Unmanned aerial vehicle cluster task optimization method based on master-slave game and steady-state matching

Abstract

The invention relates to an unmanned aerial vehicle cluster task optimization method based on a master-slave game and steady-state matching, and the method comprises the steps: 1, collecting the processed task information when an unmanned aerial vehicle carries out the task position deployment, specifically, the information comprises the position information of a user and the size of a user task; step 2, clustering and dividing all users based on user task information, and performing task unloading on each user cluster by one unmanned aerial vehicle; 3, after the information is collected, the unmanned aerial vehicle performs a master-slave game with each user task, the game aims at optimizing the energy consumption of the unmanned aerial vehicle, and finally a Nash equilibrium state is achieved; 4, when the unmanned aerial vehicle cannot complete all tasks in the user cluster, performing descending order selection on the tasks to ensure optimization of energy consumption of the unmanned aerial vehicle; according to the method, the game theory and the optimization strategy are combined, and task allocation and energy consumption management of the unmanned aerial vehicle are reasonably allocated, so that the energy efficiency of the unmanned aerial vehicle during task execution is improved, the overall energy consumption is reduced, and more efficient unmanned aerial vehicle operation is realized.

Description

Technical Field

[0001] The invention relates to a method for optimizing unmanned aerial vehicle cluster tasks based on master-slave game and steady-state matching, and belongs to the technical field of computer science and engineering. Background Art

[0002] With the popularization of 5G technology and the emerging development of the Internet of Things (IoT), the existing traditional local computing and cloud computing methods cannot support user business needs. Mobile Edge Computing (MEC) places computing resources at the edge of the network, making data processing closer to end users and reducing the delay of data transmission. Therefore, it is more suitable for task processing of devices in the field of IoT. Compared with traditional MEC, drones have high mobility and flexibility, and can assist in task offloading by carrying related equipment to support time-sensitive data services. Therefore, the optimization problem of drone-assisted task offloading has been increasingly discussed and studied by academics. Drone-assisted applications are widely used in disaster response, environmental monitoring, wireless communications and other fields. Especially in disaster areas or sparsely populated areas, due to the destruction or lack of ground communication infrastructure, communication needs cannot be met. Drones are quickly deployed to act as mobile base stations to meet temporary communication and emergency communication needs. Drones are low-cost, highly mobile, and flexible to deploy. Temporary base stations can be easily built in the air to meet post-disaster emergency dispatch requirements. At the same time, by carrying sensors and other related equipment, drones can be used to draw post-disaster maps and search for personnel. With the continuous development of future technologies, drones combined with emerging artificial intelligence technologies will play an increasingly important role in the Internet of Things where everything is connected.

[0003] In recent years, with the continuous advancement of drone technology, drone swarms have shown great potential in a variety of application scenarios. However, existing drone swarm energy consumption optimization methods face many challenges in practical applications, including low task allocation efficiency, insufficient energy management, and poor cluster coordination. Traditional energy consumption optimization methods often focus on the energy consumption of a single drone, while ignoring the mutual influence between drones in the cluster, resulting in the failure to effectively improve the overall energy efficiency. In the operation of drone swarms, the impact of task complexity on energy consumption cannot be ignored. With the increase in the number of tasks, drones will face higher energy consumption requirements when performing tasks. In addition, the master-slave game theory provides a new optimization idea for drone swarms. By combining task allocation with energy consumption management, energy consumption can be reduced more effectively. The steady-state matching strategy helps to improve the collaboration efficiency between drones in the cluster. By reasonably allocating the tasks and resources of drones, the energy consumption of the cluster can be effectively reduced during the execution of tasks. Therefore, it is of great practical significance to develop a drone swarm energy consumption optimization method based on master-slave game and steady-state matching to solve problems such as improper energy management, difficult task coordination, and low cluster operation efficiency. This method will provide a more efficient and sustainable energy management solution for drone swarms when performing complex tasks. Summary of the invention

[0004] In order to overcome the shortcomings of the prior art, the present invention provides a drone cluster task optimization method based on master-slave game and steady-state matching, and all user tasks can be decomposed into different proportions for local and drone offloading respectively.

