A resource scheduling method for unmanned aerial vehicle assisted mobile edge computing
By combining the Q-learning algorithm with the cooperative evolutionary genetic algorithm to optimize UAV resource scheduling, the problem of high latency for user tasks under large-scale UAV swarms is solved, achieving fast convergence and efficient resource allocation.
Patent Information
- Application Number
- CN202310381066.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-04-11
- Publication Date
- 2026-02-17
- Estimated Expiration
- 2043-04-11
AI Technical Summary
In existing drone-assisted mobile edge computing, how to effectively reduce user task processing latency under large-scale drone swarms is a challenge. Existing algorithms are effective on a small scale but have poor scalability, and intelligent evolution algorithms are prone to getting trapped in local optima.
By combining the Q-learning algorithm with an improved co-evolutionary genetic algorithm, UAV resource scheduling is performed through prediction, deployment, and scheduling modules. The co-evolutionary genetic algorithm is used to optimize latency, and the Q-learning algorithm is combined with the strategy update to achieve efficient allocation of UAV resources.
It enables rapid convergence to a better solution in large-scale drone swarms, reduces user task latency, improves the convergence performance and exploratory nature of resource scheduling, and avoids local optima problems.
Smart Images
Figure CN116455451B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of unmanned aerial vehicle (UAV) resource allocation technology, and more particularly to a resource scheduling method for UAV-assisted mobile edge computing. Background Technology
[0002] Unmanned aerial vehicles (UAVs) originated in the 1920s, and their technology has seen unprecedented development in my country in recent years. Compared to driverless cars, UAVs have been applied in many fields such as agriculture, photography, and disaster relief. UAVs have gained immense popularity due to their fully or intermittent autonomy and low cost. The demand for civilian applications has prompted countries to vigorously develop UAV technology and actively expand its application areas. With the iterative updates of UAV technology, their manufacturing costs have continuously decreased, and their high mobility has ushered in new developments for mobile edge computing technology. However, UAVs have physical limitations such as communication range and battery power consumption. How to rationally utilize UAVs for computational assistance and how to effectively reduce latency through resource allocation when facing large-scale communication users and UAV swarms remains a challenge.
[0003] In existing resource scheduling for UAV-assisted mobile edge computing, most considerations focus on the flight path issues faced by the UAV during flight, as well as environmental factors during communication between the UAV and the user, such as signal attenuation and denial-of-service environments. These considerations also include energy consumption optimization, packet loss rate optimization, and latency optimization. However, minimizing the latency of user task processing using UAVs, which are low-cost and highly mobile nodes, remains a challenging problem. Current methods for minimizing latency in UAV-assisted mobile edge computing mainly fall into the following categories: 1. Using theoretical mathematical derivations to transform the problem step-by-step into a composite function formula. However, polynomial methods are not effective or even solve the problem well when dealing with large-scale combinatorial problems and NP-hard problems. 2. Using heuristic algorithms as scheduling methods. Heuristic algorithms perform well in small-scale UAV scheduling, but their effectiveness decreases as the number of UAVs increases. 3. Using intelligent evolutionary algorithms as scheduling methods. Intelligent evolutionary algorithms are suitable for large-scale combinatorial problems and NP-hard problems, offering unique advantages in large-scale UAV scheduling. However, they are prone to getting trapped in local optima, often resulting in mediocre output strategies. Summary of the Invention
[0004] In view of this, the purpose of this invention is to propose a resource scheduling method for UAV-assisted mobile edge computing, which combines a mobile edge computing system with UAV as the mobile node, Q-learning algorithm and improved CCGA method to solve the optimal allocation problem in UAV-assisted mobile edge computing.
[0005] The technical means employed in this invention are as follows:
[0006] A resource scheduling method for UAV-assisted mobile edge computing includes the following steps:
[0007] S1. Establish the basic framework of the UAV-assisted mobile edge computing model in dynamic task scenarios;
[0008] S2. Based on the resource scheduling method, the user computing tasks in the basic framework of the UAV-assisted mobile edge computing model are assigned tasks and resources are scheduled to obtain the scheduling scheme of UAV-assisted mobile edge computing. Finally, the UAV is replaced under the premise of ensuring that the energy can complete the next set of tasks.
[0009] Furthermore, the basic framework of the model described in S1 includes an estimation module, a deployment module, and a scheduling module;
[0010] The estimation module is used to obtain calculation task information and estimate the number of drones to be dispatched.
