A dynamic low-altitude public air route network construction method based on two-party multi-objective ant colony negotiation

By employing a multi-objective ant colony negotiation method, the problem of network combination infeasibility caused by single route planning in urban low-altitude public airway networks is solved. This method enables rapid updates and ensures safety under changing environments, and is applicable to scenarios such as urban logistics distribution, vertical take-off and landing short-haul transportation, and emergency support.

CN123072237APending Publication Date: 2026-11-10NANJING UNIV OF INFORMATION SCI & TECH
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202611429741.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-15
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

In the construction of urban low-altitude public airway networks, existing technologies lack overall consideration for the planning of individual routes, making network combination infeasible and difficult to update quickly when the environment changes, thus failing to balance operational efficiency and public safety.

Method used

A dynamic low-altitude public airway network is constructed by adopting a method based on multi-objective ant colony negotiation between both parties. This method involves synchronous construction of the entire network of ants, feasibility of continuous three-dimensional hard constraints, non-dominated optimization of the benefit and safety parties, actual three-dimensional local consensus pheromone, consensus anchoring and reorganization, and event-triggered replanning.

Benefits of technology

It enables rapid updates to the air route network in response to environmental changes, ensuring network feasibility and security while balancing operational efficiency and public safety. The output air route network is suitable for scenarios such as urban logistics distribution, vertical take-off and landing short-haul transportation, and emergency support.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123072237A_ABST
    Figure CN123072237A_ABST
Patent Text Reader

Abstract

The application provides a dynamic low-altitude public air route network construction method based on two-party multi-target ant colony negotiation, comprising the following steps: step 1, establishing an environment snapshot and expressing a complete air route network; step 2, synchronously constructing a complete network and implementing hard constraint feasibility; step 3, forming two-party non-dominated archives and updating pheromones; step 4, carrying out intra-process negotiation based on three-dimensional common air segments; and step 5, determining a representative network and implementing event-triggered re-planning. The application takes the complete air route network as a basic object for one-time planning and evaluation, so that the construction clearance, no-fly zone, take-off and landing point protection and air route spacing between different ODs are uniformly checked under the same environment snapshot, the situation of repeatedly repairing due to combination conflicts after single-line generation can be reduced, and the method is suitable for scheme generation of fixed or periodic air route networks such as urban logistics distribution, vertical take-off short-distance transportation and emergency support.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of UAV route network planning, and in particular relates to a method for constructing a dynamic low-altitude public route network based on multi-target ant colony negotiation between two parties. Background Technology

[0002] Applications such as low-altitude logistics, urban air traffic, emergency rescue, and short-haul transportation are shifting from demonstration flights of single aircraft to networked operations involving multiple takeoffs and landings, multiple routes, and multiple missions. The "Interim Regulations on the Management of Unmanned Aerial Vehicle Flights" have already set forth requirements for safety first, service development, classified management, and collaborative supervision for unmanned aerial vehicle flight activities. The "Implementation Plan for Innovative Applications of General Aviation Equipment (2024-2030)" further proposes the construction of short-haul transportation and electric vertical takeoff and landing passenger networks, unmanned aerial vehicle delivery networks, and corresponding route networks and dynamic monitoring capabilities. Civil aviation authorities are promoting the integration of elements such as low-altitude route planning, aerial obstacle avoidance, airspace management, and traffic control, and have proposed the construction of a low-altitude three-dimensional digital grid geographic information system. Therefore, planning for actual operations cannot be limited to a single temporary path for an aircraft; it also requires the formation of a verifiable, publishable, and sustainably updated three-dimensional public route network.

[0003] Low-altitude path and route planning has developed into a relatively rich technical system. Search methods based on graphs, grids, or sampled spaces can generate feasible paths using explicit spatial topology and cost rules, characterized by clear processes, stable results, and ease of constraint checking. Trajectory optimization, rolling planning, and local repair methods facilitate continuous adjustments to existing tracks to adapt to changes in obstacles or airspace conditions. Swarm intelligence methods such as genetic algorithms, particle swarm optimization, and ant colony optimization possess strong global search capabilities, comprehensively considering nonlinear factors such as range, energy consumption, turning, and flight risk. Ant colony optimization, in particular, accumulates historical experience in selecting high-quality flight segments through pheromone accumulation, uses heuristic information to evaluate the next transition, and gradually constructs paths according to probabilistic rules, making it suitable for combinatorial planning problems consisting of multiple flight segments. Related published patents have also proposed different implementation approaches. For example, CN109948834B utilizes preset altitude slices and ant colony search to construct low-altitude public airways; CN110597276B employs three-dimensional discrete nodes and a two-level ant colony method to plan unmanned aerial vehicle (UAV) air safety corridors; and CN115657721A improves three-dimensional UAV trajectory planning from aspects such as path length, trajectory variation, and pheromone enhancement, providing a corresponding technical foundation for public airway construction, three-dimensional safety corridor search, and single-line trajectory optimization. In terms of multi-objective planning, mechanisms such as Pareto optimization, non-dominated sorting, and multi-swarm collaboration can handle the trade-offs between range, energy consumption, safety risks, and environmental impact, allowing the planning system to retain multiple candidate solutions with different performance focuses, without having to pre-merge all objectives into a single weighted index. The aforementioned methods each have their advantages in applications such as single-route planning, multi-aircraft collaboration, dynamic obstacle avoidance, and multi-objective decision-making. However, when the planning object is further expanded to an urban low-altitude public airway network composed of multiple take-off and landing points and multiple OD routes, it is not only necessary to ensure that each route meets the requirements of building clearance, no-fly zones, take-off and landing point protection, and aircraft maneuverability, but also to coordinate the space occupancy and safety separation of different routes in a unified network, while taking into account the concerns of operators regarding energy consumption and maneuverability efficiency, as well as the concerns of public management entities regarding the risk of injury or death to ground personnel and noise exposure. Therefore, how to achieve the joint construction of a complete airway network, the evaluation of sub-entity objectives, and effective collaboration in the search process based on existing path search and multi-objective optimization technologies remains a problem that needs further research in public airway network planning.

