A task scheduling and resource allocation method of a multi-unmanned aerial vehicle cooperative computing power network

Abstract

The application discloses a kind of multi-unmanned aerial vehicle collaborative computing power network task scheduling and resource allocation method, by joint optimization multi-unmanned aerial vehicle between task scheduling and resource allocation to make communication network system total energy consumption minimization, the steps of the method are as follows: step one, establish multi-unmanned aerial vehicle collaborative computing power network system model, deduce collaborative computing power network system total energy consumption;Step two, describe the joint task scheduling and resource allocation optimization problem with system total energy consumption minimization as target;Step three, adopt Lyapunov optimization and Lagrange dual method to solve the joint optimization problem of task scheduling and resource allocation.The application realizes the task scheduling and resource allocation optimization of multi-unmanned aerial vehicle collaborative computing power network, realizes the total energy consumption minimization of computing power network system.

Description

Technical Field

[0001] This invention relates to the field of multi-UAV collaborative computing network communication technology, specifically a method for task scheduling and resource allocation in a multi-UAV collaborative computing network. Background Technology

[0002] Computing power networks represent a systemic, architectural, and capability innovation within China's information and communication industry, adapting to industry development trends and emerging market demands. With the development of computing power networks, computationally intensive and latency-critical applications are constantly emerging. However, resource-constrained equipment cannot provide sufficient computing power to meet the demands of these applications for low latency, low power consumption, and massive computational loads. Reducing energy consumption and improving resource utilization have become urgent needs for the operation of computing power networks.

[0003] Unmanned aerial vehicle (UAV)-assisted communication (UAV-assisted communication) is increasingly attracting attention from academia and industry as a supplement to terrestrial communication. UAVs, as typical segmental communication devices, are characterized by low trajectory loss and high line-of-sight probability. Therefore, airborne UAV base stations can provide reliable, high-quality communication coverage for ground user equipment. Furthermore, UAVs equipped with edge servers can perform computational tasks and provide edge computing for tasks offloaded from ground user equipment. Utilizing trajectory planning to control UAVs can improve terminal access and network service quality in hotspot areas.

[0004] Therefore, joint optimization design of task scheduling and resource allocation for multi-UAV collaborative computing networks is crucial. This problem is highly non-convex, and the coupling relationship between UAV task scheduling and UAV resource allocation variables is complex. When the environment changes, the optimization process must be recalculated, and the computational complexity increases exponentially. To address this, this invention proposes a task scheduling and resource allocation method for multi-UAV collaborative computing networks. It describes a joint task scheduling and resource allocation optimization problem with the objective of minimizing total system energy consumption. Subsequently, this invention utilizes Lyapunov optimization to handle long-term queue delay constraints, decomposing the original problem to avoid directly solving the complex and highly non-convex optimization problem. Finally, the Lagrange duality method is employed to solve the joint optimization problem of task scheduling and resource allocation. Summary of the Invention

[0005] This invention discloses a task scheduling and resource allocation method for a multi-UAV collaborative computing network. Addressing the issue that ground user equipment cannot provide sufficient computing power to meet the low latency, low power consumption, and massive computational demands of communication processes, this invention proposes a joint task scheduling and resource allocation optimization method to minimize the total system energy consumption. The method comprises the following steps: Step 1: Establishing a multi-UAV collaborative computing network system model, considering the task offloading energy consumption and task computation energy consumption of the UAVs, and deriving the total energy consumption of the collaborative computing network system; Step 2: Describing the joint task scheduling and resource allocation optimization problem with the goal of minimizing the total system energy consumption; Step 3: Using Lyapunov optimization to handle long-term queue delay constraints, decomposing the original problem, and then using the Lagrange duality method to solve the joint optimization problem of task scheduling and resource allocation. This invention fully utilizes the flexibility and maneuverability of UAVs, effectively ensuring the security and reliability of network communication. The specific process is as follows:

[0006] The multi-UAV collaborative computing network communication system model proposed in this invention includes M associated UAVs, N auxiliary UAVs, and one ground base station. The associated UAVs and auxiliary UAVs are represented by sets. and The system involves an associated UAV continuously collecting computational tasks uploaded from ground user equipment; however, it cannot maintain a valid communication connection with the ground base station. An auxiliary UAV assists the associated UAV in task computation and offloading, and can further offload tasks to the ground base station for computation. Computational tasks can be executed in parallel by the associated UAV, auxiliary UAV, and ground base station, thereby reducing processing latency. Considering the system operates in time-slot mode, the entire optimization period is divided into τ time slots, each with a duration of T, and the entire time period can be represented as {1,...,t,...,T}. Assume the UAV is at a fixed altitude H, and the horizontal coordinates of the associated UAV, auxiliary UAV, and ground base station are respectively... and v S (t). Assuming the drone operates along a pre-designed trajectory, and The distance can be calculated as Similarly, it can be calculated Distance from ground base station

