Unmanned aerial vehicle scheduling optimization method and system based on improved ascidian colony algorithm

Abstract

The unmanned aerial vehicle scheduling optimization method and system based on an improved ascidian colony algorithm belong to the technical field of low-altitude economic unmanned aerial vehicle scheduling, and solve the technical problems that the existing technology generally has slow convergence speed, is prone to falling into local optimization, and is difficult to balance the global search ability and local development ability in the solving process. Step 1, a network counting model of the unmanned aerial vehicle task is constructed; step 2, based on the network counting model, a target function of the team orientation problem is established; step 3, the improved ascidian colony algorithm is used to solve the team orientation problem, and the unmanned aerial vehicle scheduling scheme corresponding to the global optimal position is obtained. The present application is used for efficient convergence and global optimization of unmanned aerial vehicle scheduling.

Description

Technical Field

[0001] This invention relates to the field of low-altitude economic unmanned aerial vehicle (UAV) scheduling technology, specifically to a UAV scheduling optimization method and system based on an improved tunicate swarm algorithm. Background Technology

[0002] The core problem facing drone scheduling has shifted from single-point technological breakthroughs to systemic operational challenges. The drone scheduling problem is essentially an NP-hard (nondeterministic polynomial) combinatorial optimization problem, which means that as the number of drones and tasks increases, the solution space grows exponentially, making real-time solutions extremely difficult.

[0003] In the prior art, Chinese patent document CN117744994A discloses a "Surfing UAV-Nest Allocation and Scheduling Method Based on the Slug Algorithm" to determine the scheduling of inspection UAVs based on their flight cost and maintenance time. Nest allocation and scheduling objective function; determine the constraints of the objective function; establish multi-UAV scheduling. Multi-nest allocation and scheduling model; constructing a master based on the ability division of the population leader group. The approach involves an evolutionary mechanism based on heterogeneous thinking and a local optimum prevention mechanism designed using the principles of human brainstorming and collaborative learning strategies. However, this technical solution does not consider the constraint of repeated access to task points, and the algorithm improvement relies primarily on master-slave heterogeneous thinking and brainstorming, resulting in slow convergence speed, a tendency to get trapped in local optima, and difficulty in meeting the needs of large-scale UAV scheduling.

[0004] In summary, existing intelligent optimization algorithms generally suffer from technical problems such as slow convergence speed, easy getting trapped in local optima, and difficulty in balancing global search capability and local exploitation capability during the solution process. Summary of the Invention

[0005] This invention solves the technical problems commonly found in existing technologies, such as slow convergence speed, easy getting trapped in local optima, and difficulty in balancing global search capability and local exploitation capability.

[0006] The UAV scheduling optimization method based on the improved tunic swarm algorithm described in this invention includes the following steps: Step 1: Construct a network counting model for drone missions; Step 2: Based on the network counting model, establish the objective function for the team orientation problem; Step 3: The improved tunic group algorithm is used to solve the team orientation problem and obtain the UAV scheduling scheme corresponding to the globally optimal position.

[0007] Furthermore, in one embodiment of the present invention, the network counting model in step 1 is specifically as follows: Each task point within the drone mission area is set as a network node. Each network node is expanded into a corresponding number of virtual task replicas according to the preset maximum number of allowed accesses. Each virtual task replica has the same scoring attributes as the original network node, and each virtual task replica can be accessed at most once. The total number of all virtual task replicas constitutes the decision variable dimension of the team orientation problem.

[0008] Furthermore, in one embodiment of the present invention, the network counting model further includes: Construct an undirected weighted graph with network nodes as vertices and the shortest path distance between network nodes as edge weights, and calculate the shortest path distance matrix between all pairs of network nodes to constrain the flight time of the UAV.