[0004] Meanwhile, the urban low-altitude operating environment is not always static; population exposure, road and vehicle hotspots, temporary activity areas, and no-fly zone status can all change over time. Existing networks may remain feasible in the new environment, or only some flight segments may be affected, or new airspace restrictions may necessitate replanning. If the entire network is rebuilt from scratch for each environmental change, a large number of unaffected flight segments will be repeatedly calculated; conversely, making only local geometric adjustments to old routes may alter existing trade-offs or introduce new inter-route conflicts. Summary of the Invention

[0005] Purpose of the Invention: This invention aims to address the core problem in the current construction of urban low-altitude public airway networks: the isolated generation of air routes and insufficient consultation between operational and safety stakeholders make it difficult to balance operational efficiency and public safety in the network planning results. Specifically, traditional methods often use individual origin-destination routes as the basic solution object, obtaining paths separately and then superimposing them into a network. This lacks a holistic consideration of shared airspace, building clearance, no-fly zones, vertical take-off and landing protection zones, and airway spacing during the unified construction and evaluation process. This easily leads to the problem that individual routes may be feasible, but the combined network cannot be directly operated. Existing multi-objective methods often combine indicators such as energy consumption, maneuverability, personnel risk, and noise through manual weighting, or only output a mixed frontier, which makes it difficult to reflect the optimization basis of the two subjects of operational efficiency and public safety, and also makes it difficult to explain the concessions made to form the final solution.

[0006] Meanwhile, even if the two populations can search separately, they may end up at two separate extreme frontiers due to the lack of a communication mechanism based on real spatial location: jointly selected local flight segments cannot be used to guide subsequent searches, and accidental intersections or identical action codes may be mistaken for consensus; when population, road exposure, and temporary no-fly zones change, the fixed route network cannot reflect the new operating conditions in a timely manner, and a complete recalculation will reduce update efficiency, making it difficult to form a planning closed loop for continuous operation.

[0007] To address the aforementioned systemic problems, this invention constructs a complete technical chain from environmental data access, network generation, constraint verification, negotiation between the two parties, determination of representative schemes, to runtime updates. This chain utilizes synchronous construction of complete network ants, continuous three-dimensional hard constraint feasibility, dual-population non-dominated optimization of benefit and safety parties, actual three-dimensional local consensus pheromone, consensus anchoring reorganization, frontier balance selection of both parties, and event-triggered cold and hot budget replanning. The generated results can serve as candidate networks for urban low-altitude airway planning platforms, allowing management entities to conduct review, simulation verification, version release, and subsequent dynamic maintenance.

[0008] This invention specifically provides a method for constructing a dynamic low-altitude public airway network based on multi-target ant colony negotiation between two parties, including the following steps:

[0009] Step 1: Create an environmental snapshot and represent the complete route network;

[0010] Step 2: Simultaneously construct the complete network and implement hard constraint feasibility.

[0011] Step 3: Create non-dominant files for both parties and update pheromones;

[0012] Step 4: Conduct in-process negotiations using the three-dimensional common route;

[0013] Step 5: Identify the representative network and implement event-triggered replanning.

[0014] Step 1 includes: acquiring the planning area boundary, building occupancy and height, population and road / vehicle exposure, important facilities, static and temporary no-fly zones, take-off and landing points, and protection zones under a unified coordinate reference, and recording the permitted flight altitude, aircraft speed and mass, maximum climb angle, maximum descent angle, airway envelope width, and safety separation; wherein, the take-off and landing points include the take-off and landing points of vertical take-off and landing manned aircraft, and the protection zone is a restricted airspace that unrelated UAVs are not allowed to intrude into; building and geographic data are formed from urban 3D geographic information or surveying data, population, road, and vehicle exposure are formed from population grids, road networks, and traffic status data, no-fly zones are formed from airspace management or operational management status, and take-off and landing points are formed from confirmed site data;

[0015] Data from the same sampling time and under the same spatial coordinate reference are combined into an environmental snapshot with time and version identifiers; based on the environmental snapshot, inputs for ground personnel casualty risk assessment, noise assessment, and spatial occupancy indexes for buildings, no-fly zones, and take-off and landing point protection zones are generated; the spatial occupancy indexes are used to query buildings and restricted airspace near the flight segment envelope without changing the judgment criteria for the corresponding constraints;

[0016] All complete network ants in the same planning round read the same environmental snapshot to ensure that candidate complete route networks are constructed and evaluated under the same environmental conditions; when subsequent monitoring data arrives, the environmental snapshot used to generate the currently published network or the most recent actual replanning is used as the benchmark for judging whether the environment has undergone substantial changes.

[0017] Step 1 further includes: assuming the complete route network contains K origin-destination routes, these origin-destination routes are called OD routes, where OD is an abbreviation for Origin-Destination and represents a pair of origins to destinations; the kth origin-destination route contains B kEach network ant has a decision variable that simultaneously stores the horizontal state and continuous altitude state of K origin-end routes. Both beneficial and safe ants belong to the complete network ant family. They use the same network coding and hard constraint verification, differing only in their dominant target, pheromone, heuristic information, and the file they enter. The horizontal state is accumulated segment by segment within the local coordinate system of each origin-end route's origin and destination, using five horizontal increments: shifting two units to the left, shifting one unit to the left, maintaining the current state, shifting one unit to the right, and shifting two units to the right. The pheromone address includes the origin-end route number, segment number, current actual horizontal-altitude state, and next actual horizontal-altitude state, to distinguish different actual segments corresponding to the same action number in different preceding states.

