Unmanned aerial vehicle multi-target task scheduling method based on distributed cooperative game

Abstract

The invention discloses an unmanned aerial vehicle multi-target task scheduling method based on a distributed cooperative game, and the method comprises the following steps: S1, defining the state information of an unmanned aerial vehicle, and generating basic data; s2, constructing a task competition game model; s3, individual task optimization is carried out, and an improved artificial protozoa optimization algorithm is adopted; s4, performing global task scheduling optimization by adopting an improved egret group optimization algorithm; s5, performing adaptive adjustment in a task execution process; and S6, completing the task scheduling method. According to the method, the resource utilization rate and global task income of the unmanned aerial vehicle are remarkably improved, and the method can be widely applied to unmanned aerial vehicle cluster task management scenes of multiple industries such as logistics distribution, disaster rescue, environment monitoring and military reconnaissance.

Description

Technical Field

[0001] The present invention relates to the technical field of multi-target task scheduling for unmanned aerial vehicles, and particularly to a multi-target task scheduling method for unmanned aerial vehicles based on distributed cooperative game. Background Art

[0002] In modern unmanned aerial vehicle (UAV) cluster systems, multi-target task scheduling is a key research direction, which involves the optimal allocation of multiple UAVs to perform multiple tasks in a complex environment. With the rapid development of UAV technology, UAV clusters are gradually undertaking more and more tasks in fields such as military reconnaissance, environmental monitoring, emergency rescue, and logistics transportation. The task scheduling of UAVs not only needs to consider factors such as task timeliness, priority, and resource allocation, but also needs to be intelligently adjusted in a dynamic environment to adapt to changes in task requirements. However, there are still many challenges and deficiencies in the existing technology for multi-target task scheduling of UAVs, resulting in problems such as low scheduling efficiency, limited task completion rate, and serious waste of UAV resources.

[0003] Currently, UAV task scheduling methods are mainly divided into two categories: centralized scheduling and distributed scheduling. Centralized scheduling relies on a central controller to uniformly plan tasks and allocate them to each UAV. Although this method can optimize globally, it has problems such as large communication delay, heavy computational burden, and poor system robustness. Once the central controller fails, the entire system will face the risk of paralysis. In addition, the centralized method is difficult to adapt to dynamic environmental changes. When situations such as insufficient battery power, task failure, or environmental mutation occur to UAVs, it is unable to quickly adjust task allocation, resulting in a decrease in task completion rate. In contrast, distributed scheduling methods improve the flexibility and adaptability of the system by allowing UAVs to make autonomous decisions on task allocation, avoiding the single-point failure problem of centralized scheduling. However, existing distributed scheduling methods usually rely on simple heuristic algorithms or rule-based strategies, lacking in-depth modeling of factors such as task competition, cooperation between UAVs, and adaptation to dynamic environments, and it is difficult to achieve optimal task allocation.

[0004] In terms of optimization methods for UAV task scheduling, existing research mostly uses intelligent optimization algorithms such as genetic algorithms, ant colony optimization, and particle swarm optimization. These algorithms have good convergence and global search capabilities in solving complex optimization problems, but there are still certain limitations in multi-target task scheduling problems. For example, traditional optimization algorithms are prone to falling into local optima in high-dimensional task allocation problems, and the computational complexity is relatively high, making it difficult to meet the high-efficiency scheduling requirements of UAV clusters in real-time environments. In addition, most optimization algorithms only optimize the task selection of individual UAVs, lacking a global coordination mechanism, which easily leads to the situation that some tasks are not executed while there are multiple UAVs competing for other tasks, resulting in unbalanced task load and a decline in the overall system benefit.

[0005] In recent years, game theory has gradually been introduced into the field of UAV task scheduling. By modeling the competitive and cooperative relationships among UAVs, optimal task allocation is achieved. However, existing task scheduling methods based on game theory usually adopt static game models and fail to fully consider the dynamic decision-making process of UAVs and the real-time changes in the task environment. In addition, most game methods only focus on maximizing individual benefits while ignoring the balance of global task allocation, resulting in the overall system benefit being difficult to reach the optimal. At the same time, the computational complexity of existing game methods is relatively high, and the computational cost is too high in large-scale UAV cluster task scheduling, affecting the real-time performance and scalability of the system.

[0006] Aiming at the deficiencies of existing UAV multi-objective task scheduling methods, the present invention proposes a UAV multi-objective task scheduling method based on distributed cooperative game, constructs a task competition game model, combines an improved artificial protozoa optimization algorithm and an egret flock optimization algorithm to achieve individual task optimization and global task collaborative optimization of UAVs. The present invention enables UAVs to autonomously make decisions on task selection based on local information by constructing a distributed task competition mechanism based on game theory, and reduces task conflicts through task negotiation and benefit equilibrium strategies, improving the fairness and global benefit of task scheduling. In order to further enhance the intelligence and adaptability of task scheduling, the present invention improves the artificial protozoa optimization algorithm, introduces a reinforcement dormancy strategy and an adaptive evolutionary reproduction strategy, enabling UAVs to dynamically adjust task selection strategies according to task competition situations and energy consumption states, and improving the success rate of task execution. At the same time, to solve the problem of uneven global task allocation, the present invention constructs a global task fitness function based on the task competition game model in combination with an improved egret flock optimization algorithm to measure the adaptability of the UAV group to tasks, and constructs a global task scheduling priority through a task urgency weighting factor to ensure that high-priority tasks are executed first. In addition, the present invention proposes an adaptive task adjustment mechanism based on evolutionary game. During the task execution process, if a UAV encounters environmental changes (such as insufficient power, task failure, etc.), the task adjustment strategy is automatically triggered, and the task allocation is re-optimized based on the global benefit function until the system reaches the optimal equilibrium state.