[0009] Furthermore, in one embodiment of the present invention, the objective function for establishing the team orientation problem in step 2 is specifically as follows: The objective is to maximize the sum of scores for all virtual mission replicas accessed by all drones, while satisfying the following constraints: Each drone's flight path starts from and returns to the dispatch center. Each virtual mission copy is accessed at most once, and the total time for each flight path does not exceed the preset maximum time budget.

[0010] Furthermore, in one embodiment of the present invention, the improved tunicate group algorithm in step 3 includes the following steps: Initialize the population and velocity, calculate the fitness of each individual, divide the population into leaders and followers according to the fitness value, perform the leader update mechanism for individuals in the leader position, and perform the follower update mechanism for individuals in the follower position. Output the global optimal position after the termination condition is met.

[0011] Furthermore, in one embodiment of the present invention, the leader update mechanism specifically comprises: For each individual leader, a temporary position is generated according to the leader rules of the tunic group algorithm. The temporary position is subjected to a wave-drift perturbation with a preset probability. The perturbed temporary position is then fused with the individual's historical best and global best information through historical memory cross-fusion to obtain a new position.

[0012] Furthermore, in one embodiment of the present invention, applying a wave-drift disturbance to the temporary location with a preset probability specifically includes: ; ; in, For the drift step size, Let A be an independent random number that is uniformly distributed in the interval [0,1]. As a wave impact factor, This is a temporary location before the update. This is the upper bound of the search space. This is the lower bound of the search space.

[0013] Furthermore, in one embodiment of the present invention, the step of fusing the perturbed temporary position with the individual historical optimal and global optimal information through historical memory cross-fusion specifically involves: ; ; in, To memorize the guiding vector, Let j be the component of the globally optimal position in the j-th dimension. As a memory fusion factor, Let be the component of the historical best position of individual i in the j-th dimension. For mixed weights, Let be the probability of memory crossover.

[0014] The UAV scheduling optimization system based on the improved tunic swarm algorithm described in this invention includes the following modules: The network counting model building module constructs a network counting model for UAV missions. The objective function establishment module, based on the network counting model, establishes the objective function for the team orientation problem; The improved tunic swarm algorithm solution module uses the improved tunic swarm algorithm to solve the team orientation problem and obtain the UAV scheduling scheme corresponding to the globally optimal position.

[0015] This invention solves the technical problems commonly found in existing technologies, such as slow convergence speed, susceptibility to local optima, and difficulty in balancing global search capability with local exploitation capability. Specific beneficial effects of this invention include: 1. This invention proposes an UAV scheduling optimization method based on an improved tunic swarm algorithm. By introducing a wave-drift disturbance mechanism and a historical memory cross-fusion mechanism, the standard tunic swarm algorithm is deeply improved, effectively solving the above-mentioned technical problems, realizing a faster convergence speed and a stronger ability to escape local optima, and significantly improving the total score of UAV scheduling tasks.

[0016] 2. This invention proposes an UAV scheduling optimization method based on an improved tunic swarm algorithm. To prevent the algorithm from getting stuck in local optima, after generating temporary tunic swarm positions, a wave-driven drift disturbance is introduced with a certain probability. By simulating the large-scale drift phenomenon caused by the sudden impact of strong waves on tunic swarms, the asymmetric large-scale drift characteristics caused by ocean currents endow the algorithm with the ability to jump globally and escape local optima, overcoming the shortcomings of traditional algorithms that are prone to premature convergence. 3. This invention proposes a UAV scheduling optimization method based on an improved tunicate swarm algorithm. In order to introduce individual memory and social collaboration without maintaining a cumbersome velocity matrix, this invention uses the global optimal position as the basis vector, integrates the deviation between the individual's historical optimal position and the current temporary position, constructs a memory guidance vector, and achieves the collaboration between individual experience and social experience through weighted fusion. By introducing individual memory without maintaining a complex velocity matrix, this invention balances the autonomous exploration experience of individual tunicates with the social wisdom of the group, accelerates convergence, and improves the solution accuracy. Attached Figure Description