[0018] Highly employing continuously normalized variables Encode the data and decode it into the actual physical height according to equation (1):

[0019] (1),

[0020] in This represents the continuously normalized altitude variable of the i-th control point on the k-th OD route. ∈[0,1]; The actual altitude of the i-th control point on the k-th OD route. and These are the lower and upper limits of the permitted flight altitude, respectively.

[0021] Step 2 includes:

[0022] In the joint construction round of the s-th segment, all OD routes that have not yet reached their destination form a legal next transfer set for the s-th segment based on the same generation's initial frozen pheromone snapshot, the same environmental snapshot, and the actual state at the end of their respective s-1 segments. After each OD route completes its transfer selection in this round, it is submitted uniformly and then enters the joint construction round of the s+1 segment, thereby avoiding changes in the pheromone basis of the same generation by altering the order of OD route processing.

[0023] The legal next transfer set is directly screened from the planning area boundary, destination accessibility, static and temporary no-fly zones, non-associated take-off and landing point protection zones, and the envelope requirements of this OD route; among them, destination accessibility refers to the ability to reach the fixed destination within the remaining number of segments and the allowable lateral increment.

[0024] For each transfer within the set of legitimate next transfers, the benefit ant forms its own heuristic information based on the incremental evaluation of propulsion energy consumption and 3D maneuver costs, and reads the benefit ant's pheromone; the safety ant forms its own heuristic information based on the incremental evaluation of ground personnel casualty risk and noise exposure, and reads the safety ant's pheromone; the opponent's heuristic correction is formed by the normalized incremental cost of the other party's corresponding target group in the same set of legitimate next transfers, and participates with a bounded factor with a preset upper limit; the 3D consensus pheromone is formed by the actual 3D overlapping segments with the same origin-destination route and the same segment number in the first non-dominated frontier of the benefit archive and the first non-dominated frontier of the safety archive in the previous iteration; then, the pseudo-random proportional transfer rule of the ant colony system is adopted: in the utilization branch, the transfer with the greatest comprehensive attractiveness is selected, and in the exploration branch, the transfer is selected from the remaining legitimate next transfers other than the transfer with the greatest comprehensive attractiveness according to the normalized attractiveness probability; two or more complete network ants can be constructed independently in parallel working process; after a single complete network ant is completed, all origin-destination routes and the spatial relationships between different origin-destination routes included in the complete network ant are evaluated at once;

[0025] For each horizontal segment, a local forward axis is established based on the start and end points of each route. A local lateral axis orthogonal to this local forward axis is then established. Reference control points are set along the local forward axis according to the target segment length. The lateral offset of the internal control points is accumulated segment by segment using five increments {-2, -1, 0, 1, 2} and multiplied by the lateral offset unit. The starting and ending control points are fixed at the starting point S. k and endpoint E k Adjacent control points are connected to form a horizontal flight segment; based on the space occupancy index established by the environmental snapshot, only the buildings and restricted airspace units covered by the bounding box after the current flight segment envelope expansion are queried to calculate the minimum allowable altitude of each control point; the continuous altitude sequence is obtained by decoding the continuous normalized altitude variables of each control point according to formula (1); under the constraints of fixed origin and destination altitude, allowable flight altitude range, building height and vertical clearance, maximum climb angle and maximum descent angle, forward projection and reverse projection are alternately performed on the continuous altitude sequence, under the condition that the upper and lower limits of the altitude of each control point and the adjacent altitude difference limit are met. This minimizes the total adjustment of the projected altitude relative to the original decoded altitude; the range of consecutive segments where different origin-destination route envelopes simultaneously fail to meet both horizontal and vertical safety intervals constitutes a conflict zone; if there are two or more consecutive altitude adjustment schemes q that can eliminate the conflict, then the additional propulsion energy consumption Δf1(q) and the additional three-dimensional maneuver cost Δf2(q) are calculated separately, schemes whose increments Δf1(q) and Δf2(q) are simultaneously dominated by other schemes are deleted, and the two increments Δf1(q) and Δf2(q) of the remaining schemes are normalized according to the value range of this candidate set, and then selected. The minimum solution; among which This represents the normalized increase in propulsion energy consumption under scheme q. This represents the normalized increment of the cost of adding new three-dimensional maneuvers under scheme q.

[0026] If the difference between the maximum normalized increments of the two schemes does not exceed the preset numerical precision, then calculate the total height change H(q) = Σ k Σ i |z' k,i (q)-z k,i |And select the scheme with smaller H(q) from the two schemes, where z' k,i (q) represents the altitude of the i-th control point on the k-th origin-destination route after the projection of scheme q, z k,i Before adjusting the altitude; if the difference in total altitude change does not exceed the preset altitude accuracy, select according to the pre-fixed scheme number; if the conflict cannot be eliminated within the allowable altitude and climb and descent constraints, mark the corresponding complete route network as infeasible;

[0027] The projected complete route network is recalculated based on the actual three-dimensional route segments to include four target values: propulsion energy consumption, three-dimensional maneuvering cost, ground personnel casualty risk, and noise exposure. At the same time, positive residuals are calculated for altitude overruns, insufficient clearance in buildings or no-fly zones, intrusion into non-associated take-off and landing point protection zones, yaw angle overruns, climb or descent angle overruns, and insufficient intervals between different OD routes. The sum of the positive residuals is recorded as the constraint violation amount, and the target values ​​and constraint violation amounts are recorded independently.