[0007] Subsequently, the link channels between the associated UAV and the auxiliary UAV, and between the auxiliary UAV and the ground base station, are modeled. A probabilistic line-of-sight (LoS) path loss model is considered to characterize the wireless channel, given by the free-space path loss model. The associated UAV continuously receives computational tasks from ground user equipment and selects the auxiliary UAV for task offloading. The binary offloading strategy from the associated UAV to the auxiliary UAV is expressed as follows: Where ym,n (t) = 1 indicates that the drone is associated in the t-th time slot. Select auxiliary drone Perform task uninstallation. Afterwards, arrive The achievable unloading rate can be expressed as

[0008] R m,n (t)=B m,n log2(1+P m,n g m,n (t) / σ 2 )

[0009] Among them, B m,n for arrive Communication link channel bandwidth, σ 2 P is the noise power. m,n For the drone's transmit power, g m,n (t) represents the channel gain, which can be calculated using the following formula.

[0010]

[0011] Where ρ0 is the channel gain when the reference distance is 1m.

[0012] After receiving tasks from associated drones, the auxiliary drone performs local calculations while also offloading some tasks to the base station, where the base station's server performs the task calculations. The information transmission rate between the base station and the base station is

[0013] R n,S (t)=B n,S log2(1+P n,S (t)g n,S (t) / σ 2 )

[0014] Among them, B n,S for The bandwidth of the communication link to the base station, P n,S (t) is Transmission power to the base station For channel gain, L n,S (t) is The path loss to the base station is calculated using the following formula.

[0015]

[0016] Among them, f c c and 'c' represent the carrier frequency and the speed of light, respectively. and These represent the additional losses caused by free-space propagation loss to line-of-sight links and non-line-of-sight links, respectively. yes The line-of-sight probability of the link to the base station is calculated by the following formula.

[0017]

[0018] Among them, ρ1, ρ2, and It is determined by the environment.

[0019] Assuming associated drones The computational task collected from a ground user facility in a time slot is represented as W. m (t), and divide it into tasks computed by its own processor. And the task of further unloading to auxiliary drones. Right now

[0020]

[0021] Meanwhile, the associated drone maintains a queue cache for user-uploaded tasks. (Definition) For the mission in The backlog of its own processor, To unload the task backlog from the queue, its update is as follows:

[0022]

[0023]

[0024] in, For tasks processed and leaving the buffer queue in time slot t, the calculation is as follows:

[0025]

[0026] Where, λ m for The number of CPU cycles required per bit for the task, f m (t) represents The computing resources. Furthermore... for The workload of unloading tasks to the auxiliary drone is calculated as follows:

[0027]

[0028] exist At that location, the data received in time slot t from... The task is These tasks are expressed in terms of quantities for assisted drone processing.

[0029]

[0030] Among them, f m,n (t) is Give Allocated computing resources. To ensure service quality for users, tasks offloaded to the auxiliary drone are no longer cached but processed promptly. To ensure the auxiliary drone offloads unprocessed tasks to the base station, its transmission power needs to meet certain requirements.

[0031]

[0032] Given that the base station is powered by a stable power grid and equipped with a powerful processor, the processing latency of the task at the base station is negligible. Furthermore, since the amount of data in the computation results is small, we consider the latency of result feedback to be negligible.

[0033] Throughout the optimization process, system energy consumption consists of transmission energy consumption and computing energy consumption. The associated UAV energy consumption includes computing energy consumption and task offloading energy consumption. Computational energy consumption is expressed as...

[0034]

[0035] Where κ is a computing power parameter that depends on the chip architecture. Furthermore, the task transfer energy consumption can be expressed as...

[0036]

[0037] The energy consumption of the assisted drone is expressed as

[0038]

[0039] Therefore, the total energy consumption of the system in each time slot is expressed as follows:

[0040]

[0041] Based on this, the joint task offloading decision y(t) and task splitting are aimed at minimizing the total system energy consumption. and computing resource allocation The optimization problem is described as follows:

[0042] P1:

[0043]

[0044]

[0045]

[0046]

[0047]

[0048]

[0049]

[0050]

[0051] Constraints 1 and 2 represent the task offloading decision of the associated drone as a binary variable and the fact that an associated drone can select at most one auxiliary drone, respectively. Constraint 3 refers to the service quality constraint of the associated drone, ensuring the effective uploading of computationally intensive tasks. R... min The minimum transmission rate per time slot; constraint 4 represents the task splitting constraint; constraint 5 represents the causal constraint, i.e., the tasks processed by the auxiliary UAV are no greater than the tasks unloaded by the associated UAV; and constraint 6 is the computational resource constraint. and These represent the available resources for the associated drone and the auxiliary drone, respectively. Constraint 7 restricts the transmission power of the associated drone and the auxiliary drone. Constraint 8 indicates that the queue backlog of the associated drone remains stable in the long term, ensuring that all computing tasks arriving at the associated drone are processed or further offloaded in a timely manner.

