76 results about "Lyapunov optimization" patented technology

The invention discloses a cloud side-end cooperative task scheduling and efficiency optimization method and system for heterogeneous patrol resources, and relates to the technical field of intelligent scheduling and resource optimization. According to the method, accurate perception of a resource state is realized by constructing a digital twinborn and federated learning mechanism, resource contention conflicts are solved by adopting a space-time diagram attention network and multi-agent reinforcement learning, and multi-target optimization and trusted execution are realized in combination with a quantum genetic algorithm and a block chain smart contract. Finally, the stability of the system is verified through Lyapunov optimization, a complete scheduling system from resource perception and conflict resolution to steady state maintenance is formed, and the task scheduling efficiency and the system stability in the heterogeneous resource environment are remarkably improved.

The invention relates to an intelligent water affair remote monitoring and control system based on the Internet of Things, in particular to the field of intelligent water affair remote monitoring and control, and the system comprises the steps: firstly, pre-training a long and short-term memory network model through a cloud, and generating a lightweight edge prediction model to reduce calculation and transmission loads; the data transmission module effectively reduces the occupation of transmission bandwidth based on a dynamic quantization threshold value of a residual error and a differential coding technology, meanwhile, the high precision of the data is kept, the verification feedback module monitors the accuracy of the transmission data in real time through confidence coefficient detection, and triggers model re-calibration when the confidence coefficient is reduced, so that the long-term stability of the system is ensured, and the reliability of the system is improved. The sampling optimization module dynamically adjusts the sampling frequency according to the confidence coefficient and the residual fluctuation characteristics, balance is achieved between bandwidth occupation and reconstruction precision by using the Lyapunov optimization theory, the real-time performance and accuracy of water affair monitoring are improved, and meanwhile bandwidth consumption and energy consumption are effectively reduced.

The invention discloses an industrial edge reasoning task-oriented model segmentation collaborative unloading method, which comprises the following steps of: obtaining a preferentially processed reasoning task, selecting a reasoning model, constructing a model segmentation candidate set, carrying out retraining and calibration on each segment, establishing a multi-edge collaborative unloading model, and according to a task scale, a segmentation position and a network state, carrying out multi-edge collaborative unloading on the multi-edge collaborative unloading model. Deciding an execution mode of each segment on a local node, an edge node or a cloud node; a Lyapunov optimization framework is introduced, a long-term precision constraint model is constructed, an initial resource allocation scheme is generated, collaborative optimization is carried out in combination with an improved multi-agent near-end strategy optimization algorithm, and an HCRA-AT-MAPPO fusion scheduling result is obtained; based on a Lyapunov optimization decision and an HCRA-AT-MAPPO fusion scheduling result, a model segmentation and unloading strategy is dynamically updated, and network fluctuation and task load change are adapted in real time. The assembling quality can be improved, and the production period can be shortened.

The invention relates to the field of micro-services, and provides an elastic resource scheduling method and device for micro-services and electronic equipment.The method comprises the steps that a to-be-scheduled micro-service system is determined, and the micro-service system comprises a server and a micro-service instance deployed on the server; constructing a long-term scheduling optimization model by taking the minimization of the long-term average request time delay of the micro-service system as a target; converting the long-term scheduling optimization model into a time slot scheduling optimization model in each time slot based on a Lyapunov optimization method; and obtaining an initial solution of the optimization model based on a near-end optimization strategy, then optimizing the initial solution for a dynamically changing request through a pre-trained deep reinforcement learning model, and performing elastic resource scheduling of the micro-service based on an optimization result. According to the elastic resource scheduling method and device for the micro-service and the electronic equipment provided by the invention, scaling and migration of the micro-service instance are performed aiming at the dynamic request arrival rate, and multi-objective optimization of micro-service scaling and migration cost and user request time delay is realized.

The invention discloses a perception-assisted unmanned aerial vehicle track and power distribution joint optimization method and system, and the method comprises the steps: firstly, extracting environment related features from high-dimensional CSI data, constructing a weighted graph, and calculating a shortest path distance matrix; and the CSI is mapped to a two-dimensional space through a twin neural network to maintain distance consistency, and user distribution is obtained by combining with known position correction. On the basis, an optimization model taking the minimization of the energy consumption of the unmanned aerial vehicle as a target and the user rate demand as a constraint is established, the optimization model is decomposed into a power distribution and trajectory optimization sub-problem, and the Lyapunov optimization and successive convex approximation are used for alternate solution. According to the method, high-precision positioning is realized through the channel map under the scene that the user position is unknown, the energy consumption of the unmanned aerial vehicle system is reduced by 27% or above through joint optimization, and meanwhile, 98% of user communication requirements are guaranteed.