[0017] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart of the drone scheduling optimization method described in Implementation Method 1; Figure 2 This is a comparison chart of the parallel convergence curves of the multiple algorithms described in Implementation Method 2; Figure 3 This is a performance comparison chart of multiple algorithms described in Implementation Method 2. Detailed Implementation

[0018] Various embodiments of the present invention will now be clearly and completely described with reference to the accompanying drawings. The embodiments described with reference to the drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0019] Implementation Method 1: Currently, the existing technology has the following technical problems: 1. How to strike a balance between convergence speed and solution accuracy: Existing methods either converge quickly but have low accuracy, or have strong local search but weak global exploration. There is a lack of an optimization mechanism that can take both into account.

[0020] 2. How to avoid the algorithm getting trapped in local optima too early during the optimization process: Existing algorithms such as the tunic swarm algorithm and the whale optimization algorithm lack individual historical memory mechanisms, and the population diversity decreases rapidly with iteration, which prevents the overall task value of the scheduling scheme from being further improved.

[0021] To address the aforementioned technical problems, this implementation proposes an improved tunicate swarm optimization (M-IHSSA) algorithm for solving UAV scheduling under a network counting model, aiming to solve the UAV task allocation and path planning problem under multiple constraints. First, the real-world UAV scheduling problem is abstracted into a team orientation problem with time windows and repeated access constraints using a network counting model. Then, an improved tunicate swarm optimization algorithm is designed to efficiently search and optimize the team orientation problem. Figure 1 As shown, it specifically includes the following: Step 1: Construct a network counting model for drone missions; This step is responsible for abstracting the drone scheduling problem into building a network counting model that can be optimized by algorithms.

[0022] Step 11, Simulate UAV mission network data generation: Network node settings: Set the total number of network nodes to [number]. Network node 1 is fixed as the starting and ending point (dispatch center) for all drones. Network node 2 to... The coordinates are randomly generated on a two-dimensional plane.

[0023] Network node score calculation: for each network node The base score is not a fixed value, but is dynamically generated based on the task priority. For example, high-value nodes (accounting for 20%) are assigned a score of 1.5, ordinary nodes (accounting for 70%) are assigned a score of 0.8, and the scores of the remaining nodes are randomized between 0.8 and 1.5 to simulate the value differences of real tasks.

[0024] Maximum number of accesses: per network node Configure a maximum number of allowed accesses For example, according to probability From the set The mid-sampling parameter represents the number of times a task point can be served repeatedly. This parameter can be imported from an external CSV / JSON file or database.

[0025] Step 12, Network connectivity graph construction: Construct an undirected weighted graph based on the coordinates of network nodes. .side weight Represents the node arrive The shortest path distance.

[0026] Calculate the shortest path distance matrix between all pairs of network nodes. This step is used to calculate the time cost of all subsequent paths. It simulates flight constraints in real airspace, rather than simple straight-line distances.

[0027] Step 13, Node Expansion: Traditional methods for achieving multiple node visits typically employ static node replication or directly relaxing inflow constraints. These methods are ineffective and lead to several problems: First, node replication causes the network size to expand exponentially, resulting in a surge of decision variables and constraints, easily triggering the curse of dimensionality. Second, blindly relaxing flow restrictions without dynamic tracking of sequences and time can cause UAVs to loop between nodes during scheduling, producing ineffective symmetric solutions. Ultimately, this can lead to the solver's branch and bound tree becoming unsearchable due to the infinite expansion of the search space, resulting in an exponential combinatorial explosion in computation time. This significantly reduces the system's real-time responsiveness, hindering the rapid decision-making process when implementing scheduling schemes in practical engineering applications.

[0028] Therefore, to solve the problem of repeated node access mentioned above, this implementation proposes a node expansion strategy. Each original network node... Based on its maximum number of visits Expand as There are 10 virtual task copies with the same scoring attribute. The total number of all virtual copies. These constitute the decision variable dimensions for team orientation issues.

