A demand-aware-driven collaborative optimization method and system for low-altitude dynamic airway networks

Abstract

This invention discloses a demand-aware driven collaborative optimization method and system for low-altitude dynamic airway networks, relating to the fields of low-altitude airspace management and intelligent transportation planning. The scheme involves: discretizing the low-altitude airspace into a three-dimensional directed network graph; constructing a multi-objective optimization model containing topology, capacity, and flow variables; initializing three groups (topology, capacity, and flow), a historical memory database, and an external database; determining the intensity of environmental changes based on a time-varying environmental detection function; and using a decoupled approach to collaboratively evolve the three groups to obtain candidate solutions; evaluating the candidate solutions for multiple objectives; performing local repair based on minimum cost and maximum flow for solutions that do not meet constraints; performing non-dominated screening and environmental feature extraction on the repaired solutions; and updating the external database and historical memory database; and outputting the non-dominated solution set in the external database when the termination condition is met. This invention achieves efficient collaborative optimization of airway networks under demand and environmental changes, improves search efficiency and solution feasibility, and ensures low-altitude operational safety.

Description

Technical Field

[0001] This invention relates to the fields of low-altitude airspace management and intelligent transportation planning, and in particular to a demand-aware-driven method and system for collaborative optimization of low-altitude dynamic airway networks. Background Technology

[0002] With the rapid development of applications such as the low-altitude economy, urban air traffic, drone logistics, emergency inspection, and electric vertical takeoff and landing (EVTOL) aircraft, low-altitude airspace is gradually shifting from a traditional low-density, manually controlled operation mode to a high-density, networked, and intelligent collaborative operation mode. To improve the efficiency of low-altitude airspace resource utilization, related technologies are beginning to incorporate ground traffic network optimization, air traffic flow management, multi-objective path planning, dynamic capacity management, and intelligent optimization algorithms into the low-altitude airway network planning process. By constructing mapping relationships between airspace nodes, flight segments, sectors, and demand flows, unified modeling of flight paths, traffic capacity, and operational flow is achieved. Simultaneously, with the development of meteorological sensing, communication, navigation, and surveillance, digital twin airspace, and intelligent scheduling platforms, the low-altitude airway network is gradually acquiring the technological foundation for dynamic adjustments based on traffic demand, meteorological conditions, control constraints, and changes in airspace capacity, providing support for the safe and efficient operation of low-altitude aircraft in complex and time-varying environments.

[0003] However, existing low-altitude airway network optimization technologies still struggle to meet the demands of collaborative optimization under high-density, highly time-varying, and multi-constraint conditions. First, existing methods often focus on single-route planning or static network capacity allocation, typically treating route topology, segment capacity, and traffic allocation as relatively independent problems. They lack a unified optimization mechanism addressing the coupling relationship among these three elements, making it difficult to simultaneously consider demand service rates, network stability, and congestion delays when traffic demand surges, local capacity decreases, or airspace is temporarily restricted. Second, existing optimization methods are insufficiently responsive to environmental changes, often relying on fixed-period replanning or single global searches. They fail to effectively distinguish between stable changes and drastic jumps, resulting in slow convergence speeds and insufficient reuse of historical experience under dynamic demand and capacity conditions. Furthermore, existing multi-objective optimization schemes often employ a unified coding method to handle mixed decision variables involving segment initiation / departure, capacity allocation, and flow distribution. The large number of infeasible solutions in the high-dimensional search space leads to severe waste of computational resources. When candidate solutions violate constraints on segment capacity, node flow conservation, or sector capacity, solutions are easily eliminated or regenerated, resulting in high computational costs and difficulty in retaining effective information from existing feasible topologies and capacity configurations. In addition, existing archive maintenance and historical memory mechanisms are inadequate, making it difficult to accumulate, filter, and reuse high-quality solution sets formed under different environmental characteristics, thus failing to continuously provide prior guidance for rolling optimization in subsequent periods.

[0004] How to solve the above-mentioned technical problems is the challenge facing this invention. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a demand-aware-driven method and system for collaborative optimization of low-altitude dynamic airway networks, which achieves efficient collaborative optimization of airway networks under changing demands and environments, improves search efficiency and solution feasibility, and ensures low-altitude operational safety.

[0006] The technical solution adopted by this invention to solve its technical problem is as follows: This invention provides a demand-aware driven collaborative optimization method for low-altitude dynamic airway networks, comprising the following steps: S1. Discretize the low-altitude airspace into a three-dimensional directed network graph, and construct a multi-objective optimization model for the low-altitude airway network by defining system decision variables, constructing a dynamic multi-objective function, and setting physical and air traffic control safety constraints. S2. Based on the multi-objective optimization model of low-altitude airway network, define and initialize the topology population, capacity population, and flow population. At the same time, construct a historical memory bank and an external archive. The historical memory bank is used to store the environmental characteristics in the population evolution and their corresponding optimization solution set. The external archive is used to save the repaired airway network scheme and non-dominated screening results. S3. Determine the intensity of environmental change based on the preset environmental time-varying detection function, and determine the evolution strategy of the current population based on the intensity of environmental change; according to the evolution strategy of the current population, use a decoupling method to perform co-evolution of the topology population, capacity population and flow population to obtain candidate route network schemes. S4. By calculating the fitness, perform multi-objective evaluation on the candidate airway network schemes, and perform local flow repair on the candidate airway network schemes that do not meet the physical and air traffic control safety constraints to obtain the repaired airway network schemes. S5. Perform non-dominated screening and environmental feature extraction on the repaired route network scheme to obtain non-dominated screening results and environmental features. Update the external archives and historical memory bank based on the repaired route network scheme, non-dominated screening results and environmental features. S6. When the preset termination condition is met, stop the population evolution and output the non-dominated solution set in the external archive as the result of low-altitude dynamic airway network collaborative optimization for the current period.

[0007] Preferably, the system decision variables in step S1 include topology state variables, capacity configuration variables, flow allocation variables, and admission flow variables; the topology state variables are used to represent the segment start / stop status; the capacity configuration variables are used to represent the segment's physical capacity; the flow allocation variables are used to represent the flow allocation of demand on the segment; and the admission flow variables are used to represent the actual flow demand; in step S2, the topology population adopts a binary encoding method, the capacity population adopts a real number encoding method, and the flow population adopts a real number matrix encoding method.