[0052] To address the challenging time-averaged steady-state form of the objective function and constraint 8 in problem P1, we employ Lyapunov optimization to transform the problem. Specifically, the queue backlog set is represented as... The Lyapunov function is defined as follows:

[0053]

[0054] Lyapunov drift is defined as the expectation of the change of the Lyapunov function at two consecutive time points, expressed as:

[0055] Δ(Θ(t))=E[L(Θ(t+1))-L(Θ(t))|Θ(t)]

[0056] To minimize the system's long-term average total energy consumption while ensuring queue stability, the Lyapunov drift plus penalty is defined as the Lyapunov drift plus the expected long-term average total energy consumption multiplied by the weighting parameter, i.e.

[0057] Δ V L(Θ(t))=Δ(Θ(t))+VE{E total (t)|Θ(t)}

[0058] Where V is non-negative, it is used to balance minimizing the penalty and queue stability.

[0059] Δ V The upper bound of L(Θ(t)) can be derived as follows:

[0060]

[0061] in, It is a constant that has an upper bound.

[0062] According to the Lyapunov optimization method, the defined drift plus penalty function Δ can be minimized in each time slot under constraints 1 to 7. V The upper bound of L(Θ(t)). Furthermore, we decompose the problem into two sub-problems by distinguishing the optimization variables. Sub-problem 1 optimizes the task offloading decision of the associated UAV, described as...

[0063] SP1:

[0064] stC1~C3

[0065] This subproblem is a discrete unloading-related variable optimization problem, which can be solved using an online search method.

[0066] In subproblem 2, we optimize task partitioning and computational resource allocation, referred to as SP2:

[0067] stC5~C7

[0068]

[0069] The objective function of SP2 and constraint 7 are non-convex, making it difficult to solve.

[0070] To solve the nonconvex problem SP2, we introduce auxiliary variables. Reconstruct the objective function as follows:

[0071]

[0072] And satisfy

[0073]

[0074] It can be reorganized into

[0075]

[0076] And power constraint 7 is transformed into

[0077] Subsequently, the Lagrange multipliers corresponding to constraints 5, 6, 9, and 10 are defined as ω. m,n (t), and ν n (t), construct the Lagrange function as

[0078]

[0079] Based on the Karush-Kuhn-Tucker (KKT) conditions, we can derive the optimal closed-form solution for the optimization and auxiliary variables as follows:

[0080]

[0081]

[0082]

[0083]

[0084] in, In addition, the Lagrange multipliers are updated using the sub-gradient method.

[0085] The technical method of the present invention has the following advantages:

[0086] This invention discloses a task scheduling and resource allocation method for a multi-UAV collaborative computing network. The method includes: First, establishing a multi-UAV collaborative computing network system model to fully utilize the flexibility and maneuverability of UAVs and effectively ensure the security and reliability of network communication. Second, considering the task offloading energy consumption and task computing energy consumption of UAVs, deriving the total energy consumption of the collaborative computing network system, and describing a joint task scheduling and resource allocation optimization problem with the goal of minimizing the total system energy consumption, thereby improving system energy efficiency. Finally, this invention utilizes Lyapunov optimization to handle long-term queue delay constraints, decomposes the original problem, and employs Lagrange duality to solve the joint optimization problem of task scheduling and resource allocation. Therefore, the proposed method has strong feasibility.

[0087] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0088] To more clearly illustrate the technical methods in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0089] Figure 1 This is a graph showing the relationship between the average backlog of drone queues and time.

[0090] Figure 2This is a graph showing the relationship between the total system energy consumption and the average computing resources of the UAV. Detailed Implementation

[0091] This invention proposes a method for task offloading and trajectory optimization in a space-to-ground network for direct mobile phone-to-satellite connections. The embodiments are described in detail below with reference to the accompanying drawings.

[0092] The specific implementation scenario of this invention is in a 1000×1000m area, including 2 associated UAVs, 2 auxiliary UAVs, and 1 base station. The UAVs fly at an altitude of H = 100m, with a total service time of T = 20s, a time slot length of τ = 0.1s, and the UAVs have a computing resource of 90GHz and a noise power σ. 2 = -100dBm, additive loss along the line of sight path Additive loss for non-line-of-sight paths The environmental parameters are ρ1 = 4.9 and ρ2 = 0.4, respectively.

[0093] The specific implementation steps of this invention are as follows:

[0094] 1) Establish a multi-UAV collaborative computing network system model, calculate all link data based on the distribution of associated UAVs, auxiliary UAVs and base stations in the system model, and generate the required implementation parameters.

