155] P1.1.3: It meets C4-C6 requirements;

[0156] Because only and Depends on ,make ,Right now:

[0157] ;

[0158] Then the optimal It can be represented as:

[0159] ;

[0160] optimal It can be calculated as:

[0161] .

[0162] 4. Optimization :fixed , and This leads to a one-dimensional optimization problem:

[0163] P1.1.4: Satisfies C4-C6, C9-C10;

[0164] Because only and Depends on ,make ,Right now:

[0165] ;

[0166] Then the optimal It can be represented as:

[0167] ;

[0168] optimal It can be calculated as: .

[0169] Based on the above approach, the original four-dimensional optimization problem P1.1 is decomposed into a series of one-dimensional subproblems. According to the load balancing principle, each one-dimensional subproblem can be solved efficiently. An iterative algorithm based on the coordinate descent (CD) algorithm is designed to determine the optimal computation time. The detailed implementation process is shown in Algorithm 1 (LORA: Load Balancing Offloading and Resource Allocation Based on Coordinate Descent):

[0170] Input: The amount of sensing data from ground-based sensing devices and drones. and

[0171] Output:

[0172] 1. Initialize tolerance error initial value

[0173] 2. When Time loop:

[0174] 3.

[0175] 4.

[0176] 5.

[0177] 6. End the loop

[0178] In an exemplary embodiment, under the constraints of preset conditions, the perception communication subproblem is solved to obtain the perception communication strategy, including: acquiring the state space data of the agent; the state space data includes the amount of perception data, channel gain, UAV position information and computing resource allocation strategy; using the policy network corresponding to the perception communication subproblem, the action space data of the agent is determined based on the state space data; the action space data includes the perception communication strategy.

[0179] Subproblem P1.2 is characterized by sequential decision-making in a dynamic environment and is solved using Deep Reinforcement Learning (DRL). First, this subproblem is modeled as a Markov Decision Process (MDP). Within this framework, the agent can learn the optimal policy through interaction with the environment. The core elements of the MDP are mapped from the optimization variables and objective in P1.2 to the corresponding state space, action space, and reward function. The specific definitions are as follows:

[0180] The state space includes: the amount of sensing data, the channel gain from the ground sensing device to the UAV, the channel gain from the UAV to the aerostat, the UAV's position, and the computation time of the aerostat. It can be represented as: .

[0181] Action space: Actions represent the behavior and decisions of an agent. The action space includes the drone's perception duration, service sequence, flight altitude, and transmission power, and can be represented as: .

[0182] Reward Function: In MDP, the reward function is used to evaluate the impact of an agent's actions on the environment given a state. The reward is defined as: .in , , These are weighting coefficients used to balance the goals of sensing performance, latency, and energy consumption. A discount factor is also taken into account. The proposed optimization objective can be expressed as:

[0183] .

[0184] Understandably, in order to avoid the curse of dimensionality and high computational complexity of traditional optimization algorithms, this application aims to propose an efficient intelligent algorithm framework based on deep reinforcement learning (DRL). By decomposing complex problems and solving them using algorithms such as proximal policy optimization (PPO), this framework can adapt to the highly dynamic and uncertain environment of aerospace networks and achieve fast and efficient resource decision-making.

[0185] In an exemplary embodiment, after determining the agent's action space data based on the state space data using the policy network corresponding to the perception and communication subproblem, the method further includes: determining the agent's reward function value based on the computational resource allocation policy and the perception and communication policy; storing the state space data, action space data, and reward function value as empirical data; and updating the parameters of the agent's policy network and evaluation network using an LSTM-enhanced proximal policy optimization algorithm based on the empirical data.

[0186] To determine the optimal CSCC strategy, a centralized, model-free deep reinforcement learning algorithm based on PPO is proposed. Due to the heterogeneity of perception capabilities between ground-based sensing devices and UAVs, the amount of sensing data dynamically changes; simultaneously, differences in priority between different target areas lead to changes in the UAV's flight trajectory, altitude, and channel state information. These factors introduce uncertainty in state transitions, thereby reducing the learning efficiency of the DRL algorithm. To address this issue, this application introduces an LSTM-based state representation layer and embeds it into the PPO algorithm to capture long-term dependencies in uncertain sequential states.

[0187] Understandably, LSTM includes a memory unit to retain information from previous input sequences and is equipped with a gating mechanism to regulate the flow of information between input, output, and memory. LSTM adjusts the utilization of new information and uses it to process the state input sequence. Hidden State The following composite function can be used to update:

[0188] .