[0008] Preferably, the dynamic multi-objective function in step S1 includes a first minimization objective, a second minimization objective, and a third minimization objective; The first minimization objective is to minimize the total amount of demand rejection, expressed by the following formula:

[0009] in, For the current decision-making period, For the requirements of indexing, For the set of all OD demand pairs, For the first Time period requirements The original traffic demand, For the first Actual demand for admission during the time period Traffic; The second minimization objective is to minimize the overall cost of dynamic network reconstruction, expressed by the following formula:

[0010] in, The preset reconstruction cost weighting coefficient, For flight segment index, For segment collection, For the first Time Segment Start-stop status, For the first Time Segment Start-stop status, For the first Time Segment Physical capacity For the first Time Segment Physical capacity; The third minimization objective is to minimize the overall network congestion delay, expressed by the following formula:

[0011] in, For the first Time period requirements In the flight segment Traffic allocation on For the segment Free circulation time, For congestion sensitivity coefficient, This is a very small positive number used for congestion delay calculation; The physical and air traffic control safety constraints include resource capacity constraints, elastic node flow conservation constraints, and sector capacity constraints. The resource capacity constraint formula is expressed as follows:

[0012] in, For the first Time Segment The dynamic maximum capacity; The formula for the elastic nodal flow conservation constraint is expressed as follows:

[0013] in, and They represent respectively with A collection of flight segments with a starting point and an ending point; The sector capacity constraint formula is expressed as follows:

[0014] in, To belong to the sector The collection of flight segments, For the set of all sectors, For the first Time Sector The air traffic control safety threshold.

[0015] Preferably, the preset environmental time-varying detection function formula in step S3 is expressed as follows:

[0016] in: and These represent the traffic demand vectors for the current time period and the previous time period, respectively. and These represent the upper limit vectors of the physical capacity of the flight segment in the current time period and the previous time period, respectively; These are the preset environmental detection weighting coefficients; The value should be a very small positive number to prevent the denominator from being zero. Represents the L2 norm; The evolutionary strategy for determining the current population based on the intensity of environmental change includes: When the intensity of environmental change is less than or equal to the preset threshold, the environment is determined to be stable, and the non-dominated solution set in the external file of the previous period is injected into the current topology population, capacity population and flow population according to the preset ratio. When the intensity of environmental change exceeds a preset threshold, it is determined to be a drastic environmental change. The non-dominated solution set corresponding to the top few historical scenarios with the highest similarity to the current environmental features is extracted from the historical memory bank and injected into the current topology population, capacity population and flow population according to the preset injection ratio to complete the population reset. The cooperative evolution is performed in a decoupled manner, specifically including: Genetic mutation operations are performed independently on the topological population, capacity population, and flow population respectively. For any topological individual in the topological population, the compatibility with the capacity individual in the capacity population is calculated, and the capacity individual with the highest compatibility with the topological individual is selected to form a topological capacity combination. The topology capacity combination is paired with the traffic population to determine a traffic individual, and candidate route network schemes are determined based on the topology capacity combination and the traffic individual.

[0017] Preferably, the local flow repair in step S4 includes: For candidate airway network schemes that do not meet the physical and air traffic control safety constraints, the topology and capacity individuals remain unchanged, and they are reduced in dimension and mapped to a linear impedance network. The linear impedance network uses the currently active air segments as traversable edges, the available remaining capacity of the air segments as the capacity limit, and the free passage time of the air segments as the cost. The starting point and the ending point correspond to the source and sink points of each demand. The minimum cost maximum flow algorithm is called to redistribute the flow in the linear impedance network to obtain the repaired flow individuals. The repaired flow individuals are then used to replace the flow individuals in the candidate route network scheme to obtain the repaired route network scheme. The repaired route network scheme is constrained and verified. If all constraints are met, it is retained; otherwise, it is marked as unrepairable and eliminated in the current generation of evolution.

[0018] Preferably, the environmental features in step S5 include a normalized current time period traffic demand vector and a normalized current time period segment capacity limit vector; When the historical memory bank reaches a preset storage limit, it is updated using an eviction mechanism; the eviction mechanism specifically includes: When each memory entry is stored in the historical memory bank, an applicability weight is assigned to the memory entry; the applicability weight is attenuated according to a preset attenuation coefficient. When a memory entry is extracted for population reset during the co-evolution process in step S3, if the non-dominated solution set of the memory entry has not entered the first layer of the non-dominated frontier in several consecutive generations of evolution, the applicability weight of the memory entry is reduced by a preset penalty factor. When the applicability weight is lower than the preset elimination threshold, the memory entry is deleted.

[0019] Preferably, the elimination mechanism further includes: During the historical memory bank update process, the similarity between the current environmental features and the environmental features of existing memory entries is calculated. If the similarity exceeds a preset similarity threshold, no new or deleted memory entries are added. Instead, the non-dominated solution set in the existing memory entries is replaced with the non-dominated solution set corresponding to the current environmental features.

[0020] This invention also provides a demand-aware driven low-altitude dynamic airway network cooperative optimization system, comprising: The model building module is used to discretize the low-altitude airspace into a three-dimensional directed network graph, and to build a multi-objective optimization model of the low-altitude airway network by defining system decision variables, constructing dynamic multi-objective functions, and setting physical and air traffic control safety constraints. The initialization module is used to define and initialize the topology population, capacity population, and flow population based on the multi-objective optimization model of the low-altitude airway network. At the same time, it builds the history memory bank and external archives. The population evolution module is used to determine the intensity of environmental change based on a preset environmental time-varying detection function, and to determine the evolution strategy of the current population based on the intensity of environmental change; based on the evolution strategy of the current population, the topology population, capacity population and flow population are co-evolved in a decoupling manner to obtain candidate route network schemes. The evaluation and repair module is used to perform multi-objective evaluation of candidate route network schemes by calculating fitness, and to perform local flow repair on candidate route network schemes that do not meet physical and air traffic control safety constraints, so as to obtain the repaired route network scheme. The update execution module is used to perform non-dominated screening and environmental feature extraction on the repaired route network scheme, obtain non-dominated screening results and environmental features, and update the external archives and historical memory bank based on the repaired route network scheme, non-dominated screening results and environmental features. The results output module is used to stop population evolution and output the non-dominated solution set in the external archive when the preset termination condition is reached, as the result of low-altitude dynamic airway network collaborative optimization for the current period.

[0021] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the above-described demand-aware driven low-altitude dynamic airway network collaborative optimization method.

[0022] The present invention also provides a computer storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the above-described demand-aware driven low-altitude dynamic airway network collaborative optimization method.