[0028] Step 3 includes:

[0029] For a complete airway network that has passed full verification and whose constraint violation amount is no greater than the preset numerical tolerance, four minimization objectives are calculated: propulsion energy consumption, three-dimensional maneuvering cost, ground personnel casualty risk, and noise exposure. Propulsion energy consumption is calculated based on the changes in flight segment length and altitude, as well as the aircraft performance model. Three-dimensional maneuvering cost is calculated based on changes in horizontal turning and climb and descent. Ground personnel casualty risk is calculated based on the exposure of population, road vehicles, and important facilities below the airway and the accident consequences model. Noise exposure is calculated based on the aircraft flight status, propagation distance, and ground-sensitive receptors.

[0030] The benefit population and the safety population are two pre-defined complete network ant colonies, not automatically generated from environmental data; the benefit population is non-dominated and sorted according to minimizing propulsion energy consumption and minimizing three-dimensional maneuver costs, while the safety population is non-dominated and sorted according to minimizing the risk of casualties to ground personnel and minimizing noise exposure.

[0031] Benefits File A E Formed by a complete route network that has undergone full verification and is non-dominatedly sorted according to propulsion energy consumption and three-dimensional maneuver costs, Safety File A SA complete route network, fully validated and sorted according to ground personnel casualty risk and noise exposure, is formed. Benefit profiles and safety profiles are updated after each ant colony iteration. When a profile exceeds its capacity, it is sorted according to the following rules: networks with smaller non-dominated level numbers have higher priority; networks with larger congestion distances within the same non-dominated level have higher priority, and the sorted network priorities are retained until the profile capacity limit is reached. Boundary networks have infinite congestion distances and are prioritized; when congestion distances are the same within the same level, networks with smaller normalized distances to the corresponding target group's ideal point are prioritized. The benefit ideal point is determined by the minimum propulsion energy consumption and the minimum three-dimensional maneuver cost in the first non-dominated front of the current benefit profile. The safety ideal point is determined by the minimum ground personnel casualty risk and the minimum noise exposure in the first non-dominated front of the current safety profile. The normalized distance from a candidate network to its corresponding ideal point is calculated by normalizing the corresponding two targets within the range of the current first non-dominated front. Simultaneously, a negotiation candidate profile A is maintained. B And four target record files A4, A B A4 is used to store complete feasible networks that can access both sides' comparisons and representative network selections. It records the joint non-dominated relationships and overall trade-offs among the four minimization objectives: propulsion energy consumption, three-dimensional maneuver costs, ground personnel casualty risk, and noise exposure. It does not replace A4. E and A S Their respective sorting criteria;

[0032] Pheromones are stored separately according to the actual state transition e and the target category, and the update relationship is as follows:

[0033] (2),

[0034] in, Let p represent the pheromone intensity of searcher p at the actual state transition e before the end of the g-th ant colony iteration; p∈{E,S}, where E represents the beneficiary and S represents the safe. This represents the pheromone evaporation rate of searcher p; This represents the current generation deposition amount for the actual state transition e in the g-th ant colony iteration; the actual three-dimensional consensus pheromone uses independent pheromone memory, intensity upper limit, and volatility;

[0035] After each ant colony iteration, a batch local update is first performed on the actual state transitions visited in this generation, and then evaporation is executed. Finally, the current generation's elite ants and benefit profile A are processed. E The sparse, high-quality network in the middle is responsible for the deposition of pheromones by the benefit side, which is composed of the current generation of elites and security archives A. SThe sparse, high-quality networks in the data are used for pheromone deposition by the safe side; among them, the current-generation elite of the benefit side is selected from the contemporary feasible complete route network after non-dominated ranking according to propulsion energy consumption and three-dimensional maneuver costs; the current-generation elite of the safe side is selected from the contemporary feasible complete route network after non-dominated ranking according to ground personnel casualty risk and noise exposure; the sparse, high-quality networks in the historical archive are selected in ascending order of non-dominated level number and descending order of congestion distance at the same level; when a complete feasible network has not yet been formed, only local origin-destination routes that have passed the checks of flight altitude, building clearance, climb and descent angles, yaw angles, endpoint requirements, and different origin-destination route intervals are allowed to provide structural clues at a preset proportion lower than the deposition amount of the complete feasible network; infeasible complete route networks are not allowed to enter A. E A S A B Or A4.

[0036] Step 4 includes:

[0037] Actual 3D consensus detection only compares benefit profile A E First Non-Domination Frontier and Security Archive A S In the first non-dominated frontier, two actual three-dimensional centerline segments corresponding to the same origin-destination route and the same segment number are considered. When the directional difference between the two actual three-dimensional centerline segments is no greater than the directional tolerance, the horizontal distance is no greater than the route envelope tolerance, the altitude difference is no greater than the altitude tolerance, and they have a positive overlap length along the line, the overlapping portion is determined to be the actual three-dimensional consensus. The benefit parent network starts from benefit file A. E Selected from the first non-dominated frontier, within the same frontier, networks with larger congestion distances are preferred. If congestion distances are the same, networks with smaller normalized distances to the ideal benefit point determined by propulsion energy consumption and three-dimensional maneuver costs are selected. The safety parent network is selected from safety file A. S In the first non-dominated frontier, networks with larger crowding distances are selected first within the same frontier. When the crowding distances are the same, networks with smaller normalized distances to the safe ideal point determined by the risk of injury or death to ground personnel and noise exposure are selected.