[0011] The deployment module is used to acquire user geographic information and deploy drones in designated areas.
[0012] The scheduling module is used to assign drones to perform corresponding tasks.
[0013] Furthermore, the working steps of the prediction module include:
[0014] S111, Loop Process: Generate a large number of drones N, and continuously reduce the number of drones during this process until one of the tasks cannot be completed under the delay requirement;
[0015] In S111, if the task cannot be completed, a co-evolutionary genetic algorithm is used to optimize the latency. If the number of drones after optimization meets the target, the number of drones is reduced; otherwise, the number of drones is increased.
[0016] Furthermore, the deployment module's operational steps include:
[0017] S121, Task Information Collection; used to obtain user's computing task request information;
[0018] S122, Geographic Information Collection; used to obtain the user's geographic location information and deploy drones;
[0019] S123, Area Coverage; The area coverage requirements are: the drone swarm covers all users, each drone covers at least one user, and each user can communicate;
[0020] S124. Deployment optimization: The block coordinate descent method is adopted, with the total number of users covered by the drone as the optimization target, so as to increase the total number of targets.
[0021] Furthermore, the working steps of the scheduling module include:
[0022] S131. Based on the geographic location information of the UAV after deployment collected in S123, the user geographic location information collected in S122, and the computing task request in S121, analyze the UAV resource scheduling scheme using a resource scheduling method.
[0023] S132. The UAV performs unloading, calculation, and data transmission of computing tasks according to the resource scheduling scheme;
[0024] S133. Computational task offloading, as a server resource scheduling mechanism, is represented by the user selecting the drone to be offloaded.
[0025] S134. Bandwidth resource scheduling is reflected in the allocation of bandwidth by the UAV for the tasks offloaded to it.
[0026] S135, Loop Process: After executing the current computational task, the drone swarm performs an energy check. If the energy requirements are met, the task continues; otherwise, the task ends and the next drone group is dispatched.
[0027] Furthermore, the resource scheduling method described in S2 includes:
[0028] S21. Using the co-evolutionary genetic algorithm as an auxiliary algorithm to the Q-learning algorithm, a combined algorithm is formed to jointly find resource scheduling schemes and task priority ranking schemes for the UAV swarm.
[0029] S22. The combined algorithm starts with the evolutionary algorithm, performs population preprocessing, quickly finds the best, and inherits the offspring into the Q-learning algorithm. The Q-learning algorithm uses this strategy as the initial strategy, uses chromosomes as agents, and interacts with the environment. The two algorithms work together alternately to find resource scheduling strategies.
[0030] Furthermore, the combination algorithm described in S21 is as follows:
[0031] Chromosomes, acting as intelligent agents, learn in their environment by changing their Q-values through action selection, thereby achieving the learning objective. Each task or state has N actions to choose from. The reward is determined by the change in fitness between two consecutive states and updates the Q-table in Q-learning.
[0032] The strategy of incorporating a co-evolutionary genetic algorithm and assigning it a larger Q-value weight ω improves the learning efficiency in the next iteration.
[0033] The output strategy of the co-evolutionary genetic algorithm changes the initial conditions of Q-learning, that is, the priority of the strategy changes, so that the chromosome of the Q-learning algorithm is constantly updated, increasing the exploratory nature of the algorithm.
[0034] Furthermore, the core principle of the chromosome is as follows:
[0035] Drone resource scheduling involves user task allocation schemes and task sequencing, and real number encoding is used to solve the encoding problem.
[0036] Chromosomes are represented by many genes, each gene representing the code for a task. The integer part represents the drone to be unloaded in the task, and the fractional part represents the priority of the task.
[0037] The entire chromosome iterates through the co-evolutionary genetic algorithm. At the algorithm interface, the current chromosome is processed to free it from the influence of coupling degree and enable it to interact with the environment as an intelligent agent.
[0038] The knowledge learned by the uncoupled chromosome during its interaction with the environment is reflected in the selection of drones for the task, and the learned strategy is ultimately output.
[0039] Furthermore, the processing principle of the interface is as follows:
[0040] The interface connects the co-evolutionary genetic algorithm with the reinforcement learning algorithm, and changes the chromosome through an encoded sorting interface; specifically, it sorts tasks according to task priority and generates new chromosomes.
[0041] Based on the encoding method of the new chromosome, the problem of future actions changing past choices is avoided; specifically, tasks are sorted according to task priority; the sorted tasks are then filled with time in a hierarchical manner.