[0007] In summary, the existing multi-objective task scheduling methods for UAVs have problems such as poor robustness of centralized methods, lack of task collaboration in distributed methods, high computational complexity of traditional optimization algorithms, and insufficient consideration of global optimization in game methods, making it difficult to meet the requirements of UAV task scheduling in complex dynamic environments. The present invention constructs a distributed cooperative game model and combines an improved intelligent optimization algorithm to propose a more adaptable and efficient multi-objective task scheduling method for UAVs, achieving balanced optimization of task allocation, intelligent avoidance of task conflicts, autonomous optimization of UAV task selection, and task adaptive adjustment in dynamic environments, significantly improving the task completion rate and system revenue of UAV swarms.

[0008] Therefore, how to provide a multi-objective task scheduling method for UAVs based on distributed cooperative game is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0009] An object of the present invention is to propose a multi-objective task scheduling method for UAVs based on distributed cooperative game, which has the advantages of high task scheduling balance, strong adaptability to dynamic environments, high task completion rate, and high utilization rate of UAV resources.

[0010] A multi-objective task scheduling method for UAVs based on distributed cooperative game according to an embodiment of the present invention includes the following steps:

[0011] S1. Construct a UAV task scheduling model, define a UAV swarm and a task set, set task attributes, initialize the state information of UAVs, and generate basic data;

[0012] S2. Based on the distributed cooperative game mechanism, define the strategy set, revenue function, and game rules of UAV agents for the basic data, and construct a task competition game model as the core basis for UAV task allocation and strategy interaction;

[0013] S3. In the task competition game model, each UAV makes individual decisions through an improved artificial protozoa optimization algorithm based on local information, evaluates the revenue of each task using the revenue function, selects the optimal task according to the strategy set, strengthens the dormant strategy, and the adaptive evolution and reproduction strategy;

[0014] S4. The task competition game model further guides the task negotiation between UAVs. Through the improved egret flock optimization algorithm, global task allocation is carried out according to the game rules. The global task fitness function is used to measure the adaptability of the UAV swarm to tasks, and a task urgency weighting factor is introduced to construct the global task scheduling priority; during the global task optimization process, a task allocation adjustment equation is constructed, the global task revenue function of the UAV swarm is set, and the UAV task scheduling scheme is optimized in combination with the task conflict factor;

[0015] S5. During the task execution, based on the game rules, strategy set, and payoff function defined in the task competition game model, if the environment changes, the UAV triggers the adaptive adjustment mechanism, re-evaluates according to the payoff function, and adjusts the task allocation according to the strategy set until the system reaches the optimal equilibrium state again;

[0016] S6. After the task execution is completed, based on the long-term payoff feedback, analyze the UAV strategy adjustment data using the task competition game model, and obtain a complete task scheduling method by updating the payoff function parameters.

[0017] Optionally, the S2 specifically includes:

[0018] S21. Define the UAV agent set U and task set T, establish a distributed game framework for UAV task scheduling, and set each UAV U i as a game player, the strategy set Π is used to select the task allocation scheme, and the individual UAV U i executes the task T j 's payoff function R(U i , T j ) is used to calculate the optimal strategy of each UAV, and define the task competition game model G=(U, T, Π, R);

[0019] S22. Set the basic rules of the task competition game. Each UAV U i can only obtain the task status information of itself and neighboring UAVs to make local optimal strategy selections. If two or more UAVs compete for the same task T j , calculate the optimal allocation strategy according to the payoff function. The UAV can dynamically adjust the strategy during the task execution, and the payoff function gradually converges to the game equilibrium state;

[0020] S23. During the task competition process, set the task selection strategy, and finally construct the UAV task competition game model. The UAV makes autonomous decisions based on the distributed cooperative game mechanism during the task allocation process to achieve the equilibrium optimization of the task allocation.

[0021] Optionally, the S3 specifically includes:

[0022] S31. Based on the task competition game model G, each UAV U i makes individual decisions based on local information, and uses the improved artificial protozoa optimization algorithm for task selection. The improvement includes strengthening the dormancy strategy and the adaptive evolutionary reproduction strategy;

[0023] S32. In the UAV U iWhen performing task selection, if facing problems such as task resource competition, insufficient power, or communication obstruction, execute the enhanced sleep strategy, calculate the optimal timing for the UAV to enter the sleep state, and optimize the sleep decision through the following two steps, specifically including:

[0024] S321. Dynamically adjust the sleep threshold based on energy consumption, and define the UAV energy threshold function:

[0025]

[0026] where E min is the minimum energy requirement for UAV task execution, E max is the maximum energy consumption that the UAV can tolerate, is the average energy consumption of the UAV's historical task execution, σ(E,t) is the energy fluctuation function, which calculates the uncertainty of energy consumption in the task environment, and γ1, γ2, γ3, γ4 are adjustment coefficients that satisfy γ1 + γ2 + γ3 + γ4 = 1;

[0027] S322. Strengthen the sleep strategy based on the task competition intensity, and define the task competition factor:

[0028]