[0038] Segments that intersect at a plane, have the same action code but different actual locations, are located at different altitudes, or belong to different OD routes do not constitute a consensus;

[0039] The actual three-dimensional overlap length and the benefit parent network in benefit file A E The cutting-edge quality and security parent network in Security Archive A SThe frontier quality in the ant colony jointly determines the amount of consensus pheromone deposition; consensus pheromone adopts an independent intensity upper limit to prevent consensus pheromone from covering the respective pheromones of the benefiting party and the safe party, and decays at an independent volatility rate in each ant colony iteration. Accidental overlaps that do not continuously obtain the support of the high-quality frontiers of the benefit archive and the safe archive decay with each iteration.

[0040] When the benefit file A E and security file A S After forming the first non-dominated frontier consisting of a complete feasible network, the endpoints of the same origin-destination route, the same segment number, directional difference, horizontal spacing and height difference are all within the tolerance and have positive overlap length, and adjacent and continuous actual three-dimensional consensus segments are merged into consensus intervals; continuous segments in the same origin-destination route other than consensus intervals constitute non-consensus intervals.

[0041] From the benefit file A E The first non-dominated frontier selects the benefit parent network and the security archive A. S The first non-dominated frontier selects a safety parent network to generate negotiated offspring: In the consensus interval, common segments where the selected benefit parent network and the selected safety parent network actually overlap in three dimensions at the same origin-destination route and segment number are retained. In the non-consensus interval, the benefit parent network or safety parent network is inherited according to the interval, and switching only occurs at compatible anchor points. The compatible anchor point is a corresponding control point with the same index j in both the benefit parent network and the safety parent network, located at the boundary between the consensus interval and the non-consensus inheritance interval in the same origin-destination route. The horizontal distance between these corresponding control points is no greater than the route envelope tolerance, and the altitude difference is no greater than the altitude tolerance. Furthermore, the altitude difference between the corresponding control points in the benefit parent network and the safety parent network, as well as the changes in direction and altitude of adjacent segments after switching, all satisfy the yaw angle after continuous altitude projection. Maximum climb angle, maximum descent angle, and constraints on the planning area boundary, static and temporary no-fly zones, non-associated take-off and landing protection zones, building clearance, permissible flight altitude, origin and destination endpoints, and horizontal and vertical safety separation between different origin and destination routes; negotiated offspring are not a third independent population. Benefit ants or safety ants constructed independently based on their own pheromones and heuristic information are called independently constructed ants; negotiated offspring and independently constructed ants are subject to the same continuous altitude projection, planning area boundary, no-fly zones, take-off and landing protection zones, building clearance, flight altitude, maneuverability, and route separation checks, and respectively calculate four minimization objectives: propulsion energy consumption, three-dimensional maneuvering cost, ground personnel casualty risk, and noise exposure. Only when they maintain competitiveness in the benefit or safety ranking will they be included in the corresponding file.

[0042] Step 5 includes:

[0043] After reaching the preset number of iterations, evaluation budget, or convergence stopping conditions, and completing the final update of the benefit and safety files, first search for benefit file A in the decision space of the complete network ant and the decoded actual three-dimensional path space. E First Non-Domination Frontier and Security Archive A S The first non-dominated front points to a common set of the same complete route network; if the common set is not empty, then the network with the smallest larger value of the benefit-side normalized loss and the safety-side normalized loss is selected from the common set; if the common set is empty, then the network that has passed complete verification and comes from A... E A S A B In the candidate complete route network of A4, the joint non-dominated hierarchy is obtained by sequentially performing non-dominated ranking based on four minimization objectives: propulsion energy consumption, three-dimensional maneuvering cost, ground personnel casualty risk, and noise exposure. The candidate network is then ranked within the A4 range. E and A S The non-dominated levels on the corresponding target groups are screened; the complete route networks whose normalized losses exceed the preset threshold or whose corresponding non-dominated levels are higher than the preset levels are excluded from the benefit target groups consisting of propulsion energy consumption and three-dimensional maneuver costs, or the safety target groups consisting of ground personnel casualty risks and noise exposure. The relative satisfaction of the remaining candidate files is calculated, and the balanced representative network is determined according to formula (3):

[0044] X*=argmax(X∈P B )min{s E (X), s S (X)}(3)

[0045] Among them, P B To form a negotiated candidate set after screening through complete feasibility, joint non-dominant hierarchy, and benefit and security archive hierarchy, s E (X) and s S (X) represents the normalized satisfaction of candidate network X with respect to the benefit side and the normalized satisfaction with respect to the safety side, respectively; this rule maximizes the satisfaction of the weaker party only in benefit profile A. E First Non-Domination Frontier and Security Archive A S The first non-dominated frontier is activated when there is no common network; X* represents the final selected balanced representative network; complete feasibility means that all network-level hard constraints are passed; the joint non-dominated hierarchy refers to the hierarchy obtained by jointly sorting the non-dominated levels according to the four minimization objectives of propulsion energy consumption, three-dimensional maneuver cost, ground personnel casualty risk and noise exposure; Equation (3) adopts the discrete Kalley-Smolotinski negotiation selection rule to select min{s E (X), s S(X)} The largest candidate network; when the difference in minimum satisfaction of two or more candidates does not exceed the preset numerical precision, the difference between the satisfaction of the benefit party and the satisfaction of the safety party, the sum of the satisfaction of the two parties, and the above-mentioned joint non-dominated level are compared in turn.