[0042] Compared with the prior art, the present invention has the following advantages:
[0043] 1. The resource scheduling method provided by this invention achieves better convergence performance and exploratory nature by combining evolutionary algorithms and reinforcement algorithms.
[0044] 2. Evolutionary algorithms are characterized by fast convergence speed, utilizing large-scale populations for search to quickly find better solutions. However, they suffer from local optima, often stagnating in later iterations. Reinforcement learning algorithms leverage the powerful computing capabilities of computers, exhibiting greater exploratory nature, but their convergence speed is slow and they also suffer from non-convergence. This invention organically combines these two approaches, achieving both rapid convergence and finding superior unloading schemes, demonstrating significant advantages over other algorithms. Attached Figure Description
[0045] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0046] Figure 1 This is a schematic diagram of the overall model framework of the present invention.
[0047] Figure 2 This is a flowchart of the resource scheduling method of the present invention.
[0048] Figure 3 This diagram illustrates different encoding methods of the present invention.
[0049] Figure 4 This is a schematic diagram of the coding principle of the present invention.
[0050] Figure 5 This is a comparison chart of the algorithm results at different scales in this invention.
[0051] Figure 6 This is a comparison chart of the algorithm results under scale two of this invention.
[0052] Figure 7 This is a comparison diagram of different encoding methods of the present invention. Detailed Implementation
[0053] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.
[0054] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.
[0055] This invention provides a technical solution: a resource scheduling method for UAV-assisted mobile edge computing, the flowchart of which is shown below. Figure 1 , 2 As shown.
[0056] S1. Establish the basic framework for a drone-assisted mobile edge computing model in dynamic task scenarios.
[0057] S11, Prediction Module: Obtains calculation task information and uses it to predict the number of drones to be dispatched.
[0058] S111, Loop Process: Generate a large number of drones N, and continuously reduce the number of drones during this process until one of the tasks cannot be completed under the delay requirement.
[0059] S112. If the process cannot be completed, the algorithm is integrated into the co-evolutionary genetic algorithm. The optimization algorithm is used to optimize the delay. If the number of drones after optimization meets the target, the algorithm returns to the previous step; otherwise, it returns to the previous quantity strategy.
[0060] S12, Deployment Module: Obtains user geographic information and performs regional deployment of drones.
[0061] S121, Task Information Collection, used to obtain user's computing task request information.
[0062] S122, Geographic Information Collection, used to obtain the user's geographic location information and deploy drones.
[0063] S123. Area Coverage Requirements: The drone swarm must cover all users; a single drone must cover at least one user; and each user must be able to communicate.
[0064] S124. Deployment optimization: The block coordinate descent method is adopted, with the total number of users covered by the drone as the optimization target, so as to maximize the total number of targets.
[0065] S13, the scheduling module, is used to assign drones to perform corresponding tasks.
[0066] S131. Based on the geographic location information of the UAV after deployment, the user's geographic location information, and the computational task request, analyze the UAV resource scheduling scheme using a resource scheduling method.
[0067] S132. The UAV performs unloading, calculation, and data transmission of computing tasks according to the resource scheduling scheme.
[0068] S133. The unloading of computing tasks, as a form of server resource scheduling, is manifested in the user selecting the drone to be unloaded.
[0069] S134. Bandwidth resource scheduling is reflected in the allocation of bandwidth by the UAV for the tasks unloaded onto it.
[0070] S135, Loop Process: After executing the current computation task, the drone swarm performs an energy check. If the energy requirements are met, the task continues; otherwise, the task ends and the next drone group is dispatched.
[0071] S2. The scheduling scheme for drone-assisted mobile edge computing will allocate tasks and schedule resources for user computing tasks in the scene according to the resource scheduling method, and finally replace the drone while ensuring that the energy can complete the next set of tasks.
[0072] S21. Using a co-evolutionary genetic algorithm as an auxiliary algorithm to the Q-learning algorithm, we jointly search for resource scheduling schemes and task priority ranking schemes for the UAV swarm.
[0073] S211. Chromosomes, acting as intelligent agents, learn in the environment by changing their Q-values through action selection, thereby achieving the learning objective. Each task (state) has N actions to choose from (N drone nodes can choose from). The reward is determined by the change in fitness between two consecutive states (specifically, the fitness changes of the current and historical states, but excluding future states) and updates the Q-table in Q-learning.
[0074] S2111. UAV resource scheduling involves user task allocation schemes and task sorting, and real number encoding is used to solve the encoding problem.