[0005] A method for optimizing UAV cluster tasks based on master-slave game and steady-state matching includes the following steps:

[0006] Step 1: When the drone is deployed at the mission location, collect the processed mission information, including the user's location information and the user's mission size;

[0007] Step 2: Based on user task information, all users are divided into clusters, and each user cluster is assigned a UAV for task offloading;

[0008] Step 3: After collecting information, the drone plays a master-slave game with each user task, the game aims to optimize its own energy consumption and eventually reach a Nash equilibrium state;

[0009] Step 4: When the drone cannot complete all the tasks in the user cluster, the tasks are selected in descending order to ensure the optimization of the drone’s energy consumption;

[0010] Step 5: Based on the heterogeneity of drone power, drones are matched with the overall situation of user clusters to achieve overall optimization of the energy consumption of the drone cluster;

[0011] Step 6: After the task is assigned, the drone moves to the designated location to unload the task, and returns to the charging facility to charge when the task is completed or the battery is low, preparing for the next round of tasks.

[0012] In the step 2, the user clustering scheme is based on the improved K-Means algorithm, and the K-Means algorithm is based on the weighted distance between the user and the drone, where the distance is weighted based on the user's location and the amount of tasks.

[0013] The master-slave game in step 3 adopts Stackelberg game, and the game function is composed of user energy consumption and benefit. and drone energy benefits Joint decision.

[0014] In step 4, tasks are selected in descending order using dynamic programming based on the 01 knapsack algorithm to select user tasks under the premise of maximizing task benefits and sufficient energy consumption of the drone.

[0015] In step 5, the overall situation of the drone and the user cluster is matched and a stable matching is performed based on the improved Gale-Shapley algorithm.

[0016] The user benefits and drone revenue The formula is as follows:

[0017] (1)

[0018] (2)

[0019] in Represents the drone number, Represents the user number, The proportion of user tasks offloaded to drones, is the user's task energy cost, is the cost of all local computations of the user task, is the cost required for drone calculation, is the time penalty cost of the drone, Transmit energy consumption costs for user tasks, Represents the energy consumption cost of the drone computing task, Represents the energy cost of the drone's flight movement, It represents the energy consumption satisfaction of the UAV with the user's task.

[0020] The optimization objective in the master-slave game is solved by the gradient descent method, and finally the optimal task offloading ratio of each drone and the energy consumption cost of the user task are obtained.

[0021] When the UAV has limited battery power, a dynamic programming method is used to solve task allocation in order to maximize the remaining battery power of the UAV and the efficiency of task processing.

[0022] The UAV's task allocation strategy optimizes the task selection within the user cluster to ensure the rationality of task allocation and the optimization of cluster energy efficiency.

[0023] An energy consumption optimization system applied to a drone cluster, comprising:

[0024] Task information collection module, used to collect relevant data such as user location information and task size;

[0025] The user clustering module is used to cluster users according to task information and assign a drone to each user for task offloading;

[0026] Game optimization module, used for master-slave game optimization between drones and users, and energy consumption benefits are analyzed through Stackelberg game;

[0027] The task selection module is used to select user tasks according to the 01 backpack algorithm to optimize the energy consumption of the drone;

[0028] Matching module, used to perform stable matching of drones and user clusters using the Gale-Shapley algorithm

[0029] Compared with the prior art, the present invention has the following beneficial effects:

[0030] In order to increase the overall energy consumption optimization of the drone cluster, the present invention optimizes the energy consumption benefits of tasks by subdividing user tasks into the proportion of local processing and drone offloading, and the system intelligently allocates resources based on the drone's own power and task requirements. This strategy not only improves the efficiency of task execution, but also significantly reduces the energy consumption of drones.

[0031] The present invention proposes to optimize the energy consumption of Stackelberg games in master-slave games through the energy consumption functions of both parties. In traditional technologies, although effective energy optimization strategies have been developed for multiple drones under the constraints of battery capacity, in-depth analysis and visualization of energy consumption benefits are usually lacking. By considering energy consumption as part of benefits, drone clusters and users can build a more dynamic and adaptive game model, so that drones can make intelligent decisions based on real-time energy consumption when performing tasks. This method can not only improve the energy efficiency of drones, but also, through the analysis of game theory in a multi-party competitive environment, enable drones to achieve better energy consumption allocation under limited resources. Through the visualized comparison of energy consumption benefits, drone clusters can more intuitively understand the impact of their respective strategies, thereby achieving more efficient collaboration and competition. This innovation provides a new perspective for autonomous decision-making of drones and helps promote the application of intelligent unmanned systems in complex environments.

