CN-122026958-A - Unmanned aerial vehicle power calculation network method assisted by dynamic partition intelligent reflecting surface

CN122026958ACN 122026958 ACN122026958 ACN 122026958ACN-122026958-A

Abstract

The invention discloses an unmanned aerial vehicle power calculation network method assisted by a dynamic partition intelligent reflecting surface, which comprises the steps of firstly constructing a system model, determining a power calculation communication matching and energy consumption model, then constructing an optimization problem aiming at minimizing user delay and unmanned aerial vehicle energy consumption, converting the optimization problem into a single-target optimization problem to be decomposed into three sub-problems of phase shift control and user association, calculation-communication resource matching and unmanned aerial vehicle flight track optimization, and finally solving the sub-problems by using a phase alignment and channel optimal matching method, a successive convex approximation method and a multi-agent adjacent strategy optimization method. The invention not only realizes the joint allocation of the reflecting unit and the computing resource, effectively improves the utilization rate of the system resource, but also effectively solves the problem of mixed integer non-convex optimization, and enables the system to realize the optimal balance of time delay and energy consumption in a dynamic environment. Simulation results show that the scheme can stably obtain the optimal comprehensive performance of the system in various scenes.

Inventors

NING ZHAOLONG
HU HAO
REN JIAXIN
HUANG YIHANG
WANG XIAOJIE
YI LING

Assignees

青海民族大学

Dates

Publication Date: 20260512
Application Date: 20260123

Claims (10)