[0075] S2112. A chromosome is represented by many genes, each gene representing the code of a task. The integer part represents the drone to be unloaded in the task, and the fractional part represents the priority of the task.
[0076] S2113. The entire chromosome iterates in the co-evolutionary genetic algorithm. At the algorithm interface (the connection between the two algorithms), the current chromosome needs to be processed to free it from the influence of coupling degree, so that it can act as an intelligent agent to interact with the environment.
[0077] S21131. The interface connects the co-evolutionary genetic algorithm and the reinforcement learning algorithm, and modifies the chromosome through an encoding and sorting interface. Specifically, tasks are sorted according to their priority to generate new chromosomes.
[0078] S21132. Based on the coding method of the new chromosome, the problem of future actions changing past choices can be avoided, such as... Figure 3 , 4 As shown. The specific solution is as follows: sort the tasks according to their priority; then fill in the time for the sorted tasks using a "building a house" method.
[0079] S2114. The "knowledge" learned by chromosomes without coupling during interaction with the environment is reflected in the selection of drones for tasks, and the learned strategy is ultimately output.
[0080] S212. At this point, a strategy of co-evolutionary genetic algorithm is added, and it is given a larger Q-value weight ω to improve the learning efficiency in the next iteration.
[0081] S213. The output strategy of the co-evolutionary genetic algorithm changes the initial conditions of Q-learning, that is, the priority of the strategy changes, so that the chromosome of the Q-learning algorithm is constantly updated, increasing the exploratory nature of the algorithm.
[0082] S22. Chromosomes, as the core of evolutionary algorithms, are an important tool for iterative optimization. In the combined algorithm, chromosomes are still used as the iterative core of the algorithm.
[0083] S23. The intelligent combination algorithm starts with the evolutionary algorithm, performs population preprocessing, quickly finds the best and inherits the offspring into the Q-learning algorithm. The Q-learning algorithm uses this strategy as the initial strategy, uses chromosomes as intelligent agents, and interacts with the environment. The two algorithms work together alternately to find resource scheduling strategies.
[0084] S24. The auxiliary role of the co-evolutionary genetic algorithm is: (1) to accelerate the convergence of the Q table (2) to find a task priority ranking strategy.
[0085] Example
[0086] This paper proposes a UAV-assisted mobile edge computing resource scheduling method based on an intelligent combination algorithm. The method uses an intelligent combination algorithm that combines an improved genetic algorithm, namely the cooperative evolutionary genetic algorithm, and the Q-learning algorithm, with the minimum latency at task completion as the optimization objective.
[0087] This embodiment conducts experiments in real-world mission scenarios, testing different sizes of drone swarms and the number of missions. The comparison algorithms used in this paper employ particle swarm optimization, cooperative evolutionary genetic algorithm, and Q-learning algorithm.
[0088] When the scale is 10 drones in a swarm and the number of tasks is 30, the comparison table of the four algorithms is shown in Table 1. Figure 5 As shown in Table 2, when the number of drones in scale 2 is 30 and the number of tasks is 150, the comparison of the four algorithms is shown in Table 2. Figure 6 As shown:
[0089] Table 1 Comparison of Scale-Assessment Results
[0090]
[0091] Table 2 Comparison of Assessment Results for Scale II
[0092]
[0093] Tables 1 and 2 show the results of the algorithm under the same number of evaluations. It can be seen that the intelligent combination algorithm outperforms other algorithms under the same number of evaluations.
[0094] Figure 7 The graph shows different encoding methods for chromosomes. The top line represents the encoding method used in the intelligent fusion algorithm, while the bottom line represents the encoding methods used in other evolutionary algorithms. As can be seen from the graph, the new encoding method performs better in reinforcement learning algorithms.
[0095] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.
[0096] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.
[0097] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.
[0098] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0099] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.
[0100] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.