The invention belongs to the field of metallurgical industry automation, and relates to an iron and steel joint enterprise energy management system scheduling method based on Lyapunov optimization and deep reinforcement learning. In order to solve the problems that an existing system depends on static rules or experience, is difficult to adapt to dynamic changes, is low in energy scheduling efficiency and the like, a multi-dimensional state space covering energy storage, a virtual queue and environment parameters is constructed, and a scheduling action space including capacity adjustment, multi-energy medium charging and discharging control and the like is defined; a Lyapunov drift function is introduced as a stability constraint, a reinforcement learning agent of an Actor-Critic framework is constructed, and global energy collaborative optimization is realized in combination with an energy sub-optimization problem solving and parallel sampling mechanism. The method can give consideration to scheduling economy and system stability, has dynamic adaptive ability, ensures project feasibility of the scheme, can improve energy utilization efficiency, reduces cost and carbon emission, and is suitable for iron and steel enterprises of different scales.

A traceable method for task replication and resource allocation of a large-scale machine type communication network belongs to the technical field of wireless communication and mobile edge computing, and comprises the following steps: based on a Lyapunov optimization theory, establishing physical and virtual queue quantization constraints, and converting long-term optimization into a drift and penalty minimization problem of each time slot; decomposing into task replication, resource block allocation and edge server queue management sub-problems, and sequentially solving; determining an optimal task replication scheme, and storing the optimal task replication scheme as a first decision-making voucher; determining a resource block allocation scheme, and storing the resource block allocation scheme as a second decision-making voucher; determining an optimal edge server queue management scheme which comprises an edge server task processing rate and a task discarding amount, and storing the optimal edge server queue management scheme as a third decision-making voucher; updating a queue state, packaging the decision-making voucher and the queue state into a periodic transaction, writing the periodic transaction into a block chain, and forming an operation auditing log; according to the method, the optimal balance of long-term operation cost is realized, and the decision-making process has traceability.

The invention discloses a dynamic state updating scheduling method in a multi-terminal wireless communication system. The method comprises the following steps of: aiming at the problem that uplink resources are limited, constructing a virtual queue to represent the historical resource use condition of a terminal; dynamic updating indexes are designed to quantify urgency of state updating; high-value terminal screening and an optimal scheduling decision are executed in each time slot; and a Lyapunov optimization framework is creatively combined with a dynamic threshold mechanism. By adopting the method provided by the invention, the average information urgency can be minimized, and the information transmission performance of the system is effectively improved.

The invention discloses a test task scheduling and management system based on a 5G power virtual private network, and relates to the technical field of 5G communication and power system automation. A two-layer distributed architecture based on core and region edge nodes is adopted, and task domain division management and resource synchronization are realized through a lightweight Raft protocol; calculating a task key index TCI by using a four-dimensional label and a key dimension index amplification method, and dynamically scheduling resources in combination with a Lyapunov optimization model; and according to the health degree index of the access equipment, adaptively correcting resource reservation through the prediction model. According to the method, the problems of high time delay, extensive resource allocation and the like of a traditional centralized architecture are solved, distributed collaborative scheduling and resource dynamic optimization of the power test tasks are realized, and the system reliability is improved.

The invention discloses a video transmission method and device based on Lyapunov optimization, electronic equipment and a storage medium. The method is applied to satellite node ends. The method comprises the following steps: acquiring initial state information of a current transmission time slot and video data to be transmitted; the initial state information comprises queue state information representing a coding environment and transmission state information representing a transmission environment; inputting the initial state information into a preset strategy action model to obtain action control parameters corresponding to the initial state information; and coding and transmitting the video data to be transmitted according to the action control parameters. Therefore, compared with the prior art that the video to be transmitted is coded and transmitted by adopting the preset fixed configuration parameters, the action control parameters output by the strategy action model dynamically adapt to the coding environment and the transmission environment of the satellite node end; the satellite node end encodes and transmits the video data to be transmitted according to the action control parameters, so that the video quality is better, and the transmission effect is better.