1. The unmanned aerial vehicle power calculation network method assisted by the dynamic partition intelligent reflecting surface is characterized by comprising the following steps of: The method comprises the steps of constructing a system model, determining a computational power communication matching and energy consumption model, wherein the system model comprises K unmanned aerial vehicles for providing computing services for N users, and intelligent reflecting surfaces for enhancing the channel environment between the users and the unmanned aerial vehicles and accelerating the transfer of computing tasks; Constructing an optimization problem aiming at minimizing user delay and unmanned aerial vehicle energy consumption, then converting the problem into a single-target optimization problem by using a linear weighting sum method and decomposing the single-target optimization problem into three sub-problems of phase shift control and user association, calculation-communication resource matching and unmanned aerial vehicle flight trajectory optimization; The method comprises the steps of solving a sub-problem of phase shift control and user association by using a phase alignment and channel optimal matching method, solving a calculation-communication resource matching sub-problem by using a method based on successive approximation under the condition of giving the unmanned aerial vehicle position, a user association strategy and an optimal phase shift control strategy, optimizing the unmanned aerial vehicle flight track sub-problem by using a method based on multi-agent proximity strategy on the basis of giving the user association, the phase control strategy and the calculation-communication resource matching scheme, and finally obtaining an optimal solution.
2. The dynamically partitioned intelligent reflective surface assisted unmanned aerial vehicle power network method of claim 1, wherein the optimization problem is: Wherein, the The flight path of the unmanned aerial vehicle is represented, Representing the association decision between user n and drone k, Representing the allocation of computing resources for the drone k, Indicating the distribution of the reflective elements, Indicating the phase shift of the reflecting element allocated by user n, the reflecting element specification of the intelligent reflecting surface is a x b, the total service time By T time slot intervals of equal length The composition of the composite material comprises the components, Representing tasks Is a maximum acceptable amount of delay; representing the total energy consumption of the drones k, V representing the speed of each drone, Representing a collection of unmanned aerial vehicles, Representing a set of users; 、 respectively representing coordinates of the unmanned aerial vehicles k and j at the time slot t; representing a minimum safe distance between unmanned aerial vehicles; representing the association between the user n and the drone k at time slot t, Representing that user n is served by drone k at time slot t, and the computing resources of drone k allocated to user n at time slot t are represented by variables The definition of the term "a" or "an" is, Representing its maximum computing power; representing the number of rows of reflective elements allocated to user n by the intelligent reflective surface at time slot t; A phase shift matrix representing the reflective elements assigned by user n; The method comprises the steps of limiting the moving distance of unmanned aerial vehicles in each time slot by constraint conditions C1, ensuring the safe distance between unmanned aerial vehicles by constraint C2, defining the value of an associated decision variable by constraint C3, ensuring the allocation of each user to one unmanned aerial vehicle by constraint C4, representing the calculation capacity allocation limit of each unmanned aerial vehicle by constraint C5, applying total delay constraint to calculation tasks by constraint C6, defining the quantity limit of reflecting elements allocated to user n by intelligent reflecting surfaces by constraint C7 and constraint C8, and defining the value of a phase shift variable by constraint C9; the single objective optimization problem using the linear weighted sum method is: Wherein, the And Is the weight.
3. The unmanned aerial vehicle power network method assisted by the dynamic zoning intelligent reflecting surface according to claim 2, wherein the phase shift control and user association sub-problem decomposed by the single-objective optimization problem can be represented by a closed-form solution, and the user association can be obtained by adopting a channel optimal matching strategy, and the closed-form solution of the optimal phase control is as follows: Wherein, the And Representing the vertical and horizontal angles of arrival from user n to the intelligent reflective surface, respectively; And Respectively representing the vertical and horizontal departure angles from the intelligent reflecting surface to the unmanned aerial vehicle k; Representing the carrier frequency; representing the speed of light; Is an imaginary unit; And Representing the column and row spacing of the IRS reflective elements, respectively; And Representing column and row indices, respectively; The computational-communication resource matching sub-problem into which the single-objective optimization problem is resolved is: the unmanned aerial vehicle flight trajectory optimization sub-problem decomposed by the single-target optimization problem is as follows: 。
4. a dynamically partitioned intelligent reflective surface assisted unmanned aerial vehicle power network method according to claim 3, wherein the phase shift control and user associated sub-problem is solved using a phase alignment and channel optimal matching method, comprising the steps of: The optimal phase control from user n to drone k at time slot t can be expressed as: the channel best match is used to obtain the user association, i.e. each user selects a drone with the best channel conditions at each time slot t.
5. The unmanned aerial vehicle power network assisted by dynamically partitioning intelligent reflective surfaces method according to claim 4, wherein the calculation-communication resource matching sub-problem is solved by a successive approximation based method, comprising the steps of: Variable of integer Relaxation to continuous variable Obtaining relaxation problem As the lower bound of P2: wherein constraint C10 is the relaxation of original constraint C8; Problems to be solved The decomposition into two pieces of iterative optimization can be expressed as: Assigning a sub-problem to the computing resource, and solving the sub-problem by adopting CVX; the sub-problem is allocated to the non-convex reflecting unit, and is processed by adopting a continuous convex approximation method, which comprises the steps of introducing a time delay upper bound Is provided with , , The channel power gain after applying the optimal phase control is expressed as: Wherein, the Representing the distance of user n from the intelligent reflecting surface i over time slot t, Representing the distance of drone k from the intelligent reflecting surface over time slot t, Representing the distance between the user n and the unmanned plane k in the time slot t; Representing the channel gain at a reference distance of 1 meter; 、 And Respectively representing path loss indexes between a user and the IRS, the IRS and the unmanned aerial vehicle and between the user and the unmanned aerial vehicle; representing the rice factor between the user and the IRS; By at local points The transmission delay term is subjected to first-order Taylor expansion to construct a convex approximation problem : Wherein the method comprises the steps of , , And ; Representing the number of offloading bits required to handle the task, B represents bandwidth, Representing a computation delay; solving problems using CVX And obtaining a continuous solution of the reflecting unit distribution sub-problem.
6. The unmanned aerial vehicle power network method assisted by dynamically partitioning intelligent reflecting surfaces according to claim 5, wherein the continuous solution of the reflecting unit allocation sub-problem further adopts a heuristic rounding strategy based on channel gain search to obtain an integer solution thereof, comprising the following steps: Rounding down the continuous solution to obtain an initial integer solution and the number of remaining reflection units ; Estimating the channel power gain variation caused by additionally distributing a reflecting unit for each user; Will remain behind The units are assigned to users with the greatest gain variation to obtain an approximate integer solution.
7. The unmanned aerial vehicle power network assisted by dynamic zoning intelligent reflecting surfaces method according to claim 5, wherein the unmanned aerial vehicle flight trajectory sub-problem is optimized by a method based on multi-agent proximity policy optimization, comprising the steps of: Modeling the unmanned aerial vehicle flight trajectory sub-problem as a decentralised part considerable Markov decision process, training by adopting a multi-agent reinforcement learning framework for centralized training-decentralized execution, and jointly training an actor-critique network by designing a reward function integrating time delay, energy consumption and safety constraint and utilizing a near-end strategy optimization method, so that each unmanned aerial vehicle can finally autonomously decide to obtain an efficient flight trajectory based on local observation.
8. The dynamically zoned intelligent reflective surface-assisted unmanned aerial vehicle power network method of claim 7, wherein modeling the unmanned aerial vehicle flight trajectory subproblems as a decentralised partial observability markov decision process comprises the steps of: the intelligent agent is each unmanned plane, and the state space thereof Comprises unmanned plane coordinates and user calculation tasks, and observation The combined action comprises the self coordinates, other unmanned plane coordinates and task information of the related user Defining a mobile decision of all unmanned aerial vehicles; in order to meet the flight safety constraint C2, if the current behavior violates the constraint, a negative penalty term is introduced in the bonus function, which is correspondingly negated because the multi-agent proximity policy optimization is trained to maximize the jackpot; Is not provided with a negative sign Representing a penalty term, the reward function being represented as: The multi-agent reinforcement learning framework for centralized training-decentralized execution comprises the steps that in a training stage, critics networks of all agents are updated based on global observation to accurately evaluate state values, in an execution stage, all unmanned aerial vehicles select actions only according to local observation to reduce communication expenditure, network parameter sharing mechanisms are adopted to improve learning efficiency in consideration of isomorphism of the agents, and the multi-agent reinforcement learning framework specifically comprises the following steps: Actor network parameters Input/output actions with local observations The loss function is optimally designed based on the near-end strategy as follows: wherein the symbols are And Representing new and old policies, respectively, functions And Representing the dominance function and the clipping function respectively, Representing clipping parameters; critics network parameters Taking global observation as input and output state value estimation, the loss function is defined as follows based on time sequence difference errors: Wherein the function is The representation is based on a critic network To the state Value assessment of (C) sign Empirical data representative of a sample is provided, A prize value that is discounted; Through alternately optimizing actor and criticizer networks, the algorithm gradually learns to a long-term optimal collaborative flight strategy, and the calculation complexity is that 。
9. The dynamically zoned intelligent reflector assisted unmanned aerial vehicle power network method of claim 1, wherein the intelligent reflector comprises a x b reflective element capable of adjusting the phase of the reflected signal to form a beam in a particular direction, wherein the receive channel between user n and the intelligent reflector can be modeled by a rice fading model, wherein at time slot t, the method comprises the steps of 、、 The method comprises the steps of respectively representing channel vectors from the UAV to the intelligent reflecting surface, from the intelligent reflecting surface to the user and from the unmanned aerial vehicle to the user link, setting all communication links as a rice fading channel model, adopting an orthogonal frequency division multiple access technology to reduce interference among users, and dividing main energy consumption of the unmanned aerial vehicle k in a time slot t into calculation energy consumption and flight energy consumption.
10. A computer readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor implements the steps of the dynamically partitioned intelligent reflector assisted unmanned aerial vehicle computational power network method of any of claims 1 to 9.