[0029] This process transforms the original problem into a standard team orientation problem variant, where each virtual mission copy can be accessed at most once, thus handling the repeated access constraint and forming the basis of the drone team orientation problem.

[0030] The node expansion strategy proposed in this implementation breaks the "only one visit" limitation of traditional team-oriented routing problems at the modeling level. It logically elevates the complex multi-visit topology in the early stages, transforming it into a standard single-visit routing problem. This leads to better synergy with subsequent algorithmic steps: it streamlines the coding logic, allowing the solution algorithm to directly utilize traditional settings and avoids runtime redundant node conflict checks; it implements structured natural pruning, directly transforming the maximum number of visits upper limit into the graph's boundary skeleton, eliminating numerous dynamic hard checks during algorithm iterations. This fundamentally resolves the computational redundancy caused by the exponential growth of the search space, improves the algorithm's convergence speed, and gives the system faster real-time response capabilities in scheduling.

[0031] Step 2: Based on the network counting model, establish the objective function for the team orientation problem; Equation (1) is the objective function, which aims to maximize the collected target value scores: (1) in, To find the maximum value of the objective function, For the first One drone, for The collection of drones For the first 1 node for The set of nodes The score that can be obtained by accessing the target node. For binary decision variables. If the drone... Node accessed ,but Otherwise .

[0032] The objective function satisfies the following constraints: Constraints (2) and (3) ensure that the flight path of each UAV originates from the depot (takeoff and landing field / base) and ultimately returns to the depot: (2) (3) in, For the first 1 node For binary decision variables. If the drone... From the base Fly directly to the target node , then Otherwise . For binary decision variables. If the drone... From the target node Fly directly back to base , then Otherwise .

[0033] Constraint (4) ensures that the flow of each target node is conserved: (4) in, and drones From node Fly to the node and from the node Fly to the node The flight segment, Let i be the set of all nodes i. For drones Do you want to access the node? , For node labels.

[0034] Constraint (5) ensures that all UAVs access each extended target node no more than once: (5) Constraint (6) Based on the maximum time budget, strictly limit the total duration of each route, including the flight time corresponding to the segment distance and the service time of each access node: (6) in, and These represent the flight times between different nodes. The service time required to execute the task at each target node. The time budget allowed for a single drone.

[0035] Constraint (7) eliminates the Miller-Tucker-Zemlin (MTZ) subloop constraint, while simultaneously tracking the arrival time of each node based on time-of-flight: (7) in, and As a continuous decision variable, representing the drone Reaching the node or The specific time.

[0036] Finally, constraint (8) defines the range of values ​​for the decision variables: .

[0037] Step 3: The improved tunic group algorithm is used to solve the team orientation problem and obtain the UAV scheduling scheme corresponding to the globally optimal position.

[0038] The improved tunic swarm algorithm (M-IHSSA) based on ocean dynamics described in this embodiment integrates the chain-guided capability of the tunic swarm algorithm (SSA), the global jump mechanism based on wave-driven drifting disturbances, and the collaborative mechanism of individual memory and social experience through cross-fusion of historical memory. Its core process is as follows: I. Algorithm Initialization Phase Parameter configuration: Set the basic running parameters of the algorithm: maximum number of iterations Population size Problem Space Dimension Upper bound of the search space and the lower world And control parameters: leader ratio Mixed weights Probability of drifting with the tide Wave impact factors Memory crossover probability Memory fusion factor .

[0039] State initialization: Randomly generate the initial location of the salps population within the search boundary. .

[0040] Initial evaluation: Calculate the objective function value (fitness) for each initial individual, and use this value to initialize the historical best position (individual memory) of each individual. The globally optimal position of the entire group (social experience) At the same time, the location of the currently best individual is set as the food source location. : .