[0023] The beneficial effects of this invention are as follows: It achieves efficient collaborative optimization of airway networks under changing demand and environment, improves search efficiency and solution feasibility, and ensures low-altitude operational safety. By treating airway topology, segment capacity configuration, and flow allocation as unified optimization objects, and employing a decoupled collaborative population evolution strategy for searching, it improves optimization efficiency and the overall performance of the solution. By distinguishing between stable and drastic change scenarios through an environmental intensity detection function and dynamically adjusting the population evolution strategy, it integrates prior knowledge injection with a historical memory bank when demand surges or capacity decreases, achieving rapid response and rolling optimization, thus solving the problems of slow response to environmental changes and the inability to reuse historical experience in existing methods. For candidate solutions that violate segment capacity, node flow conservation, or sector capacity constraints, it employs linear impedance mapping and a minimum-cost maximum-flow algorithm for local repair, while maintaining an effective combination of topology and capacity, reducing the elimination rate of infeasible solutions and improving computational efficiency and overall feasibility. Non-dominated screening combined with congestion distance maintenance of external archives and dynamic updates to the historical memory bank based on environmental characteristics enable the accumulation and reuse of high-quality Pareto solutions, providing prior guidance for subsequent rolling periods. By constructing a three-objective fitness vector, we can accurately evaluate candidate airway network schemes, ensure that the output solution has good trade-offs in multiple objectives, and enhance the constraint satisfaction capability by combining local flow repair, so as to achieve safe and efficient operation of the low-altitude airway network. Attached Figure Description

[0024] Figure 1 This is a diagram illustrating the method steps of the present invention.

[0025] Figure 2 This is a system module diagram of the present invention.

[0026] Figure 3 This is a schematic diagram illustrating the co-evolution mechanism of the topological population, capacity population, and flow population of the present invention.

[0027] Figure 4 This is a topology diagram of the low-altitude emergency flow restriction airway network in the urban central business district of Embodiment 3 of the present invention.

[0028] Figure 5 This is a topology diagram of a large-scale low-altitude airspace route network for a cross-river urban agglomeration according to Embodiment 3 of the present invention.

[0029] Figure 6 This is a diagram of the internal structure of a computer device according to Embodiment 4 of the present invention. Detailed Implementation

[0030] To clearly illustrate the technical features of this solution, the following detailed implementation method will be used to explain the solution.

[0031] Example 1: See Figure 1 , 3As shown, this embodiment is a demand-aware driven collaborative optimization method for low-altitude dynamic airway networks, including the following steps: S1. Discretize the low-altitude airspace into a three-dimensional directed network graph, and construct a multi-objective optimization model for the low-altitude airway network by defining system decision variables, constructing a dynamic multi-objective function, and setting physical and air traffic control safety constraints. It should be noted that this step aims to transform the physical low-altitude airspace into a computable mathematical model, providing a quantitative expression of the objective function and constraints for subsequent population co-evolution.

[0032] Furthermore, the target low-altitude airspace is discretized into a three-dimensional directed network graph. .in, It represents the set of spatial nodes in the airspace, including take-off and landing points, waypoints, and altitude transition points; This represents a set of directed flight segments connecting nodes, where each flight segment has a directional attribute and corresponds to an altitude layer. It should be noted that the above discretization process requires determining the node density and flight segment connectivity based on the actual airspace structure and aircraft performance.

[0033] Furthermore, in the rolling time domain Within, for each decision-making period Define the following system decision variables: : indicates the first Time Segment The start / stop status is indicated by "1" for the open segment and "0" for the closed segment. : indicates the first Time Segment Physical capacity, expressed in flights per unit time; : indicates the first Time period requirements In the flight segment Traffic allocation is measured in flights. : indicates the first Actual demand for admission during the time period The traffic, of which For the first Total original traffic demand for a given time period.

[0034] It should be noted that the above decision variables cover the three dimensions of the air route network: structure (start-stop), resources (capacity), and operation (flow), forming the variable space of the mixed integer nonlinear programming model.

[0035] It should also be noted that when constructing a dynamic multi-objective function, traffic efficiency, network stability, and operational security must be considered simultaneously. This invention sets the following three minimization objectives: The primary goal is to minimize the total amount of demand rejection, expressed as:

[0036] in, This is the set of all OD (Original Design Inventory) demand pairs. This objective is achieved through... Implement adaptive flow control to avoid network overload.

[0037] The second minimization objective is to minimize the overall cost of dynamic network reconfiguration, which is expressed as:

[0038] in, These are the preset reconstruction cost weighting coefficients. It should be noted that the first term penalizes abrupt changes in the start / stop state of a flight segment, the second penalizes significant adjustments to capacity configuration, and the third is the basic cost of keeping the flight segment open. Introducing an initial state aims to prevent unnecessary oscillations in the optimization results at time boundaries.

[0039] The third minimization objective is to minimize the overall network congestion delay, using an improved BPR function with a very small constant.

[0040] in, For the segment Free circulation time; Congestion sensitivity coefficient; This is a very small positive number used for calculating congestion delay, preventing the denominator from being zero. This function non-linearly maps the ratio of flow to capacity into additional delay, reflecting the actual physical laws governing congestion.

[0041] Furthermore, physical and air traffic control safety constraints are set to ensure the engineering feasibility of the generated airway network scheme: Constraint 1: Resource Capacity Constraint. The traffic allocated to any flight segment must not exceed the dynamic maximum capacity configured for that segment, and non-zero capacity can only be configured when the segment is open.

[0042] in, For the first Time Segment The dynamic maximum capacity.

[0043] Constraint 2: Elastic Node Flow Conservation Constraint. (Introduced) As a relaxation term, it allows the system to proactively throttle traffic when demand exceeds capacity:

[0044] in, and They represent respectively with A collection of flight segments with a starting point and an ending point.

[0045] Constraint 3: Sector Capacity Constraint. For each intersection area (sector) in the low-altitude three-dimensional airspace. The total concurrent traffic across all flight segments within the system is limited to not exceeding the air traffic control safety threshold. :

[0046] in, To belong to the sector The collection of flight segments, This is the set of all sectors.

[0047] It should be noted that, among the above constraints, and All parameters are time-varying and can be dynamically updated by external inputs such as weather forecasts and air traffic control instructions. By incorporating these time-varying parameters into the model, this invention achieves proactive perception and mathematical representation of the dynamic environment of the low-altitude airspace.

[0048] Furthermore, the low-altitude airway network multi-objective optimization model output in this step includes three objective functions and three sets of constraints, which serve as the basis for calculating population evolution fitness in the subsequent step S2. This model is constructed in a rolling time-domain manner, ensuring that the optimization for each time period is based on the latest environmental data and the baseline state of the previous time period.

