Task allocation and endurance optimization method for multi-unmanned aerial vehicle system

Abstract

The invention relates to the technical field of unmanned aerial vehicles. The invention discloses a task allocation and endurance optimization method for a multi-unmanned aerial vehicle system. The method comprises the steps of S1, task modeling, S2, initial population generation, S3, fitness function design, S4, genetic algorithm task optimization, S5, iterative optimization, S6, path planning and S7, data collection, prediction and adjustment. The multi-unmanned aerial vehicle system task allocation and endurance optimization method can process temporary tasks in real time, flexibly adjust a task allocation scheme, dynamically adjust flight parameters according to real-time flight conditions and task requirements, comprehensively consider various factors such as task priorities, flight paths, energy consumption and the like, and optimize the task allocation and endurance optimization of the multi-unmanned aerial vehicle system on the basis of actual flight data. And an energy consumption prediction model is established, and accurate energy consumption management and path planning are realized. According to the method, temporary tasks can be processed, task allocation and path planning are dynamically adjusted, and task execution efficiency and safety are ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicles, and in particular to a method for task allocation and endurance optimization of a multi-unmanned aerial vehicle system. Background Art

[0002] With the rapid development of drone technology, multi-drone systems have been widely used in logistics distribution, environmental monitoring, agricultural spraying, search and rescue and other fields. However, the task allocation and endurance optimization of multi-drone systems remain a challenging problem. Effective task allocation and endurance optimization can not only improve the efficiency of task execution, but also extend the flight time of drones, reduce energy consumption, and improve the overall performance of the system. A multi-drone system consists of multiple drones, each with different flight capabilities and task execution capabilities. These drones can work together to complete complex tasks. However, how to effectively allocate tasks, optimize flight paths, and manage the energy consumption of drones are key issues facing multi-drone systems. Traditional task allocation methods are mainly based on static task allocation, that is, all tasks are allocated at once before the task starts. This method is difficult to adjust flexibly when facing temporary tasks or emergencies, and it is easy to lead to inefficient task execution. In addition, static task allocation methods often ignore the current status of drones such as power and load, which may lead to unreasonable task allocation and affect the smooth completion of tasks. The endurance of drones is affected by many factors, including flight speed, flight altitude, load, wind speed, etc. Traditional endurance optimization methods mainly rely on fixed flight parameters and lack the ability to adapt to dynamic environments. In practical applications, drones need to dynamically adjust flight parameters according to real-time flight conditions and mission requirements to optimize endurance performance.

[0003] Existing multi-UAV systems often fail to cope with temporary tasks or emergencies in terms of allocation and endurance optimization, lack flexibility, often ignore the dynamic changes in flight conditions, and find it difficult to achieve optimal endurance performance. Some multi-UAV systems only focus on task completion time or energy consumption, fail to comprehensively consider the impact of multiple factors, and lack energy consumption prediction and path optimization based on actual flight data, making it difficult to achieve accurate energy consumption management and path planning. Summary of the invention

[0004] In order to address the shortcomings of the prior art, the present application discloses a method for task allocation and endurance optimization of a multi-UAV system that can process temporary tasks in real time, flexibly adjust task allocation plans, dynamically adjust flight parameters according to real-time flight conditions and task requirements, comprehensively consider various factors such as task priority, flight path, energy consumption, etc., and establish an energy consumption prediction model based on actual flight data to achieve accurate energy consumption management and path planning.

[0005] The present invention discloses a method for task allocation and endurance optimization of a multi-UAV system, which comprises: S1: task modeling, defining a task set and a UAV set, wherein each task and UAV have corresponding attributes, S2: initial population generation, randomly generating multiple task allocation schemes to form an initial population, S3: fitness function design, defining a fitness function, S4: genetic algorithm optimization of tasks, selection, crossover and mutation optimization of task allocation schemes, generating new task allocation schemes through selection, crossover and mutation operations, S5: iterative optimization, repeating the above steps until a termination condition is met, and selecting an optimal task allocation scheme, S6: path planning, generating a flight path of each UAV using an A* algorithm according to the task allocation scheme, S7: data collection prediction and adjustment, recording energy consumption data under various flight conditions during the flight of the UAV, extracting key features from the original data, training a random forest model using training set data, predicting the energy consumption of each UAV using the random forest model during the flight, and adjusting the flight strategy according to the prediction results.

[0006] Furthermore, the task set in S1 defines all tasks , each task T i There is position p i , priority π i , and the time required t i Attributes, drone set defines all drones , each drone D j The maximum flight distance d max , Maximum load l max , and the current power E j Attributes, multiple task allocation schemes are randomly generated in S2, and multiple task allocation schemes form the initial population P0. The fitness function f(x) in S3 is defined as follows: , x: represents a task allocation scheme in the initial population P0, w1, w2, w3: weight coefficients used to balance the importance of different goals, the weight coefficients are based on p i , π i ,t i d max , l max and E j Joint decision, T total :Indicates drone D j The total time required to complete the task, L total :Indicates drone D j The total flight distance required to complete the mission, E total :Indicates drone D j The total energy required to complete a task.