[0029] where n conf (T j ,t) represents the number of UAVs competing for task T j at time t, and n avail (T j ,t) represents the number of remaining optional tasks at time t;

[0030] S33. Set the sleep probability function of the UAV:

[0031] When E(U i ,t) < E th (U i ,t), P sleep (U i ,t) = 1;

[0032] where P sleep (U i ,t) is the probability that the UAV enters the sleep state, τ is the temperature parameter, and when the UAV energy E(U i ,t) is lower than the threshold E th (U i ,t), the sleep probability is set to 1;

[0033] S34. During the optimization of UAV mission selection, an adaptive evolutionary reproduction strategy is introduced. Based on the historical mission execution data of the UAV, the mission allocation strategy is optimized, and the individual evolution equation is set:

[0034]

[0035] Among them, is the strategy after the UAV selects tasks in the (t + 1)-th round. δ is the evolution adjustment factor, which controls the strategy update step size. is the gradient of the mission revenue function with respect to the strategy. ρ is the empirical learning factor, which optimizes the strategy based on historical mission execution experience. is the mission environment adaptation factor, which dynamically adjusts the strategy according to time. is the intelligent evolution function, which adjusts the strategy mutation direction according to the mission execution success rate and mission difficulty, enabling the UAV to evolve a more stable mission selection strategy. ζ·(E th (U i ,t)-E(U i ,t)) is the influencing term that enables the UAV to evolve a scheduling strategy that is more inclined to low-energy-consuming tasks in a low-energy state.

[0036] S35. A mission selection mechanism based on dynamic energy constraints is introduced. UAVs with lower energy will automatically reduce the revenue weight of high-energy-consuming tasks and preferentially select low-energy-consuming tasks:

[0037] R adj (U i ,T j )=R(U i ,T j )-λ E ·(E(U i ,t)-E th (U i ,t));

[0038] Among them, R(U i ,T j ) is the revenue function for the individual UAV U i to execute the mission T j . R adj (U i ,T j ) is the adjusted revenue function for the individual UAV U i to execute the mission T j . λ E is the energy adjustment factor;

[0039] S36. During the mission execution, the drone dynamically adjusts its task allocation plan by combining the enhanced sleep strategy and the adaptive evolutionary reproduction strategy. If the drone is in a dormant state for a long time, it re-adjusts its task selection strategy according to the task benefit function. During the task competition, if two or more drones compete for the same task T j a competition is carried out, and the optimal task allocation is calculated based on the game rules in the task competition game model, and the task priority is dynamically adjusted.

[0040] Optionally, the S4 specifically includes:

[0041] S41. Based on the individual task selection of the improved artificial protozoa optimization algorithm, the improved egret flock optimization algorithm is used to optimize the global task allocation;

[0042] S42. During the task competition game process, a global task fitness evaluation function is defined to measure the adaptability of the entire drone cluster to the task, and the global fitness function is set:

[0043]

[0044] where F global (T, t) is the total fitness of the drone swarm to execute the task set T at time t, and F task (U i , T j , t) is the fitness of the individual drone to the task T j ;

[0045] S43. Based on the global fitness function, the task allocation is optimized, and the task urgency weighting factor P urg (T j , t) is introduced to construct the global task scheduling priority:

[0046]

[0047] where θ1 is the maximum value of the task urgency, θ2 is the urgency change rate factor, t d is the deadline of the task T j , t is the task execution time, and P alloc (T j , t) represents the priority weight of the task T j in the global task allocation;

[0048] S44. During the global task optimization process, a dynamic task collaboration mechanism is adopted to construct the global task allocation plan:

[0049]

[0050] where Ω t+1Denote the global task allocation scheme of the UAV swarm at time t+1, Ω t Denote the global task allocation scheme of the UAV swarm at time t, α is the global fitness optimization step size, and β is the task allocation priority optimization step size. Denote the gradient function;

[0051] S45. Set the global task revenue function of the UAV swarm:

[0052]

[0053] where, R global (t) is the total revenue of UAV task allocation at time t, and C conf (U i ,T j ,t) is the task conflict factor among UAVs, and λ C is the task conflict adjustment factor.

[0054] Optionally, the S5 specifically includes:

[0055] S51. Set the UAV adaptive adjustment mechanism to trigger the task adjustment strategy according to the real-time environment information during the task execution;

[0056] S52. Define the UAV task execution status function to monitor the status of UAVs during the task execution:

[0057] S exec (U i ,T j ,t) = α1·E rem (U i ,t) + α2·R(U i ,T j ) + α3·C conf (U i ,T j ,t);

[0058] where, S exec (U i ,T j ,t) is the status value of UAV executing task T j at time t, E rem (U i ,t) is the remaining battery power of the UAV, R(U i ,T j ) is the task revenue of individual UAV U i executing task T j , and C conf (U i ,T j, t) is the task conflict factor between UAVs, and α1, α2, α3 are the task status weight coefficients, satisfying α1 + α2 + α3 = 1;

[0059] S53. Set the task status threshold, and trigger the adaptive adjustment mechanism according to the UAV task status:

[0060]

[0061] Among them, δ L is the low task status threshold. If the UAV task status evaluation is lower than this value, the adjustment strategy is triggered;

[0062] S54. After the UAV triggers the adaptive adjustment mechanism, based on the improved Egrets Flock Optimization Algorithm, perform global task reallocation, adjust the UAV task execution strategy, and optimize the individual task selection plan according to the improved Artificial Protozoa Optimization Algorithm until the system reaches the optimal equilibrium state again. The finally optimized task competition game model is defined as:

[0063] G * = (U, T, Π * , R * );

[0064] Among them, U is the UAV agent set, T is the task set, Π * is the finally optimized UAV task allocation strategy set, and R * is the finally optimized UAV task revenue function.