The present application relates to a kind of static-mobile energy storage collaborative space-time flexible regulation method, it relates to a kind of energy storage optimization scheduling method, the method includes the following steps: (1) the distribution of regional integrated energy system new energy, load access characteristic and consumption demand are analyzed, and surplus model of renewable energy is constructed;(2) according to the source and load characteristics of regional integrated energy system, static energy storage charging and discharging scheduling model is constructed;(3) according to surplus model of renewable energy, the space-time scheduling strategy of mobile energy storage is proposed;(4) according to static energy storage charging and discharging scheduling model and the space-time scheduling strategy of mobile energy storage, the queuing model of static-mobile energy storage charging and discharging strategy is constructed;(5) according to the queuing model of static-mobile energy storage charging and discharging strategy, based on improved Lyapunov optimization, static-mobile energy storage collaborative dynamic scheduling is realized.The method of the present application provides support for the collaborative consumption of distributed renewable energy between regional integrated energy system.

The application discloses a server energy efficiency dynamic optimization method based on workload prediction and a deep learning model, relates to the cross field of server energy efficiency optimization and deep learning technology, and comprises the following steps: collecting server multi-source data, performing space-time alignment on the server multi-source data, extracting causal characteristics, constructing a standardized data set, constructing a time series graph convolution network, combining a multi-head attention mechanism, outputting a probabilistic load prediction interval, constructing an energy efficiency objective function based on a Lyapunov optimization framework, quantifying performance loss and temperature drift, and outputting multi-agent optimization constraints, defining a multi-reinforcement learning agent, outputting a collaborative adjustment instruction through a counterfactual baseline algorithm, constructing a digital twin shadow model to deduce energy efficiency, combining a meta-learning fine-tuning model, forming a closed-loop optimization of physical and digital double verification, realizing dynamic and accurate optimization of server energy efficiency, and adapting to multiple computing power scenes.

The invention provides a space-air-ground integrated network resource management method and system and a storage medium, and belongs to the technical field of wireless communication and network resource dispatch. The method comprises the steps that a space-air-ground integrated network is constructed, a GU is provided with an energy collection module and a backscatter circuit, and the backscatter circuit is connected with the GU; low-power-consumption data transmission and energy collection are realized by reflecting a radio frequency carrier signal emitted by the UAV; the UAV carries a mobile edge computing server to process a task, and the task is unloaded to the LEO when the task is overloaded; a central controller collects network state and channel state information in real time, a random optimization function is constructed by taking minimization of long-term average energy consumption of a system as a target and combining constraints such as task queue stability and energy sustainability, a Lyapunov optimization framework is adopted to convert the problem into a single time slot problem, and the problem is solved through alternate optimization and concave-convex process technologies. According to the method, the stability of the task queue is guaranteed, the system energy consumption is remarkably reduced, the resource utilization rate is improved, and reliable support is provided for remote area power emergency communication.

The application discloses an edge intelligent body flexible scheduling method based on Lyapunov optimization and Stackelberg game. The application constructs a cloud-edge-end three-layer edge intelligent body federated learning network architecture to complete distributed collaborative training under the premise of protecting data privacy. In view of the problems of heterogeneous edge intelligent body network nodes, non-independent and identically distributed data and limited communication resources, a long-term joint optimization model of client selection and bandwidth allocation is established, and model accuracy, time delay and energy consumption are considered. Lyapunov optimization is used to decouple long-term random optimization into real-time decision-making per time slot, and to convert constraints into virtual queue stability control. Dynamic client selection is realized based on Stackelberg game, and high-quality nodes are adaptively selected. Adaptive bandwidth allocation is completed through an evolutionary algorithm, and communication requirements of key clients are preferentially guaranteed. The application can significantly improve model accuracy, reduce time delay and energy consumption, guarantee long-term stability of the system, and is suitable for high-dynamic edge intelligent body federated learning scenes.

The present application relates to the field of distributed large model inference, and particularly relates to a distributed large model inference method based on token perception and Lyapunov optimization. The scheme comprises: after receiving a user input prompt, predicting the output token length of the corresponding task through a length perception semantic module, calculating the corresponding total amount of work, based on the token length prediction result, the real-time state of the device and the long-term virtual queue state, generating a final unloading decision through the Lyapunov guided unloading optimization module iterative unloading algorithm; the client unloads the task to the corresponding cloud edge device according to the final unloading decision, and the device executes the large model inference; the scheduler updates the virtual queue length according to the actual task execution time and the device computing power utilization rate, and provides input for the next time slot decision. The present application is suitable for inference systems utilizing distributed large models.