[0007] Further, it includes S8: temporary task processing step, temporary task modeling: defining temporary tasks, task reallocation: when receiving a temporary task, reusing the genetic algorithm to optimize the task allocation plan and assigning the temporary task to a suitable UAV, fitness function update: updating the fitness function, considering the impact of the temporary task, path replanning: according to the new task allocation plan, using the A* algorithm to regenerate the flight path of each UAV, energy consumption prediction update: reusing the random forest model to predict the energy consumption of each UAV, including the impact of the temporary task.

[0008] Furthermore, temporary task modeling includes defining temporary tasks T temp , including the position p temp , priority π temp , time required t temp , temporary task T temp Including the path length L compared to the original task T temp And energy consumption E temp , the total path length after adding temporary tasks: L new =L total +L temp , the total energy consumption after adding temporary tasks: E new =E total +E temp , the fitness function f(x) is updated as follows when processing temporary tasks: , w4, w5, w6: The weight coefficients of each task after adding temporary tasks. The weight coefficients of each task after adding temporary tasks are calculated based on p i , π i ,t i d max , l max 、E j 、p temp , π temp and t temp Decide together, repeat the S4 genetic algorithm optimization task and S5 iterative optimization to select the best task allocation solution.

[0009] Furthermore, the adaptive control strategy adjustment includes: Predicting high energy consumption, the drone reduces flight speed or chooses a gentle flight route, reducing flight speed: v new =v current -Δv,v new : The speed after reduction, v current: Current flight speed, Δv: Reduced flight speed, Select a gentle flight route, The drone reduces altitude changes, Use the A* algorithm to regenerate the path, Select a path that minimizes altitude changes and a low wind speed altitude layer, Predict low energy consumption, Increase flight speed or choose a more direct flight route, Increase flight speed: v new =v current +Δv,V new : Reduced speed, V current : Current flight speed, Δv: Reduced flight speed, choose a more direct flight route, and use the A* algorithm to regenerate the shortest path.

[0010] Furthermore, S4 includes calculating the fitness value f(x i ), calculate the total fitness F total : , N: population size, that is, the number of individuals, f(x i ): The fitness value of the i-th individual, F total : The sum of all individual fitness values, calculate individual x i The probability of choosing P(x i ): , P(x i ): The probability of the i-th individual being selected. Generate a random number r∈[0,1] to accumulate the selection probability. When the accumulated value exceeds r, select the corresponding individual. The crossover includes: Single-point crossover: randomly select a crossover point k: k∈[1, n-1], and generate two offspring individuals x c1 and x c2 : , x p1 : The first parent individual, representing a task allocation scheme, x p2 : The second parent individual represents a task allocation scheme, x p1，i : The gene of the first parent individual at position i, representing task T i Assigned to which drone, x p2，i : The gene of the second parent individual at position i, representing task T i Assigned to which drone, x c1 : The first child individual, representing a new task allocation scheme, x c2 : The second offspring individual, indicating a new task allocation scheme, k: The position of the crossover point, indicating that gene exchange starts from this position, n: The number of tasks, that is, the length of the task allocation scheme, uniform crossover: For each gene position i, a random number r is generated in the interval [0, 1].i , set a crossover probability p c , according to the random number r i and the crossover probability p c , determines whether each gene position is exchanged. If r i <p c , then exchange the gene position, otherwise remain unchanged, for each position i, with probability p c Exchange genes: , p c : Crossover probability, a value in the interval [0, 1], mutation includes: with a certain probability p m Randomly select the task allocation scheme x of two positions to exchange genes, and for each position i with probability p m Choose another position j and swap genes: , P m : Mutation probability, x m，i : The gene of the individual at position i after mutation, x m，j : The gene of the individual at position j after mutation, x i : The gene of the original individual at position i, x j : The gene of the original individual at position j.

[0011] Furthermore, the termination condition of S5 iterative optimization is to set the maximum number of iterations G and the fitness convergence threshold ϵ. When the change in fitness is less than ϵ or reaches the maximum number of iterations G, the iteration is stopped.

[0012] Furthermore, the energy consumption data recorded in S7 includes the flight speed v current , flight altitude h current , height change rate h′ current , load l current , wind speed w current , current power E current , path length L current , each record also includes the corresponding energy consumption E actual , the task types include original task 1 and temporary task 0, and the energy consumption data is organized into feature vectors X current : , use the trained random forest model RF to predict the energy consumption E under current flight conditions predicted : ,RF: The trained random forest model is used to predict the energy consumption under the current flight conditions.

[0013] Furthermore, key features including flight speed, altitude change rate and wind speed are extracted from the original data, and the key features affecting energy consumption are selected using correlation analysis or feature importance evaluation.