[0065] The beneficial effects of the present invention are:

[0066] (1) Based on the distributed cooperative game mechanism, the present invention constructs a task competition game model. By defining the revenue function, strategy set, and game rules among UAV agents, the UAV can autonomously make task selection based on local information, and reduce task conflicts through task negotiation and revenue equilibrium strategies. In addition, combined with the improved Egrets Flock Optimization Algorithm (IESOA), the present invention introduces a global task fitness function to measure the adaptability of the UAV group to tasks, and constructs a global task scheduling priority in combination with the task urgency weighting factor to ensure that high-priority tasks are executed first, thereby improving the balance of task allocation and avoiding the problems of tasks not being executed or UAV resource waste.

[0067] (2) During the task execution process of the present invention, an adaptive task adjustment mechanism based on evolutionary game is introduced. The individual task optimization of UAVs is achieved through an improved artificial protozoa optimization algorithm (APO). By setting a dormancy strategy and an adaptive evolution and reproduction strategy, UAVs can dynamically adjust their task selection strategies according to the task competition situation, energy state, and task execution difficulty, thereby improving the task execution success rate. In addition, if the UAV encounters environmental changes such as insufficient power, task failure, or task demand changes during the task execution process, the adaptive adjustment mechanism of the present invention can trigger a task reallocation strategy, and perform global task optimization based on the profit function and strategy set, ensuring that the UAV task scheduling scheme can still maintain the optimal equilibrium in a complex environment and improving the stability of the system and the task completion rate.

[0068] (3) By combining the improved artificial protozoa optimization algorithm (APO) and the improved egret flock optimization algorithm (IESOA), the present invention proposes a scheduling method that combines individual decision optimization and global task scheduling optimization, enabling the UAV swarm to efficiently complete tasks. The APO algorithm introduces a strengthened dormancy strategy and an adaptive evolution and reproduction strategy to ensure that UAVs perform autonomous optimization according to their own power, task benefits, and task competition situation, improving task adaptability; the IESOA algorithm ensures the global optimality of task scheduling and improves the system task execution efficiency through task collaborative optimization, global task fitness evaluation, task urgency regulation, and profit function optimization. In addition, the profit equilibrium mechanism of the present invention effectively reduces the imbalance phenomenon during task execution, enables the full utilization of UAV resources, and improves the task completion efficiency and profit level of the entire UAV cluster. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the present invention, and do not constitute a limitation to the present invention. In the drawings:

[0070] Figure 1 is the overall flowchart of a multi-objective task scheduling method for UAVs based on distributed cooperative game proposed by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0071] Now, the present invention will be further described in detail with reference to the drawings. These drawings are all simplified schematic diagrams, only showing the basic structure of the present invention in a schematic way, so they only show the components related to the present invention.

[0072] Refer to Figure 1 , a multi-objective task scheduling method for UAVs based on distributed cooperative game, includes the following steps:

[0073] S1. Build a UAV mission scheduling model, define the UAV cluster and task set, set task attributes, initialize the status information of UAVs, and generate basic data;

[0074] S2. Based on the distributed cooperative game mechanism, define the strategy set, revenue function, and game rules of UAV agents for the basic data, and build a task competition game model as the core basis for UAV task allocation and strategy interaction;

[0075] S3. In the task competition game model, each UAV makes individual decisions through an improved artificial protozoa optimization algorithm based on local information, evaluates the revenue of each task using the revenue function, selects the optimal task according to the strategy set, strengthens the dormancy strategy, and adapts the evolutionary reproduction strategy;

[0076] S4. The task competition game model further guides the task negotiation between UAVs. Based on the improved egret flock optimization algorithm, global task allocation is performed according to the game rules. The global task fitness function is used to measure the adaptability of the UAV group to tasks, and a task urgency weighting factor is introduced to construct the global task scheduling priority; in the global task optimization process, a task allocation adjustment equation is constructed, the global task revenue function of the UAV group is set, and the UAV task scheduling scheme is optimized in combination with the task conflict factor;

[0077] S5. During the task execution process, based on the game rules, strategy set, and revenue function defined in the task competition game model, if the environment changes, the UAV triggers an adaptive adjustment mechanism, re-evaluates according to the revenue function, and adjusts the task allocation according to the strategy set until the system reaches the optimal equilibrium state again;

[0078] S6. After the task execution is completed, based on the long-term revenue feedback, use the task competition game model to analyze the UAV strategy adjustment data, and update the revenue function parameters to obtain a complete task scheduling method.

[0079] In this embodiment, the specific steps of S2 include:

[0080] S21. Define the UAV agent set U and task set T, establish a distributed game framework for UAV task scheduling, and set each UAV U i as a game player, the strategy set Π is used to select the task allocation scheme, and the individual UAV U i executes the task T j The revenue function R(U i , T j ) is used to calculate the optimal strategy of each UAV, and define the task competition game model G=(U, T, Π, R);

[0081] S22. Set the basic rules of the task competition game, and each UAV U iIt can only obtain the task status information of itself and neighboring drones to make local optimal strategy selection. If two or more drones compete for the same task T j generate competition, the optimal allocation strategy is calculated based on the revenue function. The drone can dynamically adjust the strategy during the task execution process, and the revenue function gradually converges to the game equilibrium state;

[0082] S23. During the task competition process, set the task selection strategy, and finally construct the drone task competition game model. The drone makes autonomous decisions based on the distributed cooperative game mechanism during the task allocation process to achieve the equilibrium optimization of task allocation.