[0046] The initial planning uses a cold-start ant population and evaluation budget, and saves the balanced representative network selected in step 5 for review, simulation, or release as the released network, and saves the benefit file A generated from the initial planning. E Security File A S Negotiation Candidate File A B The four target record files A4, and the environmental snapshot used when generating the published network are used as the baseline environmental snapshot; during operation, new risk fields and no-fly zone statuses are periodically received, and the complete feasibility of the published network is first reviewed under the current environmental snapshot, and then it is determined whether replanning is triggered according to formula (4):

[0047] trigger(t)=[ΔR(t)≥θ R ]∨[ΔZ(t)≥θ Z ]∨[V jw (t)>ε](4)

[0048] Where t represents the current monitoring time; trigger(t) indicates whether replanning is triggered at time t; ΔR(t) represents the normalized change of the current risk field relative to the risk field in the baseline environmental snapshot; ΔZ(t) represents the geometric change of the current no-fly zone or airspace restrictions relative to the no-fly zone or airspace restrictions in the baseline environmental snapshot; θ R Indicates the threshold for triggering changes in the risk field; θ Z Indicates the threshold for triggering changes in no-fly zones or airspace; V jw (t) represents the number of constraint violations of the published network under the current environmental snapshot; ε represents the feasibility numerical tolerance.

[0049] In step 5, when ΔR(t) ≥ θ R ΔZ(t)≥θ Z V jw Hot reprogramming is initiated when any of the conditions in (t)>ε are met; otherwise, the feasibility and target value of the existing network are only reviewed in the new environment. Hot reprogramming uses an independently configurable number of ants and an evaluation budget, with the most recently effective planning resulting in A. E A S A B The old file is formed by combining A4 and the network that is still feasible after being reviewed in the new environment. The network is retained as a seed, and the affected origin and destination routes or segments are disturbed and the reconstructed network is supplemented. The pheromone is reinitialized in this round. After completion, a new time stamp, network version and verification report are generated. The hot replanning is a replanning process relative to the cold start initial planning.

[0050] The present invention also provides an electronic device, including a processor and a memory, the memory storing program code that, when executed by the processor, causes the processor to perform the steps of the method.

[0051] The present invention also provides a storage medium storing a computer program or instructions that, when the computer program or instructions are run on a computer, execute the steps of the method described.

[0052] Beneficial Effects: This invention uses the complete route network as the basic object for planning and evaluation, enabling unified verification of overhead clearings, no-fly zones, takeoff and landing point protection, and route separations between different origins (ODs) under the same environmental snapshot. This reduces the need for repeated repairs due to combination conflicts after individual routes are generated separately. Both the benefit provider and the safety provider retain their respective non-dominated frontiers. The actual three-dimensional consensus only reinforces the same OD segments that are continuously mutually recognized by both parties. Negotiated offspring still need to be verified against the original ordering and complete constraints of both parties. Therefore, the operating entity can verify energy consumption and maneuverability, while the management entity can verify personnel risks, noise, and airspace constraints, and trace the selection basis of the final network relative to the frontiers of both parties. The output centerline, continuous altitude profile, route envelope, target value, verification results, and environmental version can be directly integrated into the planning review and simulation verification process, suitable for generating schemes for fixed or periodic route networks such as urban logistics distribution, vertical takeoff and landing short-haul transportation, and emergency support.

[0053] During the operation and maintenance phase, this invention distinguishes between "re-verifying the existing network" and "triggering replanning" based on the degree of change in risk fields and temporary no-fly zones: if environmental changes do not reach a threshold, a complete search is not repeated; if the threshold is reached, hot replanning is carried out using the old network and archives verified by the new environment, forming a new route network. This approach preserves the existing planning information of unaffected segments while enabling timely updates for segments affected by road traffic hotspots, temporary activity airspace restrictions, or partial airspace removal, thus providing continuous technical support for the low-altitude operation platform from initial network establishment and mutual review to dynamic maintenance. Attached Figure Description

[0054] Figure 1 This is a schematic diagram of the overall process of the dynamic low-altitude public airway network construction and replanning method of the present invention.

[0055] Figure 2 A schematic diagram illustrating the synchronous construction of all origin and destination routes for a single complete network ant.

[0056] Figure 3 This is a schematic diagram illustrating how the benefit-seeking party and the security party conduct in-process negotiations through actual three-dimensional consensus.

[0057] Figure 4A flowchart illustrating the process of environmental change verification and event-triggered hot recalculation planning.

[0058] Figure 5 This is a schematic diagram showing the spatial distribution of road vehicle exposure at twelve monitoring times.

[0059] Figure 6 This is a schematic diagram of the spatial distribution of population exposure at twelve monitoring times.

[0060] Figure 7 The time-series curves show the changes in population in commercial areas, population in residential areas, and road vehicle exposure.

[0061] Figure 8 This is a schematic diagram of a representative low-altitude public airway network that has passed the full constraint verification under static initialization conditions.

[0062] Figure 9 A schematic diagram of the dynamic risk field and deployment route network response from 00:00 to 06:00 for twelve monitoring times.

[0063] Figure 10 A schematic diagram of the dynamic risk field and deployment route network response from 08:00 to 14:00 for twelve monitoring times.

[0064] Figure 11 A schematic diagram of the dynamic risk field and deployment route network response at twelve monitoring times from 16:00 to 22:00. Detailed Implementation

[0065] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments, and the advantages of the present invention in the above and / or other aspects will become clearer.