[0014] The present invention discloses an application of a multi-UAV system task allocation and endurance optimization method, which is characterized by using any of the above multi-UAV system task allocation and endurance optimization methods in the field of UAVs. Beneficial effects of the present invention: The multi-UAV system task allocation and endurance optimization method of the present invention can handle temporary tasks in real time, flexibly adjust the task allocation scheme, dynamically adjust the flight parameters according to the real-time flight conditions and task requirements, comprehensively consider the priority of the task, the flight path, energy consumption and other factors, and establish an energy consumption prediction model based on actual flight data to achieve accurate energy consumption management and path planning. The method optimizes task allocation and path planning through genetic algorithms, and uses a random forest model to predict the energy consumption of the UAV, and adjusts the flight strategy in real time according to the prediction results to extend the endurance time. In addition, the method can also handle temporary tasks, dynamically adjust task allocation and path planning, and ensure the efficiency and safety of task execution.

[0015] The present invention constitutes a complete UAV task allocation and path planning process. Through task modeling, initial population generation, fitness function design, genetic algorithm optimization, iterative optimization, path planning and data collection prediction and adjustment, the optimization of task allocation scheme and efficient execution of UAV flight path are ensured. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A flowchart of a method for multi-UAV system task allocation and endurance optimization in an implementation manner of the present application.

[0017] Figure 2 A schematic diagram of a multi-UAV system in an embodiment of the present application.

[0018] Figure 3 Another schematic diagram of a multi-UAV system in an embodiment of the present application.

[0019] Figure 4 Another schematic diagram of a multi-UAV system in an embodiment of the present application.

[0020] Figure 5 A schematic diagram of a drone in an implementation manner of the present application. DETAILED DESCRIPTION

[0021] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the specific implementation modes of the present invention will be clearly and completely described below.

[0022] The present application discloses a multi-UAV system task allocation and endurance optimization method, which includes: S1: Task modeling, defining the task set and the drone set, each task and drone has corresponding attributes; by defining the attributes of tasks and drones, the specific content and constraints of the optimization problem are clarified, providing the necessary input data for subsequent steps.

[0023] S2: Initial population generation: randomly generate multiple task allocation schemes to form an initial population. The initial population includes multiple schemes. The diversity of the initial population helps to explore different areas of the solution space and avoid falling into local optimality. It provides a starting point for the genetic algorithm and enables the optimization process to iterate based on basic data.

[0024] S3: Fitness function design, define the fitness function, the fitness function is used to evaluate the quality of the plan, the higher the fitness, the better the plan, anyway, the plan is poor; the fitness function provides a quantitative evaluation standard for each task allocation plan, so that the optimization process has a clear goal, through the fitness function, the optimization process can be guided in the direction of improving fitness.

[0025] S4: Genetic algorithm optimizes tasks, selects, crosses and mutates to optimize task allocation schemes, and generates new task allocation schemes through selection, crossover and mutation operations; genetic algorithms can perform global searches in the solution space through selection, crossover and mutation operations to find better solutions.

[0026] S5: Iterative optimization, repeat the above steps until the termination condition is met and select the optimal task allocation plan; through multiple iterations, gradually improve the task allocation plan, improve the overall performance, set the maximum number of iterations and fitness convergence threshold, ensure that the algorithm terminates within a reasonable time, and avoid unnecessary calculations.

[0027] S6: Path planning: According to the task allocation plan, the A* algorithm is used to generate the flight path of each UAV. Through path planning, the flight path of each UAV is ensured to be the shortest and the energy consumption is the lowest, thereby improving the efficiency of task execution.

[0028] S7: Data collection, prediction and adjustment. During the flight of the drone, record the energy consumption data under various flight conditions, extract key features from the raw data, and use the training set data to train the random forest model. During the flight, use the random forest model to predict the energy consumption of each drone, and adjust the flight strategy according to the prediction results. By collecting actual flight data, establish a data-driven energy consumption prediction model to improve the accuracy of the prediction. During the flight, adjust the flight strategy in real time according to the predicted energy consumption results to ensure the smooth completion of the mission. By continuously collecting data and adjusting strategies, dynamic optimization is achieved to improve the adaptability of the system.

[0029] This application forms a complete UAV task allocation and path planning process through the above steps. Through task modeling, initial population generation, fitness function design, genetic algorithm optimization, iterative optimization, path planning and data collection prediction and adjustment, the optimization of task allocation scheme and efficient execution of UAV flight path are ensured.

[0030] In a multi-UAV system, it can effectively allocate tasks, plan paths, and optimize the flight time of UAVs. At the same time, it can handle temporary tasks and dynamically adjust task allocation and path planning to ensure efficient completion of tasks and long flight time of UAVs.

[0031] Main steps: 1. Task Modeling Task set: defines all tasks, each task has location, priority, and required time attributes.

[0032] Drone set: defines all drones, each with maximum flight distance, maximum payload, and current battery level attributes.