[0083] In this embodiment, the specific steps of S3 are as follows:

[0084] S31. Based on the task competition game model G, each drone U i makes individual decisions based on local information, and uses the improved artificial protozoa optimization algorithm for task selection. The improvement includes strengthening the dormancy strategy and the adaptive evolutionary reproduction strategy;

[0085] S32. When the drone U i performs task selection, if it faces problems such as task resource competition, insufficient power, or communication obstruction, execute the strengthened dormancy strategy, calculate the optimal timing for the drone to enter dormancy, and optimize the dormancy decision through the following two steps, which specifically include:

[0086] S321. Dynamically adjust the dormancy threshold based on energy consumption, and define the drone energy threshold function:

[0087]

[0088] where E min is the minimum energy requirement for the drone to execute the task, E max is the maximum energy consumption that the drone can bear, is the average energy consumption of the drone's historical task execution, σ(E,t) is the energy fluctuation function, which calculates the uncertainty of energy consumption in the task environment, and γ1, γ2, γ3, γ4 are adjustment coefficients that satisfy γ1 + γ2 + γ3 + γ4 = 1;

[0089] S322. Strengthen the dormancy strategy based on the task competition intensity, and define the task competition factor:

[0090]

[0091] where n conf (T j ,t) represents the number of drones competing for task T j at time t, and n avail (T j, t) represents the number of remaining optional tasks at time t;

[0092] S33. Set the sleep probability function of the UAV:

[0093] When E(U i , t) < E th (U i , t), P sleep (U i , t) = 1;

[0094] Among them, P sleep (U i , t) is the probability that the UAV enters the sleep state, τ is the temperature parameter. When the energy of the UAV E(U i , t) is lower than the threshold E th (U i , t), the sleep probability is set to 1;

[0095] S34. In the process of optimizing the UAV task selection, introduce an adaptive evolutionary reproduction strategy, optimize the task allocation strategy according to the historical task execution data of the UAV, and set the individual evolution equation:

[0096]

[0097] Among them, is the strategy after the UAV selects tasks in the (t + 1)-th round, δ is the evolution adjustment factor, which controls the strategy update step size, is the gradient of the task revenue function with respect to the strategy, ρ is the empirical learning factor, and the strategy is optimized based on the historical task execution experience is the task environment adaptation factor, which dynamically adjusts the strategy in combination with time, is the intelligent evolution function, which adjusts the strategy mutation direction according to the task execution success rate and task difficulty, so that the UAV evolves a more stable task selection strategy. ζ·(E th (U i , t) - E(U i , t)) is the influence term that enables the UAV to evolve a scheduling strategy that is more biased towards low-energy-consuming tasks in a low-energy state;

[0098] S35. Introduce a task selection mechanism based on dynamic energy constraints. UAVs with lower energy will automatically reduce the revenue weight of high-energy-consuming tasks and preferentially select low-energy-consuming tasks:

[0099] R adj (U i , T j ) = R(U i , T j ) - λ E ·(E(Ui , t)-E th (U i , t));

[0100] Wherein, R(U i , T j ) is the revenue function for the individual drone U i to execute task T j , and R adj (U i , T j ) is the adjusted revenue function for the individual drone U i to execute task T j , and λ E is the energy regulation factor;

[0101] S36. During the process of a drone executing a task, by combining the enhanced sleep strategy and the adaptive evolutionary reproduction strategy, it dynamically adjusts its own task allocation plan. If the drone is in a sleep state for a long time, it re-adjusts the task selection strategy according to the task revenue function. During the task competition process, if two or more drones compete for the same task T j , the optimal task allocation is calculated based on the game rules in the task competition game model, and the task priority is dynamically adjusted.

[0102] In this embodiment, the specific content of S4 includes:

[0103] S41. Based on the individual task selection of the improved artificial protozoa optimization algorithm, the improved egret flock optimization algorithm is used to optimize the global task allocation;

[0104] S42. During the task competition game process, a global task fitness evaluation function is defined to measure the adaptability of the entire drone cluster to the task, and the global fitness function is set:

[0105]

[0106] Wherein, F global (T, t) is the total fitness of the drone swarm to execute the task set T at time t, and F task (U i , T j , t) is the fitness of the individual drone to task T j ;

[0107] S43. Optimize the task allocation based on the global fitness function, introduce the task urgency weighting factor P urg (T j , t), and construct the global task scheduling priority:

[0108]

[0109] Among them, θ1 is the maximum value of the task urgency, θ2 is the urgency change rate factor, and t d is the deadline of task T j , t is the task execution time, and P alloc (T j , t) represents the priority weight of task T j in the global task allocation;

[0110] S44. During the global task optimization process, adopt a dynamic task collaboration mechanism to construct a global task allocation scheme:

[0111]

[0112] Among them, Ω t+1 represents the global task allocation scheme of the UAV swarm at time t + 1, and Ω t represents the global task allocation scheme of the UAV swarm at time t, α is the global fitness optimization step size, β is the task allocation priority optimization step size, represents the gradient function;

[0113] S45. Set the global task revenue function of the UAV swarm:

[0114]

[0115] Among them, R global (t) is the total revenue of UAV task allocation at time t, and C conf (U i , T j , t) is the task conflict factor between UAVs, and λ C is the task conflict adjustment factor.