[0049] S2. Based on the multi-objective optimization model of low-altitude airway network, define and initialize the topology population, capacity population, and flow population. At the same time, construct a history memory bank and an external archive. The history memory bank is used to store the environmental characteristics in the population evolution and their corresponding optimization solution set, while the external archive is used to save the repaired airway network scheme and non-dominated screening results. It should be noted that this step aims to establish the initial state of the three groups' cooperative evolution based on the low-altitude airway network multi-objective optimization model constructed in step S1, and to provide a data carrier for subsequent environmental detection, memory reuse and local repair.

[0050] Furthermore, for the decision variable types defined in step S1, three collaborative subpopulations with different encoding structures are defined and initialized respectively: (a) Topological population Using binary encoding, each individual in the population corresponds to a set of... The sequence of values. It should be noted that the binary encoding matches the 0 / 1 discrete nature of the topological state variables, and the length of each individual is equal to the total number of segments in the three-dimensional directed network graph. , of which A value of "1" indicates that the flight segment is enabled, and "0" indicates that it is disabled. During initialization, the network is set to a preset initial baseline state. Based on this, some bits are randomly flipped according to a set probability to generate an initial topological population.

[0051] (ii) Capacity population Using real number encoding, each individual in the population corresponds to a set of numbers. The sequence of values. It should be noted that real-number encoding can directly express the continuous value characteristics of the capacity, and the dimension of each individual is also [missing information]. Each gene value is a non-negative real number, representing the traffic capacity allocated to the corresponding flight segment in the current time period. During initialization, each gene value is... The interval is randomly generated according to a uniform distribution.

[0052] (III) Flow population The population uses a real-number matrix encoding method, where each individual corresponds to a set of numbers. The value matrix. It should be noted that the traffic allocation variable involves the intersection of demand and flight segment dimensions; therefore, each individual is represented as a... A real number matrix. During initialization, a shortest path flow allocation or random feasible flow generation algorithm is used to ensure that the initial flow individuals satisfy the basic boundary of the elastic node flow conservation constraint in step S1.

[0053] It should also be noted that the population sizes of all three populations were set to preset values. Furthermore, the three populations maintain the same generation count during evolution to achieve coordinated progress.

[0054] Furthermore, construct a historical memory bank. and external archives : Historical Memory Bank This is used to store historical environmental features and their corresponding optimized solution sets across time periods. Specifically, each entry in the memory is represented as... ,in For the first Environmental feature vectors of a historical scene This represents the non-dominated solution set for the corresponding scenario. It should be noted that the environmental feature vector is derived from the traffic demand vector for the current time period. Vector of the current time period's segment physical capacity limit The data is constructed by splicing and then normalized. Initially, the historical memory bank is empty, and the optimization results for the first time period are stored in the memory bank after completion.

[0055] It should be noted that the upper limit vector of the physical capacity of the flight segment in the current time period is... For the current time period, the capacity population is The maximum value randomly generated within the interval according to a uniform distribution.

[0056] External Archives This archive is used to store the non-dominated solution set generated during the evolution of the current time period and serves as a baseline reference for the next time period. The archive is updated using a Pareto non-dominated sorting and crowding distance maintenance strategy, and its capacity is preset to a certain limit. It should be noted that external archives serve both as the output carrier of the optimization results for the current period and as heuristic guidance information for population initialization in the next period when the environment is in a state of stable change.

[0057] Furthermore, after completing population initialization and data structure construction, this step outputs the initialized topological population. Capacity population Traffic population Historical Memory Bank and external archives This serves as the input for the subsequent step S3 co-evolution.

[0058] S3. Determine the intensity of environmental change based on the preset environmental time-varying detection function, and determine the evolution strategy of the current population based on the intensity of environmental change; according to the evolution strategy of the current population, use a decoupling method to perform co-evolution of the topology population, capacity population and flow population to obtain candidate route network schemes. See Figure 3 As shown, this step aims to dynamically select a population evolution strategy based on the degree of environmental change between the current time period and the previous time period, and to achieve decoupling and collaborative optimization of topological population, capacity population and flow population during the evolution process, thereby generating candidate route network schemes that meet multiple objectives.

[0059] Furthermore, in any current time period The formula for calculating the intensity of environmental change is:

[0060] in: and These represent the traffic demand vectors for the current time period and the previous time period, respectively. and These represent the upper limit vectors of the physical capacity of the flight segment in the current time period and the previous time period, respectively; These are the preset environmental detection weighting coefficients; The value should be a very small positive number to prevent the denominator from being zero. This represents the L2 norm.

[0061] Furthermore, the intensity of environmental change is compared with a preset threshold. Comparison:

[0062] In a stable evolutionary model, the previous external archive The non-dominated solution set is injected into the current population to guide the evolution of topology, capacity, and flow populations; in drastic change modes, it is retrieved from the historical memory bank. Extracting features from the current environment Highest similarity Inject the non-dominated solution set corresponding to each historical scenario into the population and reset it.

[0063] Among them, current environmental characteristics Represented as:

[0064] Each record in the historical memory bank is , This represents the historical non-dominated solution set.

[0065] Furthermore, decoupled evolution is performed on the topology population, capacity population, and flow population respectively: Specifically, the topology population evolves within the binary-coded segment start / stop state subspace; the capacity population evolves within the real-number-coded capacity subspace; and the flow population evolves within the real-number-matrix-coded flow allocation subspace. During the co-evolution process, the topology and capacity populations preferentially pair up based on compatibility to generate topology-capacity combinations. This is then passed to the traffic population to generate feasible traffic solutions. The complete solution is represented as:

[0066] It should be noted that compatibility measures whether capacity individuals are configured with effective capacity for open segments in the topology, and whether capacity individuals are not configured or only configured with near-zero capacity for closed segments. It also measures whether the capacity configuration matches the topology's start / stop status, thus avoiding incorrect combinations of individuals with "closed segments configured with large amounts of capacity" or "open segments with insufficient capacity," thereby reducing the probability of infeasible solutions arising in subsequent traffic allocation stages. In this embodiment, compatibility is obtained by calculating the Euclidean distance and structure mask matching degree between topology individuals and capacity individuals.

[0067] It should also be noted that in the initial stages of evolution... At that time, since the historical memory bank is empty and the external archive does not exist, the three groups are directly initialized using the stable evolution mode. After the evolution is completed, the generated candidate solutions are stored in the external archive and the historical memory bank for use in subsequent periods.