[0033] 2. Initial population generation Coding method: Use a table to represent the task allocation plan. Each row represents a task, each column represents a drone, and the number in the table indicates which drone the task is assigned to.

[0034] Initial population: Randomly generate multiple task allocation schemes to form an initial population.

[0035] 3. Fitness function design Fitness function: used to evaluate the quality of each task allocation scheme. Factors considered include task completion time, path length, and energy consumption.

[0036] Mission completion time: The total time required for the drone to complete all missions.

[0037] Path Length: The total distance flown by the drone.

[0038] Energy consumption: The total energy required by the drone to complete its mission.

[0039] Fitness function: Taking the above factors into consideration, the fitness function formula is obtained.

[0040] 4. Selection, crossover, mutation Selection: Select a better task allocation scheme from the current population. The scheme with high fitness is more likely to be selected.

[0041] Crossover: Partially exchange two selected solutions to generate a new solution. For example, single-point crossover is to exchange parts of two solutions at a certain position.

[0042] Mutation: Randomly change the assignment of a task with a certain probability to introduce new diversity.

[0043] 5. Iterative Optimization Termination condition: set the maximum number of iterations G or the fitness change is less than a certain threshold ϵ.

[0044] Iteration process: Calculate fitness: Calculate the fitness of each solution in the current population.

[0045] Selection, crossover, mutation: generate new task allocation schemes.

[0046] Update the population: keep the solution with the highest fitness and replace the solution with the low fitness.

[0047] Repeat the above steps until the termination condition is met, setting the maximum number of iterations G or the fitness change is less than a certain threshold ϵ.

[0048] 6. Path Planning Path generation: Based on the task allocation plan, the A* algorithm is used to generate the flight path of each drone, taking into account path length and obstacle avoidance.

[0049] 7. Battery life optimization Data collection: During the flight of the drone, energy consumption data under various flight conditions is recorded.

[0050] Feature engineering extraction: Extract key features from the data, including flight speed, altitude, load, and wind speed.

[0051] Model training: Use the random forest model to train the energy consumption prediction model.

[0052] Real-time prediction: During flight, the energy consumption of each drone is predicted in real time.

[0053] Adaptive control: Based on the predicted energy consumption, the flight strategy is adjusted in real time, such as changing the flight speed, altitude or route to extend the flight time.

[0054] 8. Temporary task processing Temporary task modeling: Define temporary tasks, including location, priority, and required time attributes.

[0055] Task reallocation: When a temporary task is received, the genetic algorithm is reused to optimize the task allocation scheme and assign the temporary task to the appropriate UAV.

[0056] Fitness function update: Update the fitness function to take into account the impact of temporary tasks.

[0057] Path replanning: Regenerate the flight path of each drone based on the new task allocation plan.

[0058] Energy consumption prediction update: re-use the random forest model to predict the energy consumption of each drone, including the impact of temporary missions.

[0059] Adaptive control: Adjust flight strategies in real time based on updated energy consumption forecasts to extend flight time.

[0060] As an implementation, the task set in S1 defines all tasks , each task T i There is position p i , priority π i , and the time required t i property; The drone set defines all drones , each drone D j The maximum flight distance d max , Maximum load l max , and the current power E j property; In S2, multiple task allocation schemes are randomly generated, and the multiple task allocation schemes form an initial population P0; The fitness function f(x) in S3 is defined as follows: ; The fitness function f(x) is used to evaluate the quality of each task allocation scheme. The higher the fitness value, the better the scheme; x: represents a task allocation scheme in the initial population P0, which includes several task allocation schemes x; w1, w2, w3: weight coefficients used to balance the importance of different objectives. The weight coefficients are based on p i , π i ,t i d max , l max and E j Joint decision making; T total :Indicates drone D j The total time required to complete the task; L total :Indicates drone D j The total flight distance required to complete the mission; E total :Indicates drone D j The total energy required to complete a task.

[0061] Definition of weight coefficient: The specific value of weight coefficient is based on previous flight experience and p i , π i ,ti d max , l max and E j Calculated using a multi-UAV system.

[0062] Position p i :The closer the mission location is to the current location of the drone, the higher the weight is. This is because missions that are close are easier to complete and can reduce flight time and energy consumption.

[0063] Consideration: The current battery level of the drone j and the maximum flight distance d max If the mission distance is too far, exceeding the remaining flight capacity of the UAV, the weight of the mission should be reduced.

[0064] Priority π i :The higher the priority, the higher the weight of the task. High-priority tasks usually need to be processed first.

[0065] Considerations: Priority is an intrinsic property of the mission and is not affected by the properties of the drone.

[0066] Time required i :The longer the task takes, the lower the weight. This is because long tasks will take up more resources and affect the completion of other tasks.

[0067] Consideration: The current battery level of the drone j If the task takes too long, exceeding the time supported by the remaining battery power of the drone, the weight of the task should be reduced.