[0189] in, and These are the weight matrix and the bias matrix, respectively. It's the sigmoid function. The algorithm framework adopts an Actor-Critic structure, consisting of three deep neural networks: two policy networks. and (The parameters are respectively) and ), and a parameter of The Critic network. The loss function and corresponding gradient obtained from policy execution are as follows: ,in:

[0190] ;

[0191] This represents the probability ratio between two policy networks. It is an estimate of the advantage function, which can be expressed as: ,in and These are the state-action value function and the state value function, respectively.

[0192] Furthermore, to balance variance and bias, the Generalized Advantage Estimator (GAE) is introduced: ,in, The weighting factors represent variance and bias, and Policy Network The loss function is defined as: .

[0193] In the specific implementation, the following Algorithm 2 (LSTM-enhanced PPO for cooperative CSCC optimization) is used for solving subproblems and updating the model:

[0194] 1. Initialize policy parameters and value function parameters

[0195] 2. for episode=1,2,3...do

[0196] 3. Reset the LSTM hidden states and cell states of the Actor and Critic networks.

[0197] 4. form=1 to M do

[0198] 5. Perform an action

[0199] 6. Calculate rewards Observe the new state

[0200] 7. Based on the current value function Calculate the advantage estimate

[0201] 8. Update the strategy by maximizing the PPO-Clip objective.

[0202] 9. Update hidden status

[0203] 10. Collect trajectory sets

[0204] 11. Fit the value function by regressing the mean squared error.

[0205] 12. Collect trajectory sets

[0206] 13. end for

[0207] 14.endfor

[0208] In one exemplary embodiment, such as Figure 3 As shown, based on the proposed algorithm framework, the steps for applying this method to solve specific communication networks are as follows:

[0209] 1. Problem Restructuring and Data Standardization (Warm Start Phase). Before solving, the problem is restructured to obtain problem P1. Z-score normalization is applied to the execution latency (flight, sensing, transmission, computation) and energy consumption (flight, sensing, transmission, hovering) of each task phase.

[0210] 2. Problem Decomposition. The normalized original optimization problem P0 is decomposed into two subproblems: P1.1 the computational resource allocation subproblem at the edge of the floating platform and P1.2 the perception and communication subproblem (perception, communication, and trajectory strategy).

[0211] 3. Solve subproblem P1.1 based on coordinate descent:

[0212] 1) Solve using an iterative algorithm based on coordinate descent (CD) (refer to Algorithm 1 above).

[0213] 2) Based on the principle of load balancing, this algorithm calculates the unloading ratio through iterative optimization. as well as GPU / CPU resource allocation ratio as well as .

[0214] 3) Output or calculate the optimal computation time. .

[0215] 4. Solving subproblem P1.2 based on DRL:

[0216] 1) Take the result from the previous step Substituting the result into P1 simplifies the problem to P1.2.

[0217] 2) Model P1.2 as a Markov Decision Process (MDP) and define its state, action, and reward functions.

[0218] 3) Use LSTM to enhance the PPO algorithm (refer to Algorithm 2 above) to train a PPO agent.

[0219] 4) The agent learns through interaction with the environment and ultimately determines the optimal perception-communication strategy, i.e., the duration of the drone's perception. Service sequence Flight altitude and transmission power .

[0220] 5. Obtain the final strategy. Combine the results of steps 3 and 4 to obtain the optimal solution for the computational resource allocation strategy and the perception-communication strategy.

[0221] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated in the embodiments of this application, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0222] Based on the same inventive concept, this application also provides a resource orchestration apparatus for implementing the resource orchestration method of the cooperative network described above. The solution provided by this apparatus is similar to the implementation described in the above method. Therefore, the specific limitations of one or more resource orchestration apparatus embodiments of cooperative networks provided below can be found in the limitations of the resource orchestration method of cooperative networks described above, and will not be repeated here.

[0223] In one exemplary embodiment, such as Figure 4 As shown, a resource orchestration device for a cooperative network is provided, comprising:

[0224] The perception module 402 is used to determine the satisfaction level of each target area based on the perception duration of each target area according to the continuous perception model, and to determine the perception energy consumption of the UAV based on the satisfaction level of each target area.

[0225] The communication module 404 is used to determine the signal-to-interference-plus-noise ratio (SIR / NNR) of each transmission stage based on the communication model, the flight altitude and transmission power of the UAV, and the transmission duration and transmission energy consumption of the UAV based on the SIR / NNR.