[0068] S4. By calculating the fitness, perform multi-objective evaluation on the candidate airway network schemes, and perform local flow repair on the candidate airway network schemes that do not meet the physical and air traffic control safety constraints to obtain the repaired airway network schemes. It should be noted that this step aims to refine the candidate route network schemes (i.e., the complete solution) generated in step S3. The system performs multi-objective evaluation on the set, filters out non-dominated solutions, and performs local repair operations on infeasible schemes that violate physical and air traffic control safety constraints, thereby obtaining feasible route network schemes that satisfy the constraints.

[0069] Furthermore, for each candidate complete solution output in step S3... This step first calculates the fitness values ​​of the three objective functions constructed in step S1. The specific calculation method is as follows: (a) Calculation of total demand rejection: Based on the expression for objective one in step S1, the current solution's first... Actual demand for admission during the time period Traffic With the Total original traffic demand during the time period Substitute the values ​​and calculate the total number of demand rejections for all OD pairs:

[0070] It should be noted that, among them Depend on The net flow from the source to the sink is obtained by inversely solving the elastic node flow conservation constraint.

[0071] (II) Calculation of Network Dynamic Reconfiguration Cost: Based on the expression for objective two in step S1, using the segment start / stop status and capacity configuration in the current solution, and the baseline status stored in the external archives of the previous time period, calculate the comprehensive cost of network dynamic reconfiguration:

[0072] in, This is the preset reconstruction cost weighting coefficient.

[0073] (III) Calculation of overall network congestion delay: Based on the expression for objective three in step S1, using the current solution... Capacity configuration variables Calculate the sum of congestion delays for all demands across all flight segments:

[0074] in, For the segment Free circulation time, For congestion sensitivity coefficient, It is a very small positive number.

[0075] It should also be noted that the above three objective function values ​​constitute the three-dimensional fitness vector of each candidate solution, which is used for subsequent non-dominated ranking and crowding distance calculation.

[0076] Furthermore, after calculating the fitness of all candidate solutions, this step performs non-dominated sorting. Specifically, a classic non-dominated sorting algorithm (such as fast non-dominated sorting) is used to divide all candidate solutions into multiple frontier levels according to Pareto dominance. It should be noted that the results of non-dominated sorting provide a selection criterion for subsequent updates to the external archives, and the non-dominated solution set located at the first front is considered the optimal compromise for the current generation.

[0077] Furthermore, for candidate solutions that do not meet the physical and air traffic control safety constraints, this step calls the local flow repair operator for correction. It should be noted that the constraints are the physical and air traffic control safety constraints set in step S1, specifically including resource capacity constraints, elastic node flow conservation constraints, and sector capacity constraints. The repair operation is as follows: (a) Constraint violation detection: Check each candidate complete solution one by one. Does it meet the following conditions: Resource capacity constraints: ; Elastic node flow conservation constraints: ; Sector capacity constraint: For each sector , .

[0078] If any condition is violated, the solution is marked as infeasible and the repair process begins.

[0079] (ii) Local Repair: While maintaining the topology and capacity configuration of the candidate route network scheme, adjustments are made to local flows that violate constraints in the flow allocation scheme. It should be noted that the local flow repair in this embodiment mainly operates on the flow allocation dimension corresponding to the flow population. Its purpose is not to regenerate the complete topology and capacity configuration, but to pull back or approximate the feasible region of flow allocations that violate capacity limits, flow conservation relationships, or sector capacity limits under a given topology-capacity combination.

[0080] Furthermore, during local flow repair, the first step is to identify the segments, nodes, or sectors that violate constraints. For segments that violate resource capacity constraints... Cut or transfer more than The flow; for nodes that violate the elastic node flow conservation constraint. Adjust relevant requirements Inflow and outflow relationships on adjacent flight segments; for three-dimensional junctions or airspace sectors that violate sector capacity constraints, reduce or redistribute concurrent traffic passing through such junctions or airspace sectors.

[0081] Furthermore, to improve the efficiency of local flow repair, the nonlinear congestion-delay relationship in the candidate route network scheme is reduced in dimension and mapped to a linear impedance network. Then, the minimum-cost maximum-flow algorithm is called on this linear impedance network for flow redistribution. The linear impedance network uses currently active routes as traversable edges, the available remaining capacity of the route as the capacity constraint, the route travel time or congestion impedance as cost information, and demand... The starting point and the end point As both the sender and receiver of traffic, the minimum cost maximum flow algorithm is used to redistribute traffic that violates constraints, while satisfying constraints on segment capacity, node flow conservation, and sector capacity.

[0082] It should be noted that using the minimum cost maximum flow algorithm for local flow repair can quickly correct unreasonable flow allocation schemes without significantly changing the topology and capacity configuration of the candidate airway network. At the same time, by approximating the nonlinear congestion delay as a linear impedance, the computational complexity of the repair process can be reduced, making the repair operation applicable to multi-generational population evolution processes in the rolling time domain.

[0083] (III) Post-repair verification: The traffic allocation matrix obtained after repair Resubstitute the constraints and verify. If all constraints are satisfied, then use... Replace the original traffic instance to obtain the repaired complete solution. If the conditions are still not met (for example, due to the topology itself causing certain requirements to be unconnected), then the solution is marked as unrepairable and eliminated in this generation of evolution (i.e. it will not participate in subsequent non-dominated sorting and file updates).

[0084] It should also be noted that, to ensure the overall feasibility level of the population, this step, after completing the local repair, involves backfeeding the topological state and capacity configuration characteristics from the verified feasible complete solution into the topological and capacity populations. Specifically, this involves backfeeding the complete solution that satisfies all constraints after repair into the topological and capacity populations. Extract the corresponding flight segment start / stop status from the data. and capacity configuration The same number of individuals with the worst fitness in both the topological and capacity populations are replaced. It should be noted that although the local repair operator only adjusts the flow allocation, the repaired feasible solution proves that the current topology-capacity combination can support a feasible flow scheme. Therefore, backfeeding it can propagate the verified structural information to the other two populations, thereby raising the search lower bound of the entire co-evolutionary system.

[0085] Furthermore, the repaired route network scheme output in this step... This approach retains the joint optimization results of topology, capacity, and flow obtained from the three group co-evolutions in step S3, and improves the degree to which the scheme satisfies capacity constraints, flow conservation constraints, and air traffic control safety constraints through local flow repair. This provides a usable scheme basis for non-dominated screening, environmental feature extraction, and external archive and historical memory bank updates in step S5.