[0068] The current battery level of the drone is E j : The current battery level of the drone directly affects its ability to complete the mission. The lower the battery level, the lower the weight of the mission should be, to avoid the drone being unable to complete the mission due to insufficient battery.

[0069] Considerations: distance and time required for the task i If the mission is far away or takes a long time, and the drone is low on battery, the weight of the mission should be significantly reduced.

[0070] The maximum flight distance of the drone max : The maximum flight distance of a drone limits the range it can reach. If the mission distance exceeds the maximum flight distance of the drone, the weight of the mission should be significantly reduced.

[0071] Consideration: The distance of the task. If the task distance exceeds d max , then the weight of the task should be zero or very low.

[0072] Maximum payload of the drone max: The maximum payload of a drone limits the weight of items it can carry. If the payload required by a mission exceeds the drone's maximum payload, the weight of that mission should be significantly reduced.

[0073] Considerations: The load required by the task i If l i More than l max , then the weight of the task should be zero or very low.

[0074] Comprehensive consideration: Position p i :If the task position p i Approach the current position of the drone j , and the remaining power of the drone is E j If the mission distance exceeds the maximum flight distance d of the UAV, the mission weight is higher. max , the weight of the task is significantly reduced.

[0075] Priority π i : Tasks with high priority have higher weights and are not directly affected by the drone’s attributes.

[0076] Time required i :The mission takes a long time and the remaining battery power of the drone is E j When the remaining battery power of the drone is less than E, the task weight is lower. The task time is shorter and the remaining battery power of the drone is E. j When it is sufficient, the task has a higher weight.

[0077] The current battery level of the drone is E j : The current battery level of the drone E j The lower the value, the weight of the task should be appropriately reduced to avoid the drone being unable to complete the task due to insufficient power.

[0078] The maximum flight distance of the drone max :If the mission distance exceeds the maximum flight distance d of the drone max , then the weight of the task should be significantly reduced or zero.

[0079] Maximum payload of the drone max :If the load required by the task is l i Exceeding the maximum load of the drone max , then the weight of the task should be significantly reduced or zero.

[0080] Task allocation: Optimize task allocation through genetic algorithm to ensure that the task load of each drone is balanced and the path is shortest.

[0081] Path planning: Use the A* algorithm to generate the optimal path to ensure that the drone can complete the task efficiently.

[0082] Endurance optimization: Use machine learning to predict energy consumption, adjust flight strategies in real time, and extend the drone’s endurance.

[0083] Temporary task processing: Ability to dynamically adjust task allocation and path planning to ensure efficient processing of temporary tasks without affecting the completion and endurance of the overall task.

[0084] Through the above steps, efficient task allocation, path planning and endurance optimization can be achieved in a multi-UAV system. Even when facing temporary tasks, dynamic adjustments can be made to ensure efficient completion of tasks and long endurance of UAVs.

[0085] As an implementation method, the multi-UAV system task allocation and endurance optimization method includes S8: a temporary task processing step; Temporary task modeling: define temporary tasks, clearly define the attributes of temporary tasks, and provide clear input for subsequent processing; Task reallocation: When receiving a temporary task, re-use the genetic algorithm to optimize the task allocation plan and assign the temporary task to the appropriate drone. Re-optimize the task allocation through the genetic algorithm to ensure that the temporary task can be assigned to the most suitable drone, make full use of existing drone resources, improve the overall task execution efficiency, and ensure that the optimization process is carried out in a direction that is more in line with actual needs by updating the fitness function; Fitness function update: Update the fitness function to take into account the impact of temporary tasks and ensure that the evaluation criteria can fully consider the impact of temporary tasks; Path replanning: Based on the new task allocation plan, the A* algorithm is used to regenerate the flight path of each drone; Energy consumption prediction update: The random forest model is reused to predict the energy consumption of each drone, including the impact of temporary missions. By reusing the random forest model and considering the impact of temporary missions, the accuracy of energy consumption prediction is improved.

[0086] As an implementation method, temporary task modeling includes defining a temporary task T temp , including the position p temp , priority π temp , time required t temp ; Temporary Tasks temp Including the path length L compared to the original task T temp And energy consumption E temp ; Total path length after adding temporary tasks: L new =L total +L temp ; Total energy consumption after adding temporary tasks: E new =E total +E temp ; The fitness function f(x) is updated when processing temporary tasks as follows: ; w4, w5, w6: The weight coefficients of each task after adding temporary tasks. The weight coefficients of each task after adding temporary tasks are calculated based on p i , π i ,t i d max , l max 、E j 、p temp , π temp and t temp Joint decision making; Repeat S4 genetic algorithm optimization task and S5 iterative optimization to select the best task allocation plan. Definition of weight coefficient: The specific value of the weight coefficient is based on previous flight experience and p i , π i ,t i d max , l max 、E j 、p temp , π temp and t temp It is calculated through a multi-UAV system. By recalculating the fitness function f(x), the optimization target can be adjusted dynamically, so that the task allocation plan is more adapted to the actual needs of the current temporary task, and the efficiency and quality of the overall task execution can be improved. Recalculating the fitness function can quickly respond to the emergence of temporary tasks, adjust task allocation and path planning in a timely manner, and improve the responsiveness and flexibility of the system. By recalculating the fitness function, UAV resources can be allocated more reasonably, ensuring that the task allocation and path planning of each UAV are optimal, and improving resource utilization. Example