[0226] The computing module 406 is used to determine the computing time of various computing resources based on the collaborative computing model.

[0227] The trajectory planning module 408 is used to determine the flight time of the UAV based on the UAV trajectory model, the service sequence and the flight altitude of the UAV, the hovering time of the UAV based on the sensing duration and the transmission duration, the flight energy consumption of the UAV based on the flight time, and the hovering energy consumption of the UAV based on the hovering time.

[0228] The optimization module 410 is used to construct an optimization model with the objectives of maximizing overall satisfaction, minimizing service duration, and minimizing UAV energy consumption. Overall satisfaction is the sum of satisfaction levels for each target area; service duration is the sum of perception duration, transmission duration, computation time, and flight time for each target area; and UAV energy consumption is the sum of perception energy consumption, transmission energy consumption, flight energy consumption, and hovering energy consumption for each target area. Under preset constraints, the optimization model is solved to obtain a computational resource allocation strategy and a perception communication strategy. The computational resource allocation strategy includes the computation time of various computational resources, and the perception communication strategy includes perception duration, service sequence, flight altitude, and transmission power.

[0229] The resource orchestration device of the aforementioned collaborative network achieves multi-objective collaborative optimization, that is, maximizing overall system satisfaction while minimizing service duration and UAV energy consumption. To achieve optimal global performance, multi-dimensional and heterogeneous resources spanning multiple stages such as sensing, communication, computing, and flight are jointly optimized. These resources include the UAV's sensing duration, transmission power, flight altitude, service sequence, and the computing resource allocation strategy of the aerostat platform. An intelligent optimization framework with real-time decision-making capabilities is provided to improve the adaptability and overall operational efficiency of the ISCC system in complex real-world scenarios.

[0230] In an exemplary embodiment, the optimization module 410 is further configured to decompose the optimization model into a computational resource allocation subproblem and a perception communication subproblem; under the constraints of preset constraints, solve the computational resource allocation subproblem to obtain a computational resource allocation strategy; and under the constraints of preset constraints, solve the perception communication subproblem to obtain a perception communication strategy.

[0231] In an exemplary embodiment, the optimization module 410 is further configured to optimize each variable in the computational resource allocation subproblem sequentially according to the load balancing strategy based on the coordinate descent iterative algorithm, thereby obtaining the computational resource allocation strategy.

[0232] In an exemplary embodiment, the optimization module 410 is further configured to acquire state space data of the agent; the state space data includes the amount of sensing data, channel gain, UAV location information and computing resource allocation strategy; and using the policy network corresponding to the sensing communication subproblem, determine the action space data of the agent based on the state space data; the action space data includes the sensing communication strategy.

[0233] In an exemplary embodiment, the resource orchestration apparatus for the cooperative network further includes an update module, which is used to determine the reward function value of the agent according to the computing resource allocation strategy and the perception communication strategy; store the state space data, action space data and reward function value as empirical data; and update the parameters of the agent's policy network and evaluation network according to the empirical data through an LSTM-enhanced proximal policy optimization algorithm.

[0234] In an exemplary embodiment, the resource orchestration device of the cooperative network further includes a preprocessing module, which is used to normalize the sensing duration, transmission duration, computing time and flight time corresponding to each target area, and to normalize the sensing energy consumption, transmission energy consumption, flight energy consumption and hovering energy consumption corresponding to each target area.

[0235] Each module in the resource orchestration device of the aforementioned collaborative network can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0236] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 5 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores empirical data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When executed by the processor, the computer program implements a resource orchestration method for a cooperative network.

[0237] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0238] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0239] In one exemplary embodiment, a computer-readable storage medium is also provided, on which a computer program is stored, which, when executed by a processor, implements the steps in the above-described method embodiments.