[0041] II. Iterative Search Phase The algorithm begins by performing a loop optimization, in each iteration... ( In this context, the specific steps are as follows: 1. Dynamic core parameter update Calculate nonlinear time-varying control parameters This parameter decays exponentially with increasing iteration count, serving to balance global exploration and local exploitation capabilities. ; in, This is the preset maximum number of iterations.

[0042] 2. Population Individual Location Update The algorithm iterates through every individual in the population. ( ), and classify them as "leaders" or "followers" according to their index order: Scenario A: Leader update mechanism (when individual index) At that time, among them, for, (Preset leader ratio) For each dimension of the current leader ( , (Assuming the search space has dimensions), the following three-layer composite update is performed sequentially: First layer: Basic position update Generate independent and identically distributed random numbers. Introducing parameters As the probability of direction determination, The probability determines whether an individual iterates towards the food source in the positive or negative direction, calculating the temporary position. : ; Second layer: Drifting and disturbed by the waves To simulate the physical phenomenon of salps being passively drifted by sudden, strong waves in the ocean, a random probability value is introduced. .like Less than the preset probability of drifting with the tide Then, the salps will deviate from its normal path and drift over a wide area.

[0043] The drift step size is determined by a wave function that follows an asymmetric long-tail characteristic, which gives the algorithm a strong transition capability, forcing individuals to escape local optima: ; ; in, For the drift step size, Let A be an independent random number that is uniformly distributed in the interval [0,1]. As a wave impact factor, This is a temporary location before the update. This is the upper bound of the search space. This is the lower bound of the search space.

[0044] The third layer: Historical Memory Crossover Fusion In this step, the algorithm does not rely on a cumbersome velocity matrix, but instead allows the individual salps to be in its "current exploration position" and "individual's best historical memory (…). ")" and "social shared experience ( The two systems will cross and integrate intelligence.

[0045] First, by combining social experience with the targeted fusion of individual historical memories, a memory guidance vector is generated. (in (Determines the depth of memory fusion) ; Subsequently, the crossover probability was memorized. The decision is made on whether to accept the guidance of this historical memory; if a crossover is triggered, then mixed weights are used. Complete the final new position Merger: ; in, To memorize the guiding vector, Let j be the component of the globally optimal position in the j-th dimension. As a memory fusion factor, Let be the component of the historical best position of individual i in the j-th dimension. For mixed weights, Let [the number] be a random number uniformly distributed in the interval [0,1]. Let be the probability of memory crossover.

[0046] Scenario B: Follower update mechanism (when individual index) (At time) Followers closely follow their predecessors through a cooperative mechanism, simulating the cluster movement of a tunic chain. Their new position Based on its current position and the new position of the previous individual Joint decision (each assigned) and (fixed weighting factor) .

[0047] 1. Boundary handling and fitness assessment Boundary constraint truncation: Perform boundary checks on the updated individual positions. If they exceed preset upper and lower bounds, forcibly truncate them to the boundary to ensure the validity of the solution. ; Calculate fitness: transform the legalized position vector Substitute into the objective function to calculate the current fitness value. : .

[0048] 2. Dynamic update of optimal state Updating individual historical memory ( If the current fitness is better than the historical best value recorded for this individual. Then, its individual historical memory is updated: ; Update social sharing of experiences and food sources ( If the current fitness is better than the global optimum of the entire population. Then, the social best experience and food source location will be updated simultaneously: ; 3. Record the convergence trajectory After the entire population completes this round of iterative updates, the globally optimal score for this round is recorded in the convergence curve vector: .

[0049] III. Algorithm Termination and Output When the iteration loop reaches the maximum number of iterations Then, the algorithm stops. Finally, it outputs the optimal solution to the objective function obtained in this optimization (the globally optimal fitness value). and the complete convergence curve vector .