[0087] Two drones D1 and D2 need to complete four tasks T1, T2, T3, and T4. The task allocation scheme x can be expressed as: x = [D1, D2, D1, D2]; Task T1 is assigned to drone D1; Task T2 is assigned to drone D2; Task T3 is assigned to drone D1; Task T4 is assigned to drone D2; The total time, total distance and total energy consumption of each drone to complete the task: Drone D1: Ttotal，1 = 10 minutes; L total，1 = 5 km; E total，1 =200 joules; Nobody D2: T total，2 = 15 minutes; L total，2 =7 km; E total，2 =250 joules; The weight coefficient is: w1=0.5, w2=0.3, w3=0.2; Then the fitness value f(x) of the task allocation scheme is: f(x)=0.5·(10+15)+0.3·(5+7)+0.2·(200+250) Calculate the parts: Time part: 0.5·(10+15)=0.5·25=12.5; Distance part: 0.3·(5+7)=0.3·12=3.6; Energy consumption: 0.2·(200+250)=0.2·450=90; Therefore, the fitness value f(x) is: f(x)=106.1.

[0088] As an implementation manner, the adaptive control strategy adjustment includes: When high energy consumption is predicted, the drone reduces its flight speed or chooses a gentle flight route; Reduce flight speed: v new =v current -Δv; v new : The speed after reduction; v current : Current flight speed; Δv: reduced flight speed; Choose a gentle flight route, reduce the altitude change of the drone, use the A* algorithm to regenerate the path, choose the path with the minimum altitude change and the low wind speed altitude layer; If energy consumption is predicted to be low, increase flight speed or choose a more direct flight route; Increase flight speed: v new =v current +Δv; v new : The speed after reduction; v current: Current flight speed; Δv: reduced flight speed; Choose a more direct flight path and use the A* algorithm to regenerate the shortest path.

[0089] As an implementation method, S4 includes calculating the fitness value f(x i ), calculate the total fitness F total : ; N: population size, i.e. the number of individuals; f(x i ): the fitness value of the i-th individual; F total : The sum of all individual fitness values; Calculate individual x i The probability of choosing P(x i ): ; P(x i ): The probability of the i-th individual being selected; Generate a random number r∈[0,1] and add up the selection probability. When the accumulated value exceeds r, select the corresponding individual. Example

[0090] There are three individuals x1, x2, and x3, and their fitness values ​​are f(x1)=10, f(x2)=20, and f(x3)=30 respectively.

[0091] Calculate the sum of fitness: F total =10+20+30=60; Calculate the selection probability for each individual: P(x1)=1060=16≈0.167; P(x2)=2060=13≈0.333; P(x3)=3060=12=0.5; Generate a random number r: random number r=0.45.

[0092] Cumulative selection probability: Starting from the first individual, the selection probability is accumulated: The cumulative value of the selection probability of the first individual x1 is 0.167; The cumulative value of the selection probability of the second individual x2 is: 0.167+0.333=0.5; The cumulative value of the selection probability of the third individual x3 is: 0.5+0.5=1.0; When the cumulative value exceeds r=0.45, the corresponding individual is selected. The cumulative value exceeds 0.45 at the second individual x2, so individual x2 is selected.

[0093] Random number r: A random number in the interval [0, 1] used to simulate the roulette wheel selection process.

[0094] Cumulative value: Starting from the first individual, the selection probability of each individual is accumulated in turn until the cumulative value exceeds r, and the corresponding individual is selected.

[0095] In this way, individuals with high fitness values ​​are more likely to be selected, thereby gradually optimizing the population and finding the optimal solution. The reason for not directly selecting the maximum fitness value is to prevent the population from converging to a local optimal solution too early instead of a global optimal solution. Although individuals with higher fitness values ​​perform better, individuals with lower fitness values ​​may also contain some useful features, which can be retained through random selection.