[0240] In one exemplary embodiment, a computer program product is also provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0241] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0242] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0243] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0244] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A resource orchestration method for a cooperative network, characterized in that, The method includes: Based on the continuous perception model, the satisfaction level of each target area is determined according to the perception duration of each target area, and the perception energy consumption of the UAV is determined according to the satisfaction level of each target area. Based on the communication model, the signal-to-interference-plus-noise ratio (SIR / NNR) of each transmission stage is determined according to the UAV's flight altitude and transmission power, and the transmission duration and UAV's transmission energy consumption are determined according to the SIR / NNR. Based on the collaborative computing model, the computation time of various computing resources is determined; Based on the drone trajectory model, the flight time of the drone is determined according to the service sequence and the drone's flight altitude. The hovering time of the drone is determined according to the sensing duration and transmission duration. The flight energy consumption of the drone is determined according to the flight time. The hovering energy consumption of the drone is determined according to the hovering time. An optimization model is constructed with the objectives of maximizing total satisfaction, minimizing service duration, and minimizing drone energy consumption. Total satisfaction is the sum of satisfaction in each target area, service duration is the sum of perception duration, transmission duration, computation time, and flight time for each target area, and drone energy consumption is the sum of perception energy consumption, transmission energy consumption, flight energy consumption, and hovering energy consumption for each target area. Under the constraints of the preset conditions, the optimization model is solved to obtain the computing resource allocation strategy and the sensing communication strategy. The computing resource allocation strategy includes the computing time of various computing resources, and the sensing communication strategy includes the sensing duration, service sequence, flight altitude and transmission power.

2. The method according to claim 1, characterized in that, The process of solving the optimization model under preset constraints to obtain a computing resource allocation strategy and a sensing communication strategy includes: The optimization model is broken down into a computational resource allocation sub-problem and a perception communication sub-problem; Under the constraints of preset conditions, the subproblem of computing resource allocation is solved to obtain a computing resource allocation strategy; Under the constraints of preset conditions, the perception communication subproblem is solved to obtain the perception communication strategy.

3. The method according to claim 2, characterized in that, The process of solving the computational resource allocation subproblem under preset constraints to obtain a computational resource allocation strategy includes: Based on the coordinate descent iterative algorithm, each variable in the computational resource allocation subproblem is optimized sequentially according to the load balancing strategy to obtain the computational resource allocation strategy.

4. The method according to claim 2, characterized in that, The process of solving the sensing communication subproblem under preset constraints to obtain the sensing communication strategy includes: Acquire the state space data of the intelligent agent; the state space data includes the amount of sensing data, channel gain, UAV location information, and computing resource allocation strategy. Using the policy network corresponding to the perception and communication subproblem, the action space data of the agent is determined based on the state space data; the action space data includes the perception and communication policy.

5. The method according to claim 4, characterized in that, After determining the agent's action space data based on the state space data using the policy network corresponding to the perception communication subproblem, the method further includes: The reward function value of the agent is determined based on the computing resource allocation strategy and the perception communication strategy; The state space data, the action space data, and the reward function value are stored as empirical data. Based on empirical data, the parameters of the agent's policy network and evaluation network are updated using an LSTM-enhanced proximal policy optimization algorithm.

6. The method according to any one of claims 1 to 5, characterized in that, Before constructing the optimization model aimed at maximizing overall satisfaction, minimizing service duration, and minimizing drone energy consumption, the method further includes: The sensing duration, transmission duration, computation time, and flight time corresponding to each target area are normalized, and the sensing energy consumption, transmission energy consumption, flight energy consumption, and hovering energy consumption corresponding to each target area are also normalized.

7. A resource orchestration device for a cooperative network, characterized in that, The device includes: The perception module is used to determine the satisfaction level of each target area based on the perception duration of each target area, and to determine the perception energy consumption of the UAV based on the satisfaction level of each target area, based on the continuous perception model. The communication module is used to determine the signal-to-interference-plus-noise ratio (SIR / NNR) for each transmission stage based on the communication model, the UAV's flight altitude and transmission power, and the transmission duration and UAV's transmission energy consumption based on the SIR / NNR. The computing module is used to determine the computing time of various computing resources based on the collaborative computing model; The trajectory planning module is used to determine the flight time of the UAV based on the UAV trajectory model, the service sequence, and the UAV's flight altitude; to determine the hovering time of the UAV based on the sensing duration and transmission duration; to determine the flight energy consumption of the UAV based on the flight time; and to determine the hovering energy consumption of the UAV based on the hovering time. The optimization module is used to construct an optimization model with the objectives of maximizing overall satisfaction, minimizing service duration, and minimizing UAV energy consumption. Overall satisfaction is the sum of satisfaction levels for each target area; service duration is the sum of perception duration, transmission duration, computation time, and flight time for each target area; and UAV energy consumption is the sum of perception energy consumption, transmission energy consumption, flight energy consumption, and hovering energy consumption for each target area. Under preset constraints, the optimization model is solved to obtain a computational resource allocation strategy and a perception communication strategy. The computational resource allocation strategy includes the computation time of various computational resources, and the perception communication strategy includes perception duration, service sequence, flight altitude, and transmission power.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images