[0086] S5. Perform non-dominated screening and environmental feature extraction on the repaired route network scheme to obtain non-dominated screening results and environmental features. Update the external archives and historical memory bank based on the repaired route network scheme, non-dominated screening results and environmental features. It should be noted that this step aims to perform non-dominated screening on the repaired route network scheme set output by step S4, extract the environmental features of the current time period, and synchronously update the external archives and historical memory bank based on the screening results and feature information, so as to provide benchmark reference and prior knowledge for the rolling optimization of the next time period.

[0087] Furthermore, for all complete solutions that have been partially repaired and verified to be feasible in step S4... This step first performs a non-dominated screening operation. Specifically, it compares the feasible solutions generated in the current generation with external files. The existing historical non-dominated solutions are merged to form a temporary merge set; then, the fast non-dominated sorting algorithm is used to perform Pareto hierarchies on the merge set to obtain multiple frontier levels. Simultaneously, crowding distances are calculated for individuals within the same frontier level to assess the distribution density of solutions. It should be noted that the goal of non-dominated screening is to preserve solutions that are non-dominated and uniformly distributed across the three objective functions, thereby maintaining the convergence and diversity of the archives.

[0088] Furthermore, based on the preset maximum file capacity... (Can be set by the implementer), individuals are selected sequentially from the sorted frontier hierarchy: priority is given to selecting All individuals; if the number is insufficient Then select in sequence Individuals within the same frontier level; when partial selection is required, individuals with greater crowding distance are prioritized to maintain archival diversity. After the above screening, updated external archives are obtained. This database stores the non-dominated feasible solutions selected during the current time period. It should be noted that external files serve both as the final candidate planning scheme set for this time period and as heuristic guidance information during the smooth evolution of the next time period.

[0089] Furthermore, this step extracts the current environmental features. Used to update the historical memory bank The current environmental characteristics are determined by the traffic demand vector for the current time period. Vector of segment capacity limit It is obtained by splicing and normalizing, that is

[0090] It should be noted that normalization can be performed using max-min normalization or Z-score normalization to ensure the comparability of features with different dimensions, thereby facilitating the similarity calculation of the historical memory bank.

[0091] Furthermore, external archives for the current time period The optimal solution set corresponding to this environmental characteristic , construct memory entries And store it in the historical memory bank It should be noted that the capacity of the historical memory bank can be preset to [value missing]. (e.g., 20), when the number of entries exceeds At certain times, an elimination mechanism can be triggered, specifically by assigning an applicability weight to each memory entry. The applicability weight decays over time according to a preset decay coefficient. The applicability weight is reduced and adjusted based on the non-dominant performance of the corresponding solution set in subsequent population evolution after the memory entry is extracted and injected. If the solution set corresponding to a memory entry fails to enter the first layer of the current non-dominant frontier in several consecutive generations of evolution, it is further multiplied by a penalty factor. Reduce the applicability weight of the memory entry; when the applicability weight falls below a preset elimination threshold. When that happens, the memory entry will be deleted from the history memory bank.

[0092] It should also be noted that when updating the historical memory database, cosine similarity or Euclidean distance is used to measure the characteristics of the current environment. The similarity to the environmental features of existing memory entries is considered if the similarity exceeds a preset similarity threshold. Instead of adding new entries, the solution set in existing memory entries is replaced or merged to maintain the timeliness and reliability of the memory.

[0093] It should be noted that, through the above processing, step S5 completes the final screening of candidate solutions for the current time period, file maintenance and environmental feature information extraction, realizing the continuity and adaptive capability of low-altitude airway network collaborative optimization in the rolling time domain.

[0094] S6. When the preset termination condition is met, stop the population evolution and output the non-dominated solution set in the external archive as the result of low-altitude dynamic airway network collaborative optimization for the current period.

[0095] It should be noted that this step aims to determine whether the population evolution in the current time period meets the termination condition, and when the condition is met, output the non-dominated solution set in the external archive as the low-altitude dynamic airway network collaborative optimization result for that time period, and at the same time decide whether to continue rolling to the next time period or end the entire planning time domain.

[0096] Furthermore, this step checks the evolutionary termination condition for the current time period after each generation of evolution is completed. The termination condition includes either of the following two scenarios: (a) Termination of maximum iteration count: When the evolutionary generation of the current time period reaches the preset maximum iteration count. When the condition is met, the algorithm is considered to have reached its termination point. It should be noted that the maximum number of iterations should be pre-determined based on network size, demand density, and computational resources to ensure sufficient convergence.

[0097] (ii) Termination of convergence criterion: When continuous In the process of generational evolution, external archives The rate of change of the hypervolume index of the non-dominated solution set in the target space is lower than the preset change threshold. At this point, the algorithm is considered to have converged, and the evolution is terminated prematurely. It should be noted that the hypervolume index measures the coverage volume of the non-dominated solution set relative to the reference point. A rate of change approaching zero indicates that the Pareto front has basically stabilized, and further evolution yields minimal benefit.

[0098] Furthermore, when any of the above termination conditions are met, this step terminates the population evolution for the current time period and updates the external archives. The non-dominated solution set stored in the middle is used as the current time period Output of the collaborative optimization results of the low-altitude dynamic airway network.

[0099] Example 2: See Figure 2 As shown, this embodiment is a demand-aware driven low-altitude dynamic airway network cooperative optimization system, including: The model building module is used to discretize the low-altitude airspace into a three-dimensional directed network graph, and to build a multi-objective optimization model of the low-altitude airway network by defining system decision variables, constructing dynamic multi-objective functions, and setting physical and air traffic control safety constraints. The initialization module is used to define and initialize the topology population, capacity population, and flow population based on the multi-objective optimization model of the low-altitude airway network. At the same time, it builds the history memory bank and external archives. The population evolution module is used to determine the intensity of environmental change based on a preset environmental time-varying detection function, and to determine the evolution strategy of the current population based on the intensity of environmental change; based on the evolution strategy of the current population, the topology population, capacity population and flow population are co-evolved in a decoupling manner to obtain candidate route network schemes. The evaluation and repair module is used to perform multi-objective evaluation of candidate route network schemes by calculating fitness, and to perform local flow repair on candidate route network schemes that do not meet physical and air traffic control safety constraints, so as to obtain the repaired route network scheme. The update execution module is used to perform non-dominated screening and environmental feature extraction on the repaired route network scheme, obtain non-dominated screening results and environmental features, and update the external archives and historical memory bank based on the repaired route network scheme, non-dominated screening results and environmental features. The results output module is used to stop population evolution and output the non-dominated solution set in the external archive when the preset termination condition is reached, as the result of low-altitude dynamic airway network collaborative optimization for the current period.