[0066] This invention provides a method for constructing a dynamic low-altitude public airway network based on multi-target ant colony negotiation, generating a continuously updated three-dimensional public airway network for scenarios such as urban logistics distribution, vertical takeoff and landing short-haul transportation, and emergency support. The overall process of this invention is as follows: Figure 1As shown, the process includes, in sequence: establishing an environmental snapshot and encoding the complete network; synchronously constructing all origin-destination routes using a single complete network ant; implementing feasibility for the complete network and calculating four objectives: propulsion energy consumption, three-dimensional maneuver costs, ground personnel casualty risk, and noise exposure; independent search by the benefit-seeking and safety-seeking parties, and conducting in-process negotiation through actual three-dimensional consensus; selecting a balanced representative network under the condition of prioritizing commonly feasible networks; reviewing environmental changes, and retaining the published network or initiating hot recalculation based on event triggering conditions, outputting an executable three-dimensional common route network and its version record. In this invention, "true three-dimensional consensus" and "actual three-dimensional consensus" both refer to three-dimensional common routes that meet the conditions of direction, horizontal distance, altitude difference, and positive overlap length at the same origin-destination route and the same segment number. The following parameters are used to illustrate one possible implementation of this invention and do not constitute a limitation on the scope of protection.

[0067] Step 1: Create an environmental snapshot and represent the complete route network;

[0068] Step 1: Create an environmental snapshot and represent the complete route network. For example... Figure 1 As shown, the embodiment writes urban 3D building data, population grid, road network and vehicle status, critical facilities, static and temporary no-fly zones, take-off and landing points, and aircraft performance into an environmental snapshot under a unified coordinate datum at the same time. Building data is used to determine the minimum airspace clearance along the route envelope; population, vehicle, and critical facility data are used to form inputs for personnel casualty risk assessment; no-fly zones and non-associated vertical take-off and landing point protection zones are used to form an intrusive space occupancy index; aircraft performance is used to set permissible flight altitude, speed, mass, and maximum yaw, climb, and descent angles. The environmental snapshot includes the sampling time and version number, which is read by all complete network ants in the same planning round.

[0069] like Figure 2 As shown, suppose the complete network contains K origin-to-origin routes, and the k-th origin-to-origin route contains B k Each ant simultaneously stores the horizontal and altitude states of K origin-end routes, rather than representing only one route. A local forward axis is established based on the start and end points of each origin-end route, and a local horizontal axis orthogonal to this local forward axis is also established. The horizontal state is accumulated segment by segment using five increments: shifting two units to the left, shifting one unit to the left, holding, shifting one unit to the right, and shifting two units to the right. The pheromone address simultaneously contains the origin-end route number, segment number, current actual state, and next actual state, ensuring that the same action number is not mistaken for the same spatial segment in different preceding states. Altitude uses a continuously normalized variable h. k,iEncode and decode according to formula (1). OD is an abbreviation for Origin-Destination, representing the origin-to-destination pair; in this embodiment, 5 take-off and landing points are set, and 10 origin-to-destination routes are formed by all unordered take-off and landing point pairs. The number of each route is determined with a target route length of 100m, the lateral offset unit is 50m, and the first and last control points are fixed. The decision vector of a complete network ant simultaneously represents the entire horizontal state and continuous altitude state of the 10 origin-to-destination routes.

[0070] (1),

[0071] in The actual altitude of the i-th control point on the k-th OD route. and These represent the lower and upper limits of the permitted flight altitude, respectively. The limited altitude box is only used to merge pheromone memories at similar altitudes and does not restrict aircraft to flying at a fixed altitude level. ∈[0,1]; and Based on the planned airspace and aircraft performance, this embodiment can take 50m and 120m respectively, and the take-off and landing point endpoint height is taken as 50m. After obtaining the original continuous height using equation (1), the building clearance, endpoint height, and maximum climb and descent angle constraints are satisfied through forward and backward projection in step 2. The height interval used to merge similar height pheromone memories does not constitute a mandatory flight altitude layer.

[0072] Step 2: Simultaneously construct the complete network and implement hard constraint feasibility.

[0073] Step 2: Simultaneously construct the complete route network and implement hard constraint feasibility. For example... Figure 2 As shown, in the joint construction round of the s-th segment, each active origin-destination route generates a legal next transfer set based on the same initial frozen pheromone snapshot and the same environmental snapshot, and submits it uniformly after all origin-destination routes have completed this round of selection. The benefit ant uses the incremental evaluation of propulsion energy consumption and three-dimensional maneuver costs as the main heuristic, while the safety ant uses the incremental evaluation of ground personnel casualty risk and noise exposure as the main heuristic. In this embodiment, the pheromone index α can be 1.0, the heuristic index β can be 1.2, the pseudo-random utilization probability q0 can be 0.40, and the upper limit of the heuristic correction gain can be 0.20. Optimal attraction reuse and probabilistic exploration are performed using pseudo-random proportional rules. Multiple complete network ants construct independently in parallel processes; the same ant independently samples each origin-destination route in the same construction round, and after all active routes have completed this round of selection, they are uniformly advanced and uniformly decoded after forming a complete network. A hard constraint check and four-item objective evaluation are performed on the 10 origin-destination routes and their mutual spatial relationships.