Description

Unmanned aerial vehicle power calculation network method assisted by dynamic partition intelligent reflecting surface Technical Field The invention relates to a dynamic zoned intelligent reflecting surface assisted unmanned aerial vehicle power calculation network in the academic field, which jointly optimizes unmanned aerial vehicle flight tracks, power distribution, reflecting element distribution, intelligent reflecting surface phase shift and unmanned aerial vehicle-user association. Background The statements in this section merely provide background information related to the present disclosure and may constitute prior art. In carrying out the present invention, the inventors have found that at least the following problems exist in the prior art. A computing network is an architecture that integrates distributed computing resources, capable of providing a wide range of computing services. Compared with the traditional mobile edge computing system, the computing network needs to arrange servers with wider coverage range, but the large-scale deployment of the edge servers easily causes resource waste. The unmanned aerial vehicle has the characteristics of flexibility and expandability, can carry various sensors and computing equipment, and provides convenient computing unloading service for users. The unmanned plane is combined with the computational power network, so that the wide coverage and flexible allocation of the computational resources are realized. Despite the advantages of unmanned aerial vehicle power networks, the unmanned aerial vehicle power networks still face the problems of limited energy, shortage of frequency spectrum resources and the like. In addition, the delay has a significant impact on the quality of service, and it is necessary to further reduce the system delay. Because the hardware condition of the unmanned aerial vehicle is limited, the improvement of the computing capacity of the unmanned aerial vehicle is not easy, and in addition, the system optimization faces challenges due to the dynamically changing environment and the user requirements. The intelligent reflecting surface provides a feasible scheme for reducing the time delay of the unmanned aerial vehicle power network. The intelligent reflecting surface can enhance the channel capacity by adjusting the phase of the reflecting element, and has the characteristics of low cost and easy deployment, so that the intelligent reflecting surface becomes an economic and effective means for reducing the transmission delay in the unmanned aerial vehicle power network. Meanwhile, the intelligent reflecting surface can dynamically adjust the signal beam direction according to the real-time change of the position of the unmanned aerial vehicle, so that the channel quality between a user and the unmanned aerial vehicle is enhanced. The patent name of the application number 202310912964.2 is 'an unmanned aerial vehicle assisted edge computing network resource allocation method based on intelligent reflection surface assistance', the average throughput of a vehicle is maximized by jointly optimizing the computing task amount, computing resources and unloading decisions of an unmanned aerial vehicle assisted edge computing system, intelligent reflection surface phase shifting and vehicle transmission power allocation, and a continuous convex approximation and block landing algorithm is adopted to optimize the maximum average throughput function so as to improve the channel gain of a link between the vehicle and the unmanned aerial vehicle. Although intelligent reflector-assisted unmanned aerial vehicle computing networks have the potential described above, there are several key issues that need to be addressed. Firstly, the intelligent reflecting surface introduces new adjustable resources for the system, so that the distribution of heterogeneous resources is more complex. Furthermore, traditional smart reflector allocation strategies have limited effectiveness in unmanned power networks, which generally assume that smart reflectors are allocated to users as a whole, which makes it difficult to meet the needs of multiple users simultaneously. Finally, the dynamically changing user requirements and the mobile environment change the association relationship between the user and the unmanned aerial vehicle, so that the problem of distribution of the intelligent reflecting surface and the computing resources is further complicated. Under the condition of solving the problems, the invention can effectively improve the utilization rate of system resources, and enables the system to realize the optimal balance of time delay and energy consumption in a dynamic environment. Disclosure of Invention In view of the above, it is an object of the present invention to solve some of the problems of the prior art, or at least to alleviate them. The unmanned aerial vehicle power calculation network method assisted by the dynamic partition intelligent re