[0116] In this embodiment, the S5 specifically includes:

[0117] S51. Set the UAV adaptive adjustment mechanism to trigger the task adjustment strategy according to the real-time environment information during the task execution process;

[0118] S52. Define the UAV task execution status function to monitor the status of the UAV during the task execution process:

[0119] S exec (U i , T j , t) = α1·E rem (U i , t) + α2·R(U i , T j ) + α3·C conf (U i , T j , t);

[0120] Among them, S exec (U i , T j , t) is the state value of the UAV executing task T at time t j E rem (U i , t) is the remaining battery power of the UAV, R(U i , T j ) is the task benefit of individual UAV U i executing task T j C conf (U i , T j , t) is the task conflict factor between UAVs, and α1, α2, α3 are task state weight coefficients, satisfying α1 + α2 + α3 = 1;

[0121] S53. Set the task state threshold and trigger the adaptive adjustment mechanism according to the UAV task state:

[0122]

[0123] Among them, δ L is the low task state threshold. If the UAV task state evaluation is lower than this value, the adjustment strategy is triggered;

[0124] S54. After the UAV triggers the adaptive adjustment mechanism, based on the improved egret flock optimization algorithm, perform global task reallocation, adjust the UAV task execution strategy, and optimize the individual task selection plan according to the improved artificial protozoa optimization algorithm until the system reaches the optimal equilibrium state again. The finally optimized task competition game model is defined as:

[0125] G * =(U, T, Π * , R * );

[0126] Among them, U is the set of UAV agents, T is the set of tasks, Π * is the finally optimized UAV task allocation strategy set, and R * is the finally optimized UAV task benefit function.

[0127] Example 1:

[0128] To verify the feasibility of the present invention in implementation, the present invention is applied to a certain UAV cluster logistics distribution system, which is used for the urban rapid distribution scenario, involves multiple UAVs executing multiple distribution tasks, optimizes the UAV scheduling scheme in a dynamic environment, and improves the task completion rate and system revenue. The main objective of this experiment is to verify the advantages of the present invention in terms of task scheduling balance, dynamic environment adaptability, task completion rate, and UAV resource utilization. The experimental site is set in a large logistics center, which needs to dispatch hundreds of UAVs to execute cross-city distribution tasks every day, involving multi-objective task allocation, different task priorities, and some tasks have time constraints.

[0129] The experiment is carried out by comparing two schemes. One is the traditional UAV task scheduling algorithm based on heuristic rules, and the other is the UAV multi-objective task scheduling method based on distributed cooperative game proposed by the present invention. In the experiment, 50 UAVs are set, and the number of distribution tasks is 200, 400, 600, 800, and 1000 respectively to test the scheduling effect of the system under different task loads. At the same time, considering the impact of dynamic task changes, 10%-30% of the tasks are randomly set to be adjusted due to environmental changes (such as weather changes, traffic control, user changed requirements) in the experiment to test the adaptability of the scheduling algorithm to the dynamic environment.

[0130] During the experiment, the scheduling method of the present invention first constructs a task competition game mechanism based on the distributed cooperative game model, enabling UAVs to independently select the optimal task based on local information. Then, combined with the improved artificial protozoa optimization algorithm, the individual tasks of UAVs are optimized, enabling UAVs to dynamically adjust the task selection scheme according to their remaining battery power, task revenue, and task competition situation, and optimize task execution through the sleep strategy and adaptive evolutionary reproduction strategy. At the same time, the improved egret flock optimization algorithm is used for global task scheduling to make the task allocation more balanced, reduce task conflicts among UAVs, and improve the overall system revenue. When UAVs encounter emergencies (such as weather changes, user order modification, UAV failure, etc.) during task execution, the adaptive adjustment mechanism of the present invention can trigger task reallocation, and perform dynamic optimization based on the task revenue function and strategy set, enabling the system to quickly return to the optimal equilibrium state.

[0131] The experimental results show that the UAV task scheduling method of the present invention is significantly superior to the traditional scheduling method in terms of task completion rate, UAV task load balance, UAV resource utilization rate, dynamic environment adaptability, and system revenue. When the total number of tasks is 1000, the task completion rate of the present invention reaches 95.2%, which is 12.8% higher than that of the traditional method; the task load balance of the UAV is increased by 20.3%, that is, the task distribution among different UAVs is more balanced, avoiding the situation of some UAVs being overloaded or idle; the resource utilization rate of the UAV is increased by 18.7%, that is, the power consumption of the UAV is more reasonable, enabling the overall system to consume less energy while executing more tasks. In addition, in the case of 30% of the tasks changing dynamically, the scheduling method of the present invention can complete task reallocation within an average of 3.7 seconds, while the traditional method requires 7.9 seconds, proving that the scheduling method of the present invention has better adaptability in a dynamic environment.

[0132] Table 1: Experimental data table of UAV task scheduling

[0133]

[0134] It can be seen from the experimental data that the task completion rate of the present invention is always higher than that of the traditional method, and the gap is more significant when the number of tasks is large (600 or more). The task completion rate of the present invention remains at 95.2% when there are 1000 tasks, while the traditional method drops to 82.4%, indicating that the scheduling method of the present invention has more advantages in high-load task scenarios.