[0101] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. A resource scheduling method for UAV-assisted mobile edge computing, characterized in that, Includes the following steps: S1. Establish the basic framework of the UAV-assisted mobile edge computing model in dynamic task scenarios; The basic framework includes an estimation module, a deployment module, and a scheduling module; The estimation module is used to obtain calculation task information and estimate the number of drones to be dispatched. The deployment module is used to acquire user geographic information and deploy drones in designated areas. The scheduling module is used to assign drones to perform corresponding tasks; S2. Based on the resource scheduling method, the user computing tasks in the basic framework of the UAV-assisted mobile edge computing model are assigned tasks and resources are scheduled to obtain the scheduling scheme of UAV-assisted mobile edge computing. Finally, the UAV is replaced under the premise of ensuring that the energy can complete the next set of tasks. The resource scheduling method includes: S21. Using the co-evolutionary genetic algorithm as an auxiliary algorithm to the Q-learning algorithm, a combined algorithm is formed to jointly find resource scheduling schemes and task priority ranking schemes for the UAV swarm. The combination algorithm is as follows: Chromosomes, acting as intelligent agents, learn from their environment by changing their Q-values through action selection, thereby achieving the goal of learning. Each task or state has... Each action can be selected; the reward is judged by the change in fitness between the two time points and the Q-table in Q-learning is updated; The strategy incorporates a co-evolutionary genetic algorithm and assigns it a larger Q-value weight. This will improve the efficiency of the next learning session; The output strategy of the co-evolutionary genetic algorithm changes the initial conditions of Q-learning, that is, the priority of the strategy changes, so that the chromosome of the Q-learning algorithm is constantly updated, increasing the exploratory nature of the algorithm. S22. The combined algorithm starts with the evolutionary algorithm, performs population preprocessing, quickly finds the best and inherits the offspring into the Q-learning algorithm. The Q-learning algorithm uses this strategy as the initial strategy, uses chromosomes as agents to interact with the environment, and the two algorithms work together alternately to find resource scheduling strategies.
2. The resource scheduling method for UAV-assisted mobile edge computing according to claim 1, characterized in that, The working steps of the prediction module include: S111, Loop Process: Generate a large number of drones N, and continuously reduce the number of drones during this process until one of the tasks cannot be completed under the delay requirement; In S111, if the task cannot be completed, a co-evolutionary genetic algorithm is used to optimize the latency. If the number of drones after optimization meets the target, the number of drones is reduced; otherwise, the number of drones is increased.
3. The resource scheduling method for UAV-assisted mobile edge computing according to claim 1, characterized in that, The deployment module's operational steps include: S121, Task Information Collection; used to obtain user's computing task request information; S122, Geographic Information Collection; used to obtain the user's geographic location information and deploy drones; S123, Area Coverage; The area coverage requirements are: the drone swarm covers all users, each drone covers at least one user, and each user can communicate; S124. Deployment optimization: The block coordinate descent method is adopted, with the total number of users covered by the drone as the optimization target, so as to increase the total number of targets.
4. The resource scheduling method for UAV-assisted mobile edge computing according to claim 1, characterized in that, The working steps of the scheduling module include: S131. Based on the geographic location information of the UAV after deployment collected in S123, the user geographic location information collected in S122, and the computing task request in S121, analyze the UAV resource scheduling scheme using a resource scheduling method. S132. The UAV performs unloading, calculation, and data transmission of computing tasks according to the resource scheduling scheme; S133. Computational task offloading, as a server resource scheduling mechanism, is represented by the user selecting the drone to be offloaded. S134. Bandwidth resource scheduling is reflected in the allocation of bandwidth by the UAV for the tasks offloaded to it. S135, Loop Process: After executing the current computational task, the drone swarm performs an energy check. If the energy requirements are met, the task continues; otherwise, the task ends and the next drone group is dispatched.
5. The resource scheduling method for UAV-assisted mobile edge computing according to claim 1, characterized in that, The core principle of the chromosome is as follows: Drone resource scheduling involves user task allocation schemes and task sequencing, and real number encoding is used to solve the encoding problem. Chromosomes are represented by many genes, each gene representing the code for a task. The integer part represents the drone to be unloaded in the task, and the fractional part represents the priority of the task. The entire chromosome iterates through the co-evolutionary genetic algorithm. At the algorithm interface, the current chromosome is processed to free it from the influence of coupling degree and enable it to interact with the environment as an intelligent agent. The knowledge learned by the uncoupled chromosome during its interaction with the environment is reflected in the selection of drones for the task, and the learned strategy is ultimately output.
6. The resource scheduling method for UAV-assisted mobile edge computing according to claim 5, characterized in that, The processing principle of the interface is as follows: The interface connects the co-evolutionary genetic algorithm with the reinforcement learning algorithm, and changes the chromosome through an encoded sorting interface; specifically, it sorts tasks according to task priority and generates new chromosomes. Based on the encoding method of the new chromosome, the problem of future actions changing past choices is avoided; specifically, tasks are sorted according to task priority; the sorted tasks are then filled with time in a hierarchical manner.