This invention discloses a federated learning method for intelligent IoT, comprising: dividing the global model into continuous fragments on the server side, configuring lightweight auxiliary head and tail segments to simulate contextual features; introducing an event-triggered resource monitoring and Lyapunov optimization-driven dynamic fragment adjustment mechanism; employing fragment-level knowledge distillation to achieve cross-fragment knowledge transfer, combined with split gradient descent optimization. This invention solves the problems of heavy communication and computational burdens and poor adaptability in resource-constrained scenarios of existing federated learning and fragmented federated learning technologies. This invention can reduce device storage, computation, and communication overhead, adapt to dynamic resource changes, improve model convergence speed and consistency, and performs excellently under both IID and Non-IID data distributions, making it suitable for distributed training scenarios of heterogeneous AIoT devices.

This invention provides a task collaborative reasoning method, apparatus, electronic device, and medium, relating to the field of computer technology. The method includes: dividing a task arriving on a mobile device into multiple subtasks and constructing an offloading strategy for each subtask to characterize execution on the mobile device or MEC server; constructing an average overhead minimization problem for all mobile devices within a time slot for each mobile device; reconstructing the average overhead minimization problem using Lyapunov optimization theory to obtain an independent time slot policy optimization problem corresponding to the average overhead minimization problem; transforming the independent time slot policy optimization problem into a Markov decision process and solving the Markov decision process using a deep reinforcement learning algorithm to obtain the optimal offloading strategy, optimal mobile device transmission power, and optimal number of threads allocated to the MEC server for each mobile device. The invention effectively utilizes distributed computing resources and significantly improves reasoning efficiency.

This invention discloses a client selection method and system for multi-task federated learning. Addressing the shortcomings of client selection and insufficient consideration of task urgency in multi-task dynamic federated learning scenarios, this invention first constructs a multi-task federated learning system model, defining a utility function that includes learning quality and penalty terms. Second, it establishes fairness constraints and introduces a fairness queue to transform the problem into a queue stability problem. Then, based on Lyapunov optimization theory, it constructs a drift-plus-utility function, transforming a long-term stochastic optimization problem into a deterministic optimization problem for each round of communication by minimizing its upper bound. Finally, it constructs an auxiliary bipartite graph to transform client selection into a minimum-weight bipartite graph matching problem. This invention, by jointly optimizing fairness, learning quality, and task urgency through the Lyapunov framework, transforms long-term constraints into a solvable problem for each round, reducing computational complexity and achieving efficient and fair dynamic client selection.

This invention relates to the field of the Internet of Things (IoT) and discloses an optimization method for energy and task scheduling in an edge computing system, comprising the following steps: First, based on a new energy and task scheduling protocol, the MEC system can adaptively switch between energy harvesting mode, IRS auxiliary task offloading mode, and IRS standby task offloading mode according to channel conditions, IRS battery energy state, and user task queue state; Second, based on the developed protocol, the system optimization problem is modeled to minimize long-term user task offloading and computing energy consumption; Third, using the Lyapunov optimization method, the problem is decomposed into a time-slot-based deterministic optimization problem, and the corresponding deterministic optimization problem is solved using convex optimization theory.

The invention discloses a sampling rate and power optimization method based on a single-cluster air federated edge learning system. According to the method, an optimization problem with the optimization objective of minimizing the difference between expected global loss and optimal loss after T times of communication is established, and an original optimization problem is converted into an approximate optimization problem containing an explicit formula. In an optimization problem solving process, long-term energy constraint is converted into queue stability constraint based on a Lyapunov optimization method, and an EARA algorithm is provided to realize online joint optimization of a data sampling rate and a transmission power scaling factor. Experimental results show that the EARA algorithm can significantly improve the convergence performance of the model under a relatively low energy budget.