[0135] In addition, the task load balance reflects the uniformity of UAV task allocation. The load balance of the present invention reaches 85.6% when the number of tasks is 1000, which is 16.4% higher than that of the traditional method, indicating that the optimization strategy of the present invention can effectively reduce the imbalance of task allocation and avoid the situation of some UAVs working overloaded while other UAVs are idle.

[0136] In terms of resource utilization rate, the UAV task scheduling method of the present invention optimizes the task execution path of the UAV, making the power consumption of the UAV more balanced, and the overall resource utilization rate is increased by 18.7%, indicating that the present invention can effectively reduce the waste of UAV resources and improve the task execution efficiency. In addition, when the number of tasks reaches 1000, the average scheduling time of the present invention is only 3.7 seconds, which is 4.2 seconds less than that of the traditional method, indicating that the present invention has a higher scheduling response ability in a dynamic environment and can quickly complete task allocation adjustment.

[0137] The multi-objective task scheduling method for drones based on distributed collaborative game of the present invention is superior to traditional methods in key indicators such as task completion rate, load balance degree, resource utilization rate, and scheduling time, proving the significant advantages of the present invention in improving task execution efficiency, reducing task conflicts, and enhancing the resource utilization rate of drones, and can be more effectively applied to the large-scale drone cluster task scheduling in a dynamic environment.

[0138] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, making equivalent replacements or changes should be covered within the protection scope of the present invention.

Claims

1. A multi-objective task scheduling method for unmanned aerial vehicles based on distributed collaborative game, characterized in that It includes the following steps: S1. Define the UAV cluster and task set, initialize the state information of the UAVs, and generate basic data; S2. Based on the distributed cooperative game mechanism, define the strategy set, payoff function, and game rules of the UAV agents for the basic data, and construct a task competition game model; S3. In the task competition game model, each UAV makes individual decisions through an improved artificial protozoa optimization algorithm based on local information, evaluates the payoff of each task using the payoff function, selects the optimal task according to the strategy set, strengthens the dormancy strategy, and adapts the evolutionary reproduction strategy; S4. The task competition game model performs global task allocation through an improved egret flock optimization algorithm, measures the adaptability of the UAV group to tasks based on the global task fitness function, introduces a task urgency weighting factor to construct the global task scheduling priority, constructs a task allocation adjustment equation during the global task optimization process, sets the global task payoff function of the UAV group, and optimizes the UAV task scheduling scheme in combination with the task conflict factor; S5. Construct an adaptive adjustment mechanism, adjust the task allocation according to the payoff function evaluation and strategy set until the system reaches the optimal equilibrium state again; S6. After the task execution is completed, update the payoff function parameters according to the long-term payoff feedback to obtain a complete task scheduling method.

2. The method for multi-objective task scheduling of unmanned aerial vehicles based on distributed collaborative game according to claim 1, wherein The specific content of S2 includes: S21. Define the set of UAV agents U and the set of tasks T, establish a distributed game framework for UAV task scheduling, and set each UAV U i as a game player, with the strategy set Π used to select a task allocation scheme, and the individual UAV U i executes the task T j The revenue function R(U i , T j ) is used to calculate the optimal strategy of each UAV, and define the task competition game model G = (U, T, Π, R); S22. Set the basic rules of the task competition game. Each drone U i can only obtain the task status information of itself and neighboring drones to make a locally optimal strategy selection. If two or more drones compete for the same task T j a competition occurs, then the optimal allocation strategy is calculated based on the payoff function. The drone can dynamically adjust the strategy during the task execution process, and the payoff function gradually converges to the game equilibrium state; S23. During the task competition process, set the task selection strategy, and finally construct a UAV task competition game model. The UAV makes autonomous decisions based on the distributed cooperative game mechanism during the task allocation process to achieve the equilibrium optimization of task allocation.