[0100] Example 3: To verify the dynamic optimization capability and computational efficiency of this invention under different scales of low-altitude airspace and sudden environmental conditions, this embodiment designed two simulation scenarios: one is a sudden low-altitude airspace rationing scenario in a city's central business district (CBD), and the other is a low-altitude airspace scenario in a large-scale urban agglomeration spanning rivers / bays. The static NSGA-II, dynamic D-NSGA-II, PPS-RM with prediction strategies, and the method of this invention were tested and compared in the two simulation scenarios.

[0101] In a sudden low-altitude traffic restriction scenario in a city's central business district (CBD), the response speed and optimization capability of this invention under sudden environmental changes can be verified. A typical city CBD low-altitude logistics and commuting hybrid simulation scenario can be set up (e.g.,...). Figure 4 (As shown). This airspace contains 25 spatial nodes, 80 potential three-dimensional flight segments, divided into 3 altitude layers, involving 18 core OD demand pairs. The time step is set to 5 minutes, and a total of 12 time periods (1 hour) are evaluated.

[0102] Dynamic event injection settings: during time period At that time, a simulated sudden severe convective weather event (such as gusts in an urban canyon) caused the maximum physical capacity of the four main air routes passing through the core business district to drop by 70% instantaneously, and some nodes were closed to flights.

[0103] The method of this invention was compared with the classical static algorithm (NSGA-II, reset over time), the classical dynamic evolution algorithm (D-NSGA-II), and the dynamic algorithm with prediction strategy (PPS-RM). Under a set of simulation settings, the comparison results are shown in Table 1 (time periods are considered). to (Average performance) Table 1 Comparison of optimization results of various algorithms in CBD sudden flow throttling simulation scenario

[0104] Simulation results show that, when faced with a sudden drop in capacity, the method of this invention exhibits a higher demand service ratio and lower reconfiguration cost in this embodiment. Its demand service ratio is expected to remain at [value missing]. The average vehicle delay is expected to be reduced to 1.9 minutes. The significant reduction in reconstruction cost (18.5) and solution time (42 seconds) verifies the effectiveness of the historical memory reuse mechanism and local flow repair operator of this invention from the simulation level, indicating that the method has the theoretical feasibility and application potential to achieve minute-level near real-time dynamic response.

[0105] In the airspace scenario of large-scale urban agglomerations spanning rivers / bays, the hybrid coding decoupling capability and scalability of the proposed method were verified, and a larger-scale low-altitude simulation scenario of a cross-river urban agglomeration (such as...) was constructed. Figure 5 (As shown). This network contains 80 nodes, 240 potential flight segments, 5 altitude levels, and involves 45 high-volume OD pairs, resulting in a sharp increase in the dimensionality of decision variables. Under this set of simulation settings, the comparison results shown in Table 2 are obtained: Table 2 Comparison of optimization results of various algorithms in a large-scale cross-river urban agglomeration simulation scenario.

[0106] Simulation results show that with the significant increase in airspace complexity, traditional centralized coding algorithms face the "curse of dimensionality," leading to a dramatic increase in solution time. In contrast, the method of this invention is expected to maintain a high service rate of approximately 92.4% and a high HV value of 0.815 in this large-scale simulation scenario. This comparative trend demonstrates the strong scalability of the three swarm cooperative evolution architectures proposed in this invention: through dimensionality reduction and decoupling of macro-topology, micro-capacity, and specific traffic, coupled with "compatibility-first pairing," the algorithm can quickly locate high-quality feasible solution regions in a vast search space, and the theoretical solution time meets the time window requirements for large-scale airway network rolling operations.

[0107] Example 4: This embodiment provides a computer device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0108] This computer device can be a server, and its internal structure diagram can be as follows: Figure 6 As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database stores server data. The network interface communicates with external terminals via a network connection. When executed by the processor, the computer program implements a demand-aware, dynamic low-altitude airway network cooperative optimization method.

[0109] Those skilled in the art will understand that Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0110] Example 5: This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0111] If the functions implemented by the method are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art or the current technical solution, can be embodied in the form of a software product. This current 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 this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0112] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0113] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0114] The technical features of this invention not described can be implemented by or using existing technology, and will not be repeated here. Of course, the above description is not a limitation of this invention, and this invention is not limited to the examples above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of this invention should also be within the protection scope of this invention.

Claims

1. A demand-aware driven low-altitude dynamic air route network collaborative optimization method, characterized in that, Includes the following steps: S1. Discretize the low-altitude airspace into a three-dimensional directed network graph, and construct a multi-objective optimization model for the low-altitude airway network by defining system decision variables, constructing a dynamic multi-objective function, and setting physical and air traffic control safety constraints. S2. Based on the multi-objective optimization model of low-altitude airway network, define and initialize the topology population, capacity population, and flow population. At the same time, construct a historical memory bank and an external archive. The historical memory bank is used to store the environmental characteristics in the population evolution and their corresponding optimization solution set. The external archive is used to save the repaired airway network scheme and non-dominated screening results. S3. Determine the intensity of environmental change based on the preset environmental time-varying detection function, and determine the evolution strategy of the current population based on the intensity of environmental change; according to the evolution strategy of the current population, use a decoupling method to perform co-evolution of the topology population, capacity population and flow population to obtain candidate route network schemes. S4. By calculating the fitness, perform multi-objective evaluation on the candidate airway network schemes, and perform local flow repair on the candidate airway network schemes that do not meet the physical and air traffic control safety constraints to obtain the repaired airway network schemes. S5. Perform non-dominated screening and environmental feature extraction on the repaired route network scheme to obtain non-dominated screening results and environmental features. Update the external archives and historical memory bank based on the repaired route network scheme, non-dominated screening results and environmental features. S6. When the preset termination condition is met, stop the population evolution and output the non-dominated solution set in the external archive as the result of low-altitude dynamic airway network collaborative optimization for the current period. The preset environmental time-varying detection function formula in step S3 is expressed as follows: wherein: and and and and is a preset environmental detection weight coefficient; is a very small positive number to prevent the denominator from being zero; denotes a two-norm; The evolutionary strategy for determining the current population based on the intensity of environmental change includes: When the intensity of environmental change is less than or equal to the preset threshold, the environment is determined to be stable, and the non-dominated solution set in the external file of the previous period is injected into the current topology population, capacity population and flow population according to the preset ratio. When the intensity of environmental change exceeds a preset threshold, it is determined to be a drastic environmental change. The non-dominated solution set corresponding to the top few historical scenarios with the highest similarity to the current environmental features is extracted from the historical memory bank and injected into the current topology population, capacity population and flow population according to the preset injection ratio to complete the population reset. The cooperative evolution is performed in a decoupled manner, specifically including: Genetic mutation operations are performed independently on the topological population, capacity population, and flow population respectively. For any topological individual in the topological population, the compatibility with the capacity individual in the capacity population is calculated, and the capacity individual with the highest compatibility with the topological individual is selected to form a topological capacity combination. The topology capacity combination is paired with the traffic population to determine a traffic individual, and candidate route network schemes are determined based on the topology capacity combination and the traffic individual.