The invention discloses a virtual power plant distributed optimization and feasible region analysis method and system. The method comprises the following steps: establishing an interactive physical model of a virtual power plant and a power distribution network; based on an interactive physical model, relaxation decoupling is carried out on wide-area and neighborhood time coupling constraints caused by the state of charge of the energy storage system through a Lyapunov optimization theory, and a multi-period optimization problem of the virtual power plant is converted into a single-period optimization sub-problem; aiming at the single-period optimization sub-problem, adopting a particle swarm optimization algorithm HSIPSO improved based on a hybrid strategy and fusing the particle swarm optimization algorithm HSIPSO into an ADMM distributed optimization framework to form an ADMM-HSIPSO algorithm, and carrying out distributed solution on the single-period optimization sub-problem; converting a multi-period virtual power plant feasible region representation problem into a calculation problem of each scheduling period feasible region; a mathematical model of the feasible region of the virtual power plant is constructed, and an ADMM-HSIPSO algorithm is adopted for solving; the coordination and unification of the optimal distribution of the dispatching power of the power distribution network among the virtual power plants and the privacy protection in the virtual power plants are realized.

The application discloses a liquid crystal phased array fast-response bistable driving and low-power scanning method, equipment and medium, and relates to the technical field of liquid crystal phased array.The application compares the deviation matrix generated by real-time acquisition of liquid crystal molecule response data with the digital twin theory value, combines the deep reinforcement learning to output the electric field parameter, generates the non-uniform composite electric field through the reverse design of topological photonics, realizes the high-precision bistable steering of molecules, reconstructs the adaptive sparse scanning path based on the Bayesian compressed sensing, optimizes the energy distribution by matching the dynamic energy spectrum shaping strategy, predicts the molecule steering state through the physical information neural network embedded with physical constraints, generates the pre-control driving sequence combined with the event triggering mechanism, and outputs the optimal control instruction through the Lyapunov optimization, so that the steering response speed and stability of the liquid crystal phased array are effectively improved, and the energy utilization efficiency is greatly improved.

The invention discloses an industrial park scheduling method and system fused with workload migration of a data center, and belongs to the technical field of energy scheduling, and the method comprises the steps: obtaining an industrial park energy constraint rule according to an energy conversion coefficient representing an energy conversion characteristic and an energy demand of an industrial park; constructing a workload queue based on the dynamic change relationship of the task receiving amount, the task processing amount and the task migration amount of the data center host, and obtaining a scheduling rule according to a migration constraint condition and a migration execution rule of the workload queue; through Lyapunov optimization, constructing a final objective function by taking the lowest total operation cost of the industrial park, the lowest operation cost of the data center and the long-term stable operation of the data center as final objectives; and based on an industrial park energy constraint rule, a scheduling rule and a final objective function, obtaining an optimal scheduling strategy by using TD3. The technical problem that it is difficult to guarantee the system stability while improving the resource utilization rate in the prior art is solved.

The application discloses a multi-modal scheduling method based on early exit under edge-limited equipment and relates to the field of resource offloading.The method comprises the following steps: establishing a multi-modal scheduling framework, including an edge equipment end and a server end; obtaining a plurality of early exit nodes by using an early exit mechanism; under the constraint of maximum delay, establishing a multi-modal offloading scheduling system model based on Lyapunov optimization according to network environment information and the early exit nodes, and taking Lyapunov drift penalty minimization as an optimization objective; generating a modal offloading strategy by using the multi-modal offloading scheduling system model; generating a first modal recognition result at the edge equipment end according to the modal offloading strategy, and transmitting the second spectral data to the server end after adjusting the resolution of the second spectral data, so that the server end generates a second modal recognition result; and performing probability integration on the first modal recognition result and the second modal recognition result at the edge equipment end, so as to obtain a prediction result.Compared with the prior art, the application can guarantee system stability and model processing accuracy and can improve the processing speed of the system.

The invention discloses a new energy state-based cross-regional inter-data-center computing collaborative scheduling method, which comprises the following steps of: firstly, constructing a regional differentiation load model, and accurately depicting characteristic differences of characteristic computing power parks of different regions; then designing an adaptive scheduling algorithm based on Lyapunov optimization, fusing a prediction-reaction mechanism, and realizing dynamic balance between the stability of computing collaborative scheduling among multiple data centers and the new energy consumption rate; a V-parameter adaptive adjustment mechanism and a hierarchical fault-tolerant guarantee system are further provided to cope with new energy fluctuation and improve the robustness of the system. And finally, through self-adaptive strategy adjustment of the V parameter, the green computing power proportion is improved while the computing power processing requirements are met in different environment scenes.