3. A multi-objective task scheduling method for unmanned aerial vehicles based on distributed collaborative game according to claim 1, characterized in that The specific content of S3 includes: S31. Based on the task competition game model G, each unmanned aerial vehicle U i makes individual decisions based on local information and uses an improved artificial protozoa optimization algorithm for task selection. The improvements include strengthening the dormancy strategy and the adaptive evolutionary reproduction strategy; S32. When the drone U i performs task selection, if it faces problems such as task resource competition, insufficient power, or communication obstruction, it executes the enhanced sleep strategy, calculates the optimal timing for the drone to enter the sleep state, and optimizes the sleep decision through the following two steps, specifically including: S321. Dynamically adjust the dormancy threshold based on energy consumption, and define the UAV energy threshold function: Among them, E min is the minimum energy requirement for UAV mission execution, and E max is the maximum energy consumption that the UAV can withstand. is the average energy consumption of the UAV's historical mission execution. σ(E, t) is the energy fluctuation function, which calculates the uncertainty of energy consumption in the mission environment. γ1, γ2, γ3, and γ4 are adjustment coefficients that satisfy γ1 + γ2 + γ3 + γ4 = 1; S322. Strengthen the dormancy strategy based on the task competition intensity, and define the task competition factor: where n conf (T j , t) represents the number of UAVs competing for task T at time t, and n j (T avail , t) represents the number of remaining tasks available at time t; j ​ S33. Set the dormancy probability function of the UAV: When E(U i , t) < E th (U i , t), then P sleep (U i , t) = 1; Among them, P sleep (U i , t) is the probability that the UAV enters the dormant state, τ is the temperature parameter. When the energy E(U i , t) of the UAV is lower than the threshold E th (U i , t), the dormant probability is set to 1; S34. During the UAV task selection optimization process, introduce an adaptive evolutionary reproduction strategy, optimize the task allocation strategy according to the historical task execution data of the UAV, and set the individual evolution equation: Among them, is the strategy of the UAV after the task selection in the (t + 1)-th round, δ is the evolutionary adjustment factor that controls the step size of strategy update, is the gradient of the task revenue function with respect to the strategy, ρ is the empirical learning factor, and the strategy is optimized based on the historical task execution experience θ is the task environment adaptation factor that dynamically adjusts the strategy according to time, is the intelligent evolution function that adjusts the strategy mutation direction according to the task execution success rate and task difficulty, enabling the UAV to evolve a more stable task selection strategy, ζ·(E th (U i ,t)-E(U i ,t)) is the influencing term that enables the UAV to evolve a scheduling strategy that is more biased towards low-energy-consuming tasks in the low-energy state; S35. Introduce a task selection mechanism based on dynamic energy constraints. UAVs with lower energy will automatically reduce the payoff weight of high-energy-consuming tasks and preferentially select low-energy-consuming tasks: R adj (U i ,T j ) = R(U i ,T j ) - λ E ·(E(U i ,t) - E th (U i ,t)); Among them, R(U i , T j ) is the revenue function for the individual unmanned aerial vehicle U i to execute task T j , R adj (U i , T j ) is the adjusted revenue function for the individual unmanned aerial vehicle U i to execute task T j , and λ E is the energy adjustment factor; S36. During the mission execution, the drone dynamically adjusts its task allocation plan by combining the enhanced sleep strategy and the adaptive evolutionary reproduction strategy. If the drone remains in the sleep state for a long time, it re-adjusts the task selection strategy according to the task benefit function. During the task competition, if two or more drones compete for the same task T j compete, the optimal task allocation is calculated based on the game rules in the task competition game model, and the task priority is dynamically adjusted.

4. A method for multi-objective task scheduling of unmanned aerial vehicles based on distributed cooperative game according to claim 1, characterized in that, The specific content of S4 includes: S41. Based on the individual task selection of the improved artificial protozoa optimization algorithm, use the improved egret flock optimization algorithm to optimize the global task allocation; S42. During the task competition game process, define the global task fitness evaluation function to measure the adaptability of the entire UAV cluster to tasks, and set the global fitness function: Among them, F global (T, t) is the total fitness of the UAV swarm executing the mission set T at time t, and F task (U i , T j , t) is the fitness of an individual UAV for the mission T j . S43. Optimize task allocation based on the global fitness function, and introduce the task urgency weighting factor P urg (T j , t), and construct the global task scheduling priority: Among them, θ1 is the maximum value of the task urgency, θ2 is the urgency change rate factor, and t d is the deadline of task T j , t is the task execution time, and P alloc (T j , t) represents the priority weight of task T j in the global task allocation; S44. During the global task optimization process, adopt a dynamic task coordination mechanism to construct a global task allocation plan: Among them, Ω t+1 represents the global task allocation scheme of the UAV swarm at time t+1, and Ω t represents the global task allocation scheme of the UAV swarm at time t. α is the global fitness optimization step size, and β is the task allocation priority optimization step size. represents the gradient function; S45. Set the global task payoff function of the UAV group: Among them, R global (t) is the total revenue of UAV mission allocation at time t, C conf (U i , T j , t) is the task conflict factor among UAVs, and λ C is the task conflict adjustment factor.

5. A method for multi-objective task scheduling of unmanned aerial vehicles based on distributed collaborative game according to claim 1, characterized in that, The specific content of S5 includes: S51. Set the UAV adaptive adjustment mechanism, and trigger the task adjustment strategy according to the real-time environment information during the task execution process; S52. Define the UAV task execution status function to monitor the status of the UAV during the task execution process; S exec (U i ,T j ,t) = α1·E rem (U i ,t) + α2·R(U i ,T j ) + α3·C conf (U i ,T j ,t); Among them, S exec (U i , T j , t) is the state value of the UAV performing mission T j at time t, E rem (U i , t) is the remaining battery power of the UAV, R(U i , T j ) is the mission benefit of individual UAV U i performing mission T j , C conf (U i , T j , t) is the mission conflict factor among UAVs, and α1, α2, α3 are mission state weight coefficients, satisfying α1 + α2 + α3 = 1; S53. Set the task status threshold, and trigger the adaptive adjustment mechanism according to the UAV task status: Among them, δ L is the low threshold of the task status. If the UAV task status assessment is lower than this value, the adjustment strategy is triggered; After the UAV triggers the adaptive adjustment mechanism, it performs global task reallocation based on the improved egret flock optimization algorithm, adjusts the UAV task execution strategy, and optimizes the individual task selection scheme according to the improved artificial protozoa optimization algorithm until the system reaches the optimal equilibrium state again. Finally, the optimized task competition game model is defined as: G * =(U, T, Π * , R * ); Among them, U is the set of UAV agents, T is the set of tasks, and Π * is the set of the finally optimized UAV task allocation strategies, and R * is the finally optimized UAV task revenue function.

Images