[0050] In summary, to prevent the algorithm from getting trapped in local optima, this implementation method introduces a certain probability of wave-driven perturbation after generating the temporary tunic location. .in The asymmetric, large-scale drift characteristics caused by ocean currents endow the algorithm with the ability to jump globally and escape local optima.

[0051] To introduce individual memory and social collaboration without maintaining a cumbersome velocity matrix, the algorithm calculates a guidance vector. And satisfy the crossover probability At that time, perform weighted merging: This operation balances the individual salps' autonomous exploration experience with the collective social wisdom.

[0052] Implementation Method 2: This implementation method compares M-IHSSA with Genetic Algorithm (GA), Whale Algorithm (WOA), Harris Eagle Algorithm (HHO), and Standard Salicylicer Group Algorithm (SSA). The experiment uses the same parameter settings (250 nodes, 22 drones, maximum duration 200, 400 iterations).

[0053] Convergence curve analysis: such as Figure 2 As shown, GA, WOA, and HHO all exhibit premature convergence to varying degrees. While the standard SSA continues to improve, M-IHSSA surpasses the final convergence value of SSA from the 50th generation onwards and maintains its lead throughout, ultimately achieving the highest score. This demonstrates that M-IHSSA possesses both rapid convergence speed and excellent ability to escape local optima.

[0054] Figure 2 This graph compares the parallel convergence curves of various algorithms (GA, WOA, HHO, SSA, Particle Swarm Optimization (PSO), and M-IHSSA) under the same experimental conditions. The horizontal axis represents the number of iterations, and the vertical axis represents the total score of the optimal scheduling scheme found by the algorithm. The graph visually illustrates the optimization process of each algorithm over time. The M-IHSSA algorithm in this implementation consistently curves above the other curves and converges to the highest score at the fastest speed, verifying its dual advantages in solution accuracy and convergence speed.

[0055] Performance Quantitative Comparison: Such as Figure 3 As shown, the optimal total score obtained by the M-IHSSA algorithm is 116.41, which is 7.8% higher than the standard SSA (108.02) and 22.5% higher than GA (95.15). This result fully demonstrates the effectiveness and superiority of the improved strategy proposed in this embodiment in solving complex UAV scheduling problems.

[0056] Figure 3A comparison of the final performance of multiple algorithms is typically presented as a bar chart or box plot. The chart clearly shows the optimal score, average score, and score distribution achievable by each algorithm after one or more independent runs. The M-IHSSA algorithm proposed in this implementation significantly outperforms other comparative algorithms in all statistical indicators, demonstrating that it not only has strong optimization capabilities but also provides stable and reliable solutions.

[0057] In summary, this implementation method systematically solves the multi-constraint problem in large-scale, multi-constraint UAV scheduling by using an improved tunic group algorithm, and has excellent solution accuracy, convergence speed and engineering practicality.

[0058] Implementation Method 3: This implementation method discloses a specific implementation plan for the collaborative scheduling optimization of low-altitude UAV clusters, which is applicable to large-scale UAV operation scenarios such as low-altitude logistics delivery, power line inspection, and security patrol. It solves the practical difficulties of traditional scheduling, such as the lack of standardized topology modeling, chaotic multi-task collaboration, and low solution efficiency.

[0059] This implementation transforms the UAV physical operation scheduling problem into an integer value network topology model, with the actual operation tasks as network nodes, each node having task benefits and time requirements; and the compliant and passable flight paths between any two task nodes as network edges, with each path having a specified flight mileage and travel time, thus completing the construction of the scheduling network.

[0060] Based on this topology, and considering practical constraints such as UAV endurance, non-repeating mission execution, and exclusive airspace paths, a team-oriented optimization problem specifically for low-altitude UAVs is constructed. Taking into account the requirements of maximizing mission benefits, minimizing flight energy consumption, and minimizing path risk, a multi-objective function is constructed, with the goal of maximizing the numerical value of the objective function. Subsequently, an improved tunicate swarm optimization algorithm is employed to quickly avoid local optima, and the optimal scheduling scheme is output after iterative convergence.