[0032] The present invention proposes to use the Gale-Shapley algorithm to ensure stable matching between users and drones, thereby ensuring efficient use of resources. Although existing research can achieve Nash equilibrium for multiple drones, it rarely considers the problem of stable matching of benefits between multiple UEs and multiple drone task allocation. By reasonably allocating tasks, idleness and overload of drones are avoided, the working efficiency of the entire cluster is improved, and it is ensured that each drone can perform its tasks in the best state. These advantages jointly promote the overall performance of drone clusters when performing complex tasks, and enhance their competitiveness and feasibility in practical applications. Compared with traditional matching algorithms, the present invention first uses improved K-means to complete user task clustering, and has the problem of high robustness in the selection of user tasks by drones and the matching of user clusters. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0034] Figure 1 It is a flow chart of the UAV cluster task optimization method based on master-slave game and steady-state matching of the present invention;

[0035] Figure 2 This is a diagram illustrating the model scenario structure of the UAV cluster task optimization method based on master-slave game and steady-state matching of the present invention;

[0036] Figure 3This is a diagram illustrating the master-slave game phase of the UAV cluster task optimization method based on master-slave game and steady-state matching of the present invention. DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0038] The flow chart of the present invention is as follows Figure 1 shown.

[0039] See also Figure 2 ,This example is conducted in a limited power scenario of user task ,allocation under a multi-UAV cluster.

[0040] The drone charging facility first receives the parameter information of the user's task and analyzes it, and optimizes energy consumption through three stages: drone location deployment, drone user master-slave game, and stable matching of drones and user clusters. Finally, the drone uses the strategy to dispatch the drone for deployment and task unloading. If the drone is low on power or the task is completed, it will return to charge and wait for the next round of instructions.

[0041] A method for optimizing UAV cluster tasks based on master-slave game and steady-state matching includes the following steps:

[0042] Step 1: When the drone is deployed at the mission location, collect the processed mission information, including the user's location information and the user's mission size;

[0043] Step 2: Based on user task information, all users are clustered, and each user cluster is unloaded by a drone; in the step 2, the user clustering scheme is based on the improved K-Means algorithm, and the K-Means algorithm is based on the weighted distance between the user and the drone, where the distance is weighted based on the user location and the task amount.

[0044] The initial location deployment selection of the drone and the user situation responsible for the location are completed based on the improved K-means algorithm. The specific steps are as follows:

[0045] S2.1: Initial generation of the estimated deployment location of the drone And normalize the task size for subsequent weighted processing.

[0046] S2.2: Calculate all users to all initial positions The distance between them is calculated and the user is assigned to the closest In the user clustering

[0047] S2.3: By formula Regenerate the drone deployment position, where represents the location information of drone k, represents the location information of user n, represents the normalized parameter of the user task, Represents user n belonging to the deployment Unload the task to the UAV at the specified location.

[0048] S2.4: Repeat S2.2 and S2.3 until the deployment positions of all drones remain unchanged.

[0049] Step 3: After collecting information, the drone plays a master-slave game with each user task, the game aims to optimize its own energy consumption and eventually reach a Nash equilibrium state;

[0050] The master-slave game in step 3 adopts Stackelberg game, and the game function is composed of user energy consumption and benefit. and drone energy benefits Joint decision.

[0051] The user benefits and drone revenue The formula is as follows:

[0052] (1)

[0053] (2)

[0054] in Represents the drone number, Represents the user number, The proportion of user tasks offloaded to drones, is the user's task energy cost, is the cost of all local computations of the user task, is the cost required for drone calculation, is the time penalty cost of the drone, Transmit energy consumption costs for user tasks, Represents the energy consumption cost of the drone computing task, Represents the energy cost of the drone's flight movement, It represents the energy consumption satisfaction of the UAV with the user's task.

[0055] (3)

[0056] In formula (3), is the CPU frequency required by task n, is the CPU frequency when the user processes tasks locally, is the unit energy consumption cost of the user, is the time required to complete the task. is the maximum allowable delay of the task, is the time penalty function, is the size of the task being transferred, is the user's transmission power, is the rate at which user tasks are transmitted, and its specific formula is calculated by the Shannon formula:

[0057] (4)

[0058] in is the signal white noise, is the channel bandwidth, Indicates the channel gain status, represents the total interference from other user channels during user n's transmission.