[0096] Crossover includes: Single-point crossover: randomly select a crossover point k: k∈[1,n-1]; Generate two offspring individuals x c1 and x c2 : ; x p1 : The first parent individual represents a task allocation scheme; x p2 : The second parent individual represents a task allocation scheme; x p1，i : The gene of the first parent individual at position i, representing task T i which drone to assign to; x p2，i : The gene of the second parent individual at position i, representing task T i which drone to assign to; x c1 : The first offspring individual, representing a new task allocation scheme; x c2 : The second offspring individual represents a new task allocation scheme; k: the position of the crossover point, indicating that gene exchange starts from this position; n: the number of tasks, i.e. the length of the task allocation plan; Uniform crossover: Uniform crossover can generate more diverse offspring individuals by randomly selecting each gene position for exchange, thereby exploring new solution space; For each gene position i, generate a random number r in the interval [0, 1]i , set a crossover probability p c , according to the random number r i and the crossover probability p c , determines whether each gene position is exchanged. If r i <p c , then the gene position is exchanged, otherwise it remains unchanged; For each position i, with probability p c Exchange genes: ; p c : Crossover probability, a value in the interval [0, 1]; Variations include: With a certain probability p m Randomly select the task allocation scheme x of two positions to exchange genes, and for each position i with probability p m Choose another position j and swap genes: ; P m : Mutation probability, the value is between 0.01-0.1; x m，i : The gene of the individual at position i after mutation; x m，j : The gene of the individual at position j after mutation; x i : The gene of the original individual at position i; x j : The gene of the original individual at position j.

[0097] As an implementation method, the termination condition of S5 iterative optimization is to set the maximum number of iterations G. Based on previous experience and experimental results, a suitable maximum number of iterations is selected, and a fitness convergence threshold ϵ is set. When the change in fitness is less than ϵ or reaches the maximum number of iterations G, the iteration is stopped. The maximum number of iterations G means that the algorithm stops iterating after reaching the predetermined maximum number of iterations. This is to avoid an infinite loop of the algorithm and ensure that a solution is obtained within a limited time.

[0098] As an implementation method, the energy consumption data recorded in S7 includes the flight speed vcurrent, the flight altitude hcurrent, and the altitude change rate h′ current , load l current , wind speed w current , current power E current , path length L current , each record also includes the corresponding energy consumption E actual ,Task types include original task 1 and temporary task 0; Organize the energy consumption data into feature vector X current : ; Use the trained random forest model RF to predict the energy consumption E under current flight conditions predicted : ; RF: The trained random forest model is used to predict the energy consumption under the current flight conditions.

[0099] As an implementation method, key features including flight speed, altitude change rate and wind speed are extracted from the original data, and correlation analysis or feature importance evaluation is used to select key features that affect energy consumption.

[0100] The present invention discloses an application of a multi-UAV system task allocation and endurance optimization method, which is characterized by using any of the above-mentioned multi-UAV system task allocation and endurance optimization methods in the field of UAVs.

[0101] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all these improvements and changes should fall within the scope of protection of the appended claims of the present invention.

Claims

1. A method for task allocation and endurance optimization of a multi-UAV system, characterized in that: include: S1: Task modeling, defining the task set and drone set, each task and drone has corresponding attributes; S2: Initial population generation, randomly generating multiple task allocation schemes to form an initial population; S3: Fitness function design, define the fitness function; S4: Genetic algorithm optimization task, selection, crossover and mutation optimization task allocation scheme, and generate new task allocation scheme through selection, crossover and mutation operations; S5: Iterative optimization, repeating the above steps until the termination condition is met and selecting the optimal task allocation solution; S6: Path planning, based on the task allocation plan, the A* algorithm is used to generate the flight path of each UAV; S7: Data collection, prediction and adjustment. During the flight of the drone, the energy consumption data under various flight conditions is recorded, and key features are extracted from the raw data. The random forest model is trained using the training set data. During the flight, the random forest model is used to predict the energy consumption of each drone, and the flight strategy is adjusted according to the prediction results.

2. A method for task allocation and endurance optimization of a multi-UAV system according to claim 1, characterized in that: The task set in S1 defines all the tasks , each task T i There is position p i , priority π i , and the time required t i property; The drone set defines all drones , each drone D j The maximum flight distance d max , Maximum load l max , and the current power E j property; In S2, multiple task allocation schemes are randomly generated, and the multiple task allocation schemes form an initial population P0; The fitness function f(x) in S3 is defined as follows: ； x: represents a task allocation scheme in the initial population P0; w1, w2, w3: weight coefficients used to balance the importance of different objectives. The weight coefficients are based on p i , π i ,t i ,d max , l max and E j Joint decision making; T total :Indicates drone D j The total time required to complete the task; L total :Indicates drone D j The total flight distance required to complete the mission; E total :Indicates drone D j The total energy required to complete a task.

3. A method for task allocation and endurance optimization of a multi-UAV system according to claim 2, characterized in that: Includes S8: temporary task processing step; Temporary task modeling: define temporary tasks; Task reallocation: When receiving a temporary task, the genetic algorithm is used again to optimize the task allocation scheme and assign the temporary task to the appropriate UAV; Fitness function update: Update the fitness function to take into account the impact of temporary tasks; Path replanning: Based on the new task allocation plan, the A* algorithm is used to regenerate the flight path of each drone; Energy consumption prediction update: re-use the random forest model to predict the energy consumption of each drone, including the impact of temporary missions.