[0061] Implementation Method 4: A UAV scheduling optimization system based on an improved tunicate swarm algorithm, comprising the following modules: The network counting model building module constructs a network counting model for UAV missions. The objective function establishment module, based on the network counting model, establishes the objective function for the team orientation problem; The improved tunic swarm algorithm solution module uses the improved tunic swarm algorithm to solve the team orientation problem and obtain the UAV scheduling scheme corresponding to the globally optimal position.

[0062] The above provides a detailed description of the UAV scheduling optimization method and system based on the improved tunic swarm algorithm proposed in this invention. Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A method for optimizing UAV scheduling based on an improved tunicate swarm algorithm, characterized in that, Includes the following steps: Step 1, construct a network counting model for UAV missions. Specifically, the network counting model is as follows: Each task point within the drone mission area is set as a network node. Each network node is expanded into a corresponding number of virtual task replicas according to the preset maximum number of allowed accesses. Each virtual task replica has the same scoring attributes as the original network node, and each virtual task replica can be accessed at most once. The total number of all virtual task replicas constitutes the decision variable dimension of the team orientation problem. Step 2: Based on the network counting model, establish the objective function for the team orientation problem; Step 3: The improved tunic group algorithm is used to solve the team orientation problem and obtain the UAV scheduling scheme corresponding to the global optimal position; The improved tunicate group algorithm in step 3 includes the following steps: Initialize the population and velocity, calculate the fitness of each individual, divide the population into leaders and followers according to the fitness value, perform the leader update mechanism for individuals in the leader position, and perform the follower update mechanism for individuals in the follower position. Output the global optimal position after the termination condition is met. The leader update mechanism is as follows: For each individual leader, a temporary position is generated according to the leader rule of the tunic group algorithm. The temporary position is subjected to a wave-drift perturbation with a preset probability. The perturbed temporary position is fused with the individual's historical best and global best information through historical memory cross-fusion to obtain a new position. The process of applying a wave-drift disturbance to the temporary location with a preset probability is specifically as follows: ； ； in, For the drift step size, Let A be an independent random number uniformly distributed in the interval [0,1]. As a wave impact factor, This is a temporary location before the update. This is the upper bound of the search space. This serves as the lower bound of the search space. The process of fusing the perturbed temporary position with individual historical optimal and global optimal information through historical memory cross-fusion is specifically as follows: ； ； in, To memorize the guiding vector, Let j be the component of the globally optimal position in the j-th dimension. As a memory fusion factor, Let be the component of the historical best position of individual i in the j-th dimension. For mixed weights, Let be the probability of memory crossover.

2. The UAV scheduling optimization method based on the improved tunic swarm algorithm according to claim 1, characterized in that, The network counting model also includes: Construct an undirected weighted graph with network nodes as vertices and the shortest path distance between network nodes as edge weights, and calculate the shortest path distance matrix between all pairs of network nodes to constrain the flight time of the UAV.

3. The UAV scheduling optimization method based on the improved tunic swarm algorithm according to claim 1, characterized in that, The objective function for establishing the team orientation problem in step 2 is as follows: The objective is to maximize the sum of scores for all virtual mission replicas accessed by all drones, while satisfying the following constraints: Each drone's flight path starts from and returns to the dispatch center. Each virtual mission copy is accessed at most once, and the total time for each flight path does not exceed the preset maximum time budget.

4. A UAV scheduling optimization system based on an improved tunicate swarm algorithm, the system being used to implement the method of claim 1, characterized in that, Includes the following modules: The network counting model building module constructs a network counting model for UAV missions. The objective function establishment module, based on the network counting model, establishes the objective function for the team orientation problem; The improved tunic swarm algorithm solution module uses the improved tunic swarm algorithm to solve the team orientation problem and obtain the UAV scheduling scheme corresponding to the globally optimal position.

Images