[0059] The energy consumption benefit function of the drone is The specific parameter formula is as follows:

[0060] (5)

[0061] in, It represents the mission benefit that can be brought by the average energy consumption of UAV m in future missions. represents the unit energy consumption cost of UAV m, Indicates the current remaining battery capacity of drone m, is the amount of power consumed by the drone’s computing. It represents the minimum energy consumption required for UAV m to safely fly to the deployment location and return. is the task satisfaction coefficient, which is a positive constant. is the computing power of the drone, is the CPU frequency when the drone is computing tasks, is the flight power of the drone, It is the time required for the drone to fly to the designated deployment location.

[0062] In order to obtain the best response function of formula (1) and formula (3), the present invention performs the following proof based on the Stackelberg game of the master-slave game to prove that there is a Nash equilibrium. The goal is to establish and The value of to prove the necessary conditions for the existence of its maximum value:

[0063] S3.1: Yes Task offloading ratio Take the second-order derivative

[0064] (6)

[0065] Where A, B, C are and The specific formula is as follows:

[0066] (7)

[0067] By observing and calculating, we can find that its second-order derivative is always less than 0, which satisfies the necessary conditions for Nash equilibrium. At the same time, the strategy set of each drone is non-empty, convex and compact. Each drone has a unique task offloading strategy in response to the user's strategy. Proof There is a unique Nash equilibrium value.

[0068] S3.2: Let right Find the first-order derivative and make it equal to 0, and we get and relationship.

[0069] (8)

[0070] Where D is and The specific expression of irrelevant parameter set is:

[0071] (9)

[0072] S3.3: Substitution , and find its second-order derivative.

[0073] (10) (11)

[0074] By observing and calculating, we can also find that its second-order derivative is always less than 0, which satisfies the necessary conditions for Nash equilibrium. At the same time, the strategy set of each drone is non-empty, convex and compact. Each drone has a unique task offloading strategy in response to the user's strategy. Proof There is a unique Nash equilibrium value.

[0075] Therefore, there is a unique The set solution satisfies formula (1) and (2) and obtains the maximum value at the same time. In order to solve its maximum value, the present invention uses the gradient descent method to train and solve it. The specific pseudo code is as follows:

[0076]

[0077] Step 4: When the drone cannot complete all the tasks in the user cluster, the tasks are selected in descending order to ensure the optimization of the drone’s energy consumption;

[0078] In step 4, tasks are selected in descending order using dynamic programming based on the 01 knapsack algorithm to select user tasks under the premise of maximizing task benefits and sufficient energy consumption of the drone.

[0079] Step 5: Based on the heterogeneity of drone power, drones are matched with the overall situation of user clusters to achieve overall optimization of the energy consumption of the drone cluster;

[0080] In step 5, the overall situation of the drone and the user cluster is matched and a stable matching is performed based on the improved Gale-Shapley algorithm.

[0081] Step 6: After the task is assigned, the drone moves to the designated location to unload the task, and returns to the charging facility to charge when the task is completed or the battery is low, preparing for the next round of tasks.

[0082] The present invention discloses an optimization algorithm for competition between drone groups and user tasks, aiming to achieve global optimal energy consumption management in a drone networking environment. In the traditional drone and user task game model, there is an optimal strategy choice between each drone and the user task. However, considering that users may choose to offload tasks to drones for the purpose of reducing their own energy consumption, this in turn triggers competition between drones and users; at the same time, in a drone networking environment, multiple drones may also compete for the optimal set of user tasks in order to maximize their own energy efficiency. Therefore, the present invention proposes a collaborative optimization algorithm for the above-mentioned competitive behavior, which ultimately achieves the goal of global optimal energy consumption by realizing collaborative game between drone groups and user tasks.