The invention relates to the technical field of satellite communication networks, in particular to a multi-strategy resource allocation method based on Lyapunov optimization, which is characterized by comprising the following steps: S1, sensing system state information in a current time frame by using a deep online unloading framework DROO; s2, a Lyapunov virtual migration cost queue is constructed; s3, fusing the plurality of candidate unloading actions based on a Softmax weight weighting and entropy regularization method; s4, executing task unloading and resource allocation by the hybrid unloading strategy; the multi-strategy fusion mechanism provided by the invention effectively overcomes the convergence problem of a single strategy, enhances the strategy exploration capability through entropy regularization, improves the response and decision-making level of the system in an abnormal scene, supports online learning and dynamic updating, can feed back and optimize DROO model parameters, accelerates the convergence speed, and improves the robustness of the system. The mechanism is suitable for non-stationary scenes such as high-speed movement of low-orbit satellites, task processing efficiency is improved, migration overhead is reduced, and service quality is guaranteed.

The application discloses a kind of electric carbon perception-based computing power network task scheduling method, system and device, the method includes: introducing carbon tax mechanism, constructs electric carbon perception-based computing power network task scheduling model;Based on the computing power network task scheduling model, the problem of minimizing system electric carbon cost is generated;Based on Lyapunov optimization, the problem is transformed, and the optimal solution of scheduling scheme is obtained by using DPRA algorithm solution.The system includes: model construction module, problem definition module and solving module.Through using the application, green and stable computing power network task scheduling can be realized.The application can be widely applied to computer computing power network field.

The invention provides a complex working condition co-electrolysis system electro-thermal-chemical stability calculation method and a related device, and belongs to the field of co-electrolysis system control. The method comprises the following steps: firstly, calculating a nonlinear tensor according to an initial state of an electro-thermal-chemical coupling field of a co-electrolysis system; predicting the system state at the next moment, and obtaining an optimization control parameter at the next moment by using a Lyapunov optimization method and taking minimization of system fluctuation as a target; a multi-field coupling state matrix at the next moment is calculated according to the optimization control parameters; and finally, evaluating the service life of the system based on thermal stress distribution, and taking the multi-field coupling state matrix, the optimization control parameter and the service life evaluation result as stable calculation results at the next moment. According to the method, the method of dynamically calculating nonlinear physical property parameters, Lyapunov optimization control and multi-field coupling state and life evaluation closed loop is adopted, electro-thermal-chemical stable operation of the co-electrolysis system under complex working conditions is achieved, the hydrogen production cost is reduced, and the service life of equipment is prolonged.

The invention relates to a satellite internet virtual network function deployment and scheduling optimization method, which is characterized by comprising the following steps of: constructing a joint time delay-resource constraint model comprising a node data queue and a node virtual cost queue in a satellite-ground fusion satellite network; deriving a Lyapunov drift-penalty upper bound based on a data queue and a virtual resource cost queue, and converting a long-term delay and resource cost joint optimization problem into a time slot-by-slot solvable optimization problem; a motion decomposition structure method based on deep reinforcement learning is adopted, and a virtual network function deployment motion, a link mapping motion and a node resource allocation motion are modeled as sub-motions respectively. According to the technical scheme, the Lyapunov optimization theory is introduced into service function chain (SFC) deployment and scheduling in the satellite internet, a dual feedback mechanism of a task data queue and a resource virtual cost queue is established, unified modeling and dynamic control of time delay stability and resource budget constraint are achieved, and the long-term stability of the system is remarkably enhanced.

The application provides a kind of unmanned aerial vehicle task unloading method and device based on deep reinforcement learning, it is related to vehicle networking technical field.The method comprises: according to the computing capacity of vehicle, task is handled by vehicle or edge server;If handled by edge server, the vehicle position is obtained by constructing the Gaussian-Markov mobility model of vehicle, according to the computing resource of unmanned aerial vehicle and the distance between vehicle and intelligent roadside facility, it is judged by unmanned aerial vehicle or by unmanned aerial vehicle and intelligent roadside facility;If by unmanned aerial vehicle, the path planning result of unmanned aerial vehicle is output according to DGBCO model;If by intelligent roadside facility, task allocation is optimized according to computing reuse technology;Unmanned aerial vehicle energy constraint problem based on Lyapunov optimization is constructed;MADDPG algorithm is used to obtain task unloading strategy.The application optimizes UAV trajectory, designs a kind of joint optimization method in combination with the dynamic change of the mobility and computing resource of vehicle, improves the comprehensive performance of system.