2. The demand-aware driven low-altitude dynamic flight corridor network collaborative optimization method according to claim 1, characterized in that, The system decision variables in step S1 include topology state variables, capacity configuration variables, flow allocation variables, and admission flow variables; the topology state variables are used to represent the segment start-stop status; the capacity configuration variables are used to represent the segment physical capacity. The flow allocation variable is used to represent the flow allocation of demand on the flight segment; the admission flow variable is used to represent the flow of actual admission demand. In step S2, the topology population uses binary encoding, the capacity population uses real number encoding, and the flow population uses real number matrix encoding.

3. The demand-aware driven low-altitude dynamic airway network cooperative optimization method according to claim 2, characterized in that, The dynamic multi-objective function in step S1 includes a first minimization objective, a second minimization objective, and a third minimization objective; The first minimization objective is to minimize the total amount of demand rejection, expressed by the following formula: in, For the current decision-making period, For the requirements of indexing, For the set of all OD demand pairs, For the first Time period requirements The original traffic demand, For the first Actual demand for admission during the time period Traffic; The second minimization objective is to minimize the overall cost of dynamic network reconstruction, expressed by the following formula: in, The preset reconstruction cost weighting coefficient, For flight segment index, For segment collection, For the first Time Segment Start-stop status, For the first Time Segment Start-stop status, For the first Time Segment Physical capacity For the first Time Segment Physical capacity; The third minimization objective is to minimize the overall network congestion delay, expressed by the following formula: in, For the first Time period requirements In the flight segment Traffic allocation on For the segment Free circulation time, For congestion sensitivity coefficient, This is a very small positive number used for congestion delay calculation; The physical and air traffic control safety constraints include resource capacity constraints, elastic node flow conservation constraints, and sector capacity constraints. The resource capacity constraint formula is expressed as follows: in, For the first Time Segment The dynamic maximum capacity; The formula for the elastic nodal flow conservation constraint is expressed as follows: in, and They represent respectively with A collection of flight segments with a starting point and an ending point; The sector capacity constraint formula is expressed as follows: in, To belong to the sector The collection of flight segments, For the set of all sectors, For the first Time Sector The air traffic control safety threshold.

4. The demand-aware driven low-altitude dynamic airway network cooperative optimization method according to claim 3, characterized in that, The execution of local flow repair in step S4 includes: For candidate airway network schemes that do not meet the physical and air traffic control safety constraints, the topology and capacity individuals remain unchanged, and they are reduced in dimension and mapped to a linear impedance network. The linear impedance network uses the currently active air segments as traversable edges, the available remaining capacity of the air segments as the capacity limit, and the free passage time of the air segments as the cost. The starting point and the ending point correspond to the source and sink points of each demand. The minimum cost maximum flow algorithm is called to redistribute the flow in the linear impedance network to obtain the repaired flow individuals. The repaired flow individuals are then used to replace the flow individuals in the candidate route network scheme to obtain the repaired route network scheme. The repaired route network scheme is constrained and verified. If all constraints are met, it is retained; otherwise, it is marked as unrepairable and eliminated in the current generation of evolution.

5. The demand-aware driven low-altitude dynamic airway network cooperative optimization method according to claim 4, characterized in that, The environmental features in step S5 include the normalized current time period traffic demand vector and the normalized current time period segment capacity limit vector. When the historical memory bank reaches a preset storage limit, it is updated using an eviction mechanism; the eviction mechanism specifically includes: When each memory entry is stored in the historical memory bank, an applicability weight is assigned to the memory entry; the applicability weight is attenuated according to a preset attenuation coefficient. When a memory entry is extracted for population reset during the co-evolution process in step S3, if the non-dominated solution set of the memory entry has not entered the first layer of the non-dominated frontier in several consecutive generations of evolution, the applicability weight of the memory entry is reduced by a preset penalty factor. When the applicability weight is lower than the preset elimination threshold, the memory entry is deleted.

6. The demand-aware driven low-altitude dynamic airway network cooperative optimization method according to claim 5, characterized in that, The elimination mechanism also includes: During the historical memory bank update process, the similarity between the current environmental features and the environmental features of existing memory entries is calculated. If the similarity exceeds a preset similarity threshold, no new or deleted memory entries are added. Instead, the non-dominated solution set in the existing memory entries is replaced with the non-dominated solution set corresponding to the current environmental features.

7. A demand-aware driven low-altitude dynamic airway network collaborative optimization system, characterized in that, The steps for performing the demand-aware driven low-altitude dynamic airway network cooperative optimization method according to any one of claims 1 to 6 include: The model building module is used to discretize the low-altitude airspace into a three-dimensional directed network graph, and to build a multi-objective optimization model of the low-altitude airway network by defining system decision variables, constructing dynamic multi-objective functions, and setting physical and air traffic control safety constraints. The initialization module is used to define and initialize the topology population, capacity population, and flow population based on the multi-objective optimization model of the low-altitude airway network. At the same time, it builds the history memory bank and external archives. The population evolution module is used to determine the intensity of environmental change based on a preset environmental time-varying detection function, and to determine the evolution strategy of the current population based on the intensity of environmental change; based on the evolution strategy of the current population, the topology population, capacity population and flow population are co-evolved in a decoupling manner to obtain candidate route network schemes. The evaluation and repair module is used to perform multi-objective evaluation of candidate route network schemes by calculating fitness, and to perform local flow repair on candidate route network schemes that do not meet physical and air traffic control safety constraints, so as to obtain the repaired route network scheme. The update execution module is used to perform non-dominated screening and environmental feature extraction on the repaired route network scheme, obtain non-dominated screening results and environmental features, and update the external archives and historical memory bank based on the repaired route network scheme, non-dominated screening results and environmental features. The results output module is used to stop population evolution and output the non-dominated solution set in the external archive when the preset termination condition is reached, as the result of low-altitude dynamic airway network collaborative optimization for the current period.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the demand-aware driven low-altitude dynamic airway network collaborative optimization method as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the demand-aware driven low-altitude dynamic airway network collaborative optimization method as described in any one of claims 1 to 6.

Images