4. A method for task allocation and endurance optimization of a multi-UAV system according to claim 3, characterized in that: Temporary task modeling includes defining temporary tasks T temp , including the position p temp , priority π temp , time required t temp ; Temporary Tasks temp Including the path length L compared to the original task T temp And energy consumption E temp ; Total path length after adding temporary tasks: L new =L total +L temp ; Total energy consumption after adding temporary tasks: AND new =E total +E temp ; The fitness function f(x) is updated when processing temporary tasks as follows: ； w4, w5, w6: The weight coefficients of each task after adding temporary tasks. The weight coefficients of each task after adding temporary tasks are calculated based on p i , π i ,t i d max , l max 、E j 、p temp , π temp and t temp Joint decision making; Repeat S4 genetic algorithm optimization task and S5 iterative optimization to select the best task allocation solution.

5. The method for task allocation and endurance optimization of a multi-UAV system according to claim 3, characterized in that: The strategy adjustment of the adaptive control includes: When high energy consumption is predicted, the drone reduces its flight speed or chooses a gentle flight route; Reduce flight speed: v new =v current -Δv; v new : The speed after reduction; v current : Current flight speed; Δv: reduced flight speed; Choose a gentle flight route, reduce the altitude change of the drone, use the A* algorithm to regenerate the path, choose the path with the minimum altitude change and the low wind speed altitude layer; If energy consumption is predicted to be low, increase flight speed or choose a more direct flight route; Increase flight speed: v new =v current +Δv; V new : The speed after reduction; V current : Current flight speed; Δv: reduced flight speed; Choose a more direct flight path and use the A* algorithm to regenerate the shortest path.

6. The method for task allocation and endurance optimization of a multi-UAV system according to claim 1, characterized in that: S4 selects including, calculating the fitness value f(x) of all individuals in the current population i ), calculate the total fitness F total : ； N: population size, i.e. the number of individuals; f(x i ): the fitness value of the i-th individual; F total : The sum of all individual fitness values; Calculate individual x i The probability of choosing P(x i ): ； P(x i ): The probability of the i-th individual being selected; Generate a random number r∈[0,1] and add the selection probability. When the accumulated value exceeds r, select the corresponding individual. Crossover includes: Single-point crossover: randomly select a crossover point k: k∈[1,n-1]; Generate two offspring individuals x c1 and x c2 : ； x p1 : The first parent individual represents a task allocation scheme; x p2 : The second parent individual represents a task allocation scheme; x p1，i : The gene of the first parent individual at position i, representing task T i which drone to assign to; x p2，i : The gene of the second parent individual at position i, representing task T i which drone to assign to; x c1 : The first offspring individual, representing a new task allocation scheme; x c2 : The second offspring individual represents a new task allocation scheme; k: the position of the crossover point, indicating that gene exchange starts from this position; n: the number of tasks, i.e. the length of the task allocation plan; Even crossover: For each gene position i, generate a random number r in the interval [0, 1] i , set a crossover probability p c , according to the random number r i and the crossover probability p c , determines whether each gene position is exchanged. If r i <p c , then the gene position is exchanged, otherwise it remains unchanged; For each position i, with probability p c Exchange genes: ； p c : Crossover probability, a value in the interval [0, 1]; Variations include: With a certain probability p m Randomly select the task allocation scheme x of two positions to exchange genes, and for each position i with probability p m Choose another position j and swap genes: ； P m : Mutation probability; x m，i : The gene of the individual at position i after mutation; x m，j : The gene of the individual at position j after mutation; x i : The gene of the original individual at position i; x j : The gene of the original individual at position j.

7. The method for task allocation and endurance optimization of a multi-UAV system according to claim 1, characterized in that: The termination condition of S5 iterative optimization is to set the maximum number of iterations G and the fitness convergence threshold ϵ. When the change in fitness is less than ϵ or reaches the maximum number of iterations G, the iteration is stopped.

8. The method for task allocation and endurance optimization of a multi-UAV system according to claim 1, characterized in that: The energy consumption data recorded in S7 include flight speed v current , flight altitude h current , height change rate h′ current , load l current , wind speed w current , current power E current , path length L current , each record also includes the corresponding energy consumption E actual ,Task types include original task 1 and temporary task 0; Organize the energy consumption data into feature vector X current : ； Use the trained random forest model RF to predict the energy consumption E under current flight conditions predicted : ； RF: The trained random forest model is used to predict the energy consumption under the current flight conditions.

9. The method for task allocation and endurance optimization of a multi-UAV system according to claim 1, characterized in that: Extract key features from the raw data including flight speed, altitude change rate and wind speed; Select key features that affect energy consumption using correlation analysis or feature importance assessment.

10. An application of a multi-UAV system task allocation and endurance optimization method, characterized in that: Application of a multi-UAV system task allocation and endurance optimization method as described in any of claims 1-9 above in the field of UAVs.

Images