[0083] Before selecting a drone to be deployed in a user cluster, the present invention takes into account that due to the limited power of the drone, when the amount of user-requested task offloading in a certain user cluster is too large, it may not be possible to fully cover all user tasks. Therefore, before deciding to deploy a drone to a certain user cluster, it is necessary to first consider which user tasks in the cluster can be processed to maximize the energy utilization of the drone itself. To this end, the present invention models this problem as a classic dynamic programming problem. Considering that the number of tasks in each user cluster is limited and does not exceed a certain number, the present invention adopts the solution method of the 01 knapsack problem, based on the benefit value of the task. , Energy consumption of drone processing tasks And the remaining battery power of the drone The optimal task allocation strategy is obtained through recursive calculation based on three state parameters.

[0084] In addition, the present invention also notes that there is a classic bilateral matching problem in the task matching problem between drone clusters and user clusters. In order to obtain the optimal matching result between drone clusters and user clusters, the present invention designs a stable matching algorithm based on the Gale-Shapley algorithm to ensure that the needs of all parties can be met, thereby further optimizing energy consumption management and task allocation efficiency. The specific pseudo code is as follows:

[0085]

[0086]

[0087] In summary, the present invention proposes a new method for optimizing the energy consumption of drone clusters by combining game theory and steady-state matching algorithm. This method not only optimizes the energy efficiency in the task allocation process, but also improves the collaborative efficiency and power management capabilities in the task execution process, which can significantly reduce cluster energy consumption, improve task execution efficiency, and ensure the optimal allocation of resources. This technical solution is widely applicable to disaster response, environmental monitoring and other fields, and can also be applied to other multi-drone collaborative tasks.

[0088] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. For those skilled in the art, various changes, modifications, substitutions and variations of these embodiments are made without departing from the principles and spirit of the present invention, and still fall within the scope of protection of the present invention.

Claims

1. A method for optimizing UAV cluster tasks based on master-slave game and steady-state matching, characterized by: The following steps are involved: Step 1: When the UAV is deployed at the mission location, collect the processed mission information, including the user's location information and the user's mission size; Step 2: Based on user task information, all users are divided into clusters, and each user cluster is assigned a UAV for task offloading; Step 3: After collecting information, the drone plays a master-slave game with each user task, the game aims to optimize its own energy consumption and eventually reach a Nash equilibrium state; Step 4: When the drone cannot complete all the tasks in the user cluster, the tasks are selected in descending order to ensure the optimization of the drone’s energy consumption; Step 5: Based on the heterogeneity of drone power, the overall situation of drone and user clustering is matched to achieve the overall optimization of energy consumption of drone clusters; Step 6: After the task is assigned, the drone moves to the designated location to unload the task, and returns to the charging facility to charge when the task is completed or the battery is low, preparing for the next round of tasks.

2. The method for optimizing the task of a drone cluster based on master-slave game and steady-state matching according to claim 1 is characterized in that: In the step 2, the user clustering scheme is based on the improved K-Means algorithm, and the K-Means algorithm is based on the weighted distance between the user and the drone, where the distance is weighted based on the user's location and the amount of tasks.

3. The method for optimizing the task of a drone cluster based on master-slave game and steady-state matching according to claim 1 is characterized in that: The master-slave game in step 3 adopts Stackelberg game, and the game function is composed of user energy consumption and benefit. and drone energy benefits Joint decision.

4. The method for optimizing the task of a drone cluster based on master-slave game and steady-state matching according to claim 1 is characterized in that: In the step 4, the tasks are selected in descending order using dynamic programming based on the 01 knapsack algorithm to select user tasks under the premise of maximizing the mission benefits and sufficient energy consumption of the drone.

5. The method for optimizing the task of a drone cluster based on master-slave game and steady-state matching according to claim 1 is characterized in that: In step 5, the overall situation of the drone and the user cluster is matched and a stable matching is performed based on the improved Gale-Shapley algorithm.

6. The method for optimizing UAV cluster tasks based on master-slave game and steady-state matching according to claim 3 is characterized in that: The user's energy consumption benefits and drone energy benefits The formula is as follows: （1） （2） in: Represents the drone number, Represents the user number, The proportion of user tasks offloaded to drones, is the user's task energy cost, is the cost of all local computations of the user task, is the cost required for drone calculation, is the time penalty cost of the drone, Transmit energy consumption costs for user tasks, Represents the energy consumption cost of the drone computing task, Represents the energy cost of the drone's flight movement, It represents the energy consumption satisfaction of the UAV with the user's task.

Images