[0095] 2) Describe the joint task scheduling and resource allocation optimization problem with the goal of minimizing the total energy consumption of the system. Based on the problem and constraints, use Lyapunov optimization to distinguish and decouple the variables of the problem.

[0096] 3) Establish subproblem 1 and solve it using an online search method to find the optimal associated UAV unloading decision.

[0097] 4) For subproblem 2, auxiliary variables are introduced to reconstruct the function, construct the Lagrangian function, and derive the optimal closed-form solution of the optimization variables and auxiliary variables according to the Karush-Kuhn-Tucker (KKT) conditions, and find the optimal task partitioning and resource allocation.

[0098] Figure 1 The graph shows the relationship between the average backlog of the UAV queue and time. It can be seen that the queue backlog of the proposed method is significantly lower than that of other benchmark methods. This is because in the proposed method, the UAV can dynamically adjust task scheduling and resource allocation according to the real-time task queue, so as to achieve the optimal task scheduling and resource allocation decision and thus reduce task backlog.

[0099] Figure 2The paper presents a comparison of the total system energy consumption of the proposed algorithm with other comparative algorithms in each time slot as the average computing resources of the UAV change. The total system energy consumption decreases with the increase of UAV computing resources because when computing resources increase, tasks can be computed locally and in a timely manner more often than they can be transported and offloaded. Experimental results also verify that the proposed method outperforms other methods.

[0100] The above description is merely a preferred embodiment of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this disclosure is not limited to technical methods formed by specific combinations of the above-described technical features, but should also cover other technical methods formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the inventive concept. For example, technical methods formed by substituting the above-described features with (but not limited to) technical features disclosed in this disclosure that have similar functions.

Claims

1. A method for task scheduling and resource allocation in a multi-UAV collaborative computing network, the steps of which are as follows: Step 1: Establish a multi-UAV collaborative computing network system model, considering the task offloading energy consumption and task computing energy consumption of the UAVs, and derive the total energy consumption of the collaborative computing network system; the multi-UAV collaborative computing network system model includes Connecting drones, The system consists of an auxiliary drone and a base station. The associated drone and the auxiliary drone are respectively set up using a set. and It is stated that the associated drone continuously collects computing tasks uploaded from ground user equipment, but it cannot maintain an effective communication connection with the ground base station. The auxiliary drone assists the associated drone in performing task calculations and offloading, and can further offload tasks to the ground base station for calculation. The entire optimization period is divided into equal parts. There are 1 time slot, and the duration of each time slot is 1. The entire time period can be represented as Throughout the optimization process, system energy consumption consists of transmission energy consumption and computing energy consumption. The energy consumption of associated UAVs includes computing energy consumption and task offloading energy consumption. The computing energy consumption of associated UAVs is expressed as follows: in The computing power parameters depend on the chip architecture. express Computing resources for The number of CPU cycles required per bit for the task. For the mission in The backlog of the processor itself and the energy consumption for task transmission can be expressed as: in To correlate the drone binary offloading strategy, To correlate the drone's transmission power, To unload the backlog of tasks in the queue, for arrive The achievable unloading rate, and the energy consumption of the auxiliary drone, are expressed as: in, for Give Allocated computing resources To support the UAV's launch power, the total system energy consumption for each time slot is expressed as follows: ； Step 2: Describe the joint task scheduling and resource allocation optimization problem with the goal of minimizing total system energy consumption; in Step 2, the joint task offloading decision is aimed at minimizing total system energy consumption. Task segmentation and computing resource allocation The optimization problem is described as follows: in, To ensure the minimum transmission rate required for intensive task transmission, To associate drones A computing task collected from ground user equipment in a time slot, The amount of work to be computed by its own processor. To further offload the workload to auxiliary drones, To assist drones in handling a larger workload, For received from The workload, and Available resources for associated drones and auxiliary drones, respectively. and These are the maximum transmit powers of the associated drone and the auxiliary drone, respectively; Step 3: Use Lyapunov optimization and Lagrange duality to solve the joint optimization problem of task scheduling and resource allocation; the solution steps in step 3 are limited to: considering the objective function and constraints in problem P1. The problem is transformed using Lyapunov optimization in its time-averaged steady-state form. Then, by distinguishing the optimization variables, the problem is decomposed into two subproblems. For the first subproblem, an online search method is used to solve it. For the second subproblem, auxiliary variables are introduced to reconstruct the function and construct the Lagrangian function. Based on the Karush-Kuhn-Tucker (KKT) conditions, the optimal closed-form solution of the optimization variables and auxiliary variables is derived.

2. The task scheduling and resource allocation method for a multi-UAV collaborative computing network according to claim 1, wherein the method utilizes Lyapunov optimization to handle long-term queue delay constraints, decomposes the original problem, and then uses the Lagrange dual method to solve the joint optimization problem of task scheduling and resource allocation.

Images