[0074] For each generated horizontal flight segment, the spatial occupancy index of the environmental snapshot is first used to query the buildings, no-fly zones, and non-associated take-off and landing point protection zones near the route envelope to obtain the minimum allowable altitude for each control point. The ratio of the altitude difference to the horizontal distance between adjacent control points must not exceed the tangent of the maximum climb angle or the maximum descent angle. In this embodiment, both the maximum climb angle and the maximum descent angle are set to 20°. Forward and backward projections are alternately performed within the upper and lower altitude limits until the maximum altitude change between two adjacent projections does not exceed 10°. -9 m; for example, if the original altitude is 70m, the lower clearance is 78m, and the upper reachable altitude is 90m, the point is projected to 78m, without further arbitrary increases. If multiple consecutive altitude adjustment schemes can eliminate conflicts between different origin and destination routes, the additional propulsion energy consumption Δf1 and the additional three-dimensional maneuver cost Δf1 for each scheme are calculated separately. Schemes that are simultaneously dominated by both increments are deleted. The remaining schemes are normalized according to the value range of the current candidate set, and the scheme with the smallest larger of the two normalized increments is selected; the difference between the above maximum normalized increments shall not exceed 10. -6 Then, continue comparing H(q)=Σ k Σ i |z' k,i (q)-z k,i Choose the option with the smaller change in total height; the difference in H(q) should not exceed 10. -6 When m, select according to the fixed scheme number. The horizontal safety separation of different origin and destination routes is 30m, and the vertical safety separation is 20m; origin and destination routes sharing the same terminal are exempt from mutual separation checks only within a 200m×200m square area centered on that terminal. If conflicts cannot be eliminated within the allowable altitude and climb / descent constraints, the complete route network is judged as infeasible; the complete feasibility numerical tolerance ε in this embodiment is 10. -12 .

[0075] Step 3: Create non-dominant files for both parties and update pheromones;

[0076] Step 3: Create non-dominated profiles for both parties and update pheromones. For a complete route network that has passed full verification, calculate propulsion energy consumption according to the existing aircraft energy consumption model, calculate the three-dimensional maneuvering cost according to horizontal turning and vertical altitude changes, calculate the risk of ground personnel casualties according to the exposure and accident consequences model of ground population, vehicles, and important facilities, and calculate noise exposure according to aircraft noise source intensity, propagation distance, and ground sensitive receptors. The benefit population updates the benefit profile A only based on propulsion energy consumption and three-dimensional maneuvering cost. E The safety population updates safety records only based on the risk of injury or death to ground personnel and noise exposure. S Negotiation Candidate File A BThe four-objective record file A4 is used for candidate comparison and joint recording of the four minimization objectives, without changing the dual-objective non-dominated ordering of the benefit file and safety file. Both the benefit file and safety file have a capacity of 100 complete route networks. When a file overflows, it is retained sequentially according to the non-dominated level number from smallest to largest, and the congestion distance at the same level from largest to smallest. The congestion distance of the boundary network is taken as infinite. When the congestion distances at the same level are the same, the benefit file selects the network with the smaller normalized distance of the ideal benefit point determined by the minimum propulsion energy consumption and the minimum three-dimensional maneuver cost in the first non-dominated frontier of the current benefit file. The safety file selects the network with the smaller normalized distance of the ideal safety point determined by the minimum ground personnel casualty risk and the minimum noise exposure in the first non-dominated frontier of the current safety file. The aforementioned normalized distance is calculated after normalization of the corresponding two objectives within the value range of the current first non-dominated frontier.

[0077] Pheromones are stored according to their origin and destination route numbers, segment numbers, current actual lateral-altitude state, and next actual lateral-altitude state. After each ant colony iteration, local updates, evaporation, and global deposition are performed; the beneficiary side consists only of the beneficiary side's current elites and beneficiary profile A. E The sparse, high-quality network deposits pheromones from the security side, which are only generated by the security side's contemporary elites and security archives A. S The sparse, high-quality network in the model deposits the local pheromone. The consensus pheromone is stored separately with independent upper limits and volatilization rates; in this embodiment, the volatilization rates of the benefit pheromone, the security pheromone, and the consensus pheromone can all be set to 0.12, and the lower and upper limits of the pheromone can each be set to 10. -3 And 20, consensus pheromones shall not exceed 30% of the upper limit of the corresponding pheromone for the benefiting party or the safe party. When a complete feasible network has not yet been formed, the amount of local origin-destination route clues shall not exceed 10% of the amount of complete feasible network clues, and infeasible complete route networks shall not enter any non-dominated file.

[0078] (2),

[0079] Where, τ e,pg ρ represents the pheromone intensity of searcher p at the actual state transition e in the g-th ant colony iteration. p Δτ represents the pheromone evaporation rate of searcher p. e,pgThis represents the amount of deposition for the actual state transfer in the g-th ant colony iteration; p=E represents the beneficiary side, and p=S represents the safe side. After each generation, batch local updates and evaporation are first performed on the transfers visited in the current generation, and then global deposition is performed by the beneficiary side's current generation elites, the sparse high-quality network in the beneficiary archive, and the safe side's current generation elites and the sparse high-quality network in the safe archive, respectively; the consensus pheromone formed in step 4 is updated separately according to the consensus rules. When a complete feasible network has not yet been formed, only local origin-destination routes that have passed physical checks and route conflict checks are allowed to provide construction clues with low intensity, and infeasible complete networks are not allowed to enter the non-dominated archive.

[0080] Step 4: Conduct in-process negotiations using the three-dimensional common route;

[0081] Step 4: Conduct in-process negotiations based on the actual three-dimensional common route. For example... Figure 3 As shown, this embodiment only compares benefit file A. E First Non-Domination Frontier and Security Archive A S In the first non-dominant frontier, the actual three-dimensional centerline segments with the same origin and destination routes and the same segment number are considered; the directional tolerance is 15°, the horizontal tolerance is 10m, and the altitude tolerance is 10m, with a positive overlap length required. Segments with planar intersections, different origin and destination segments, the same action code but different actual positions, and altitude differences exceeding the tolerances are not included in the consensus. The benefit parent network originates from benefit file A. E Selected from the first non-dominated frontier, within the same frontier, networks with larger congestion distances are preferred. If congestion distances are the same, networks with smaller normalized distances to the ideal benefit point determined by propulsion energy consumption and three-dimensional maneuver costs are selected. Safety parent networks are selected in the same order from safety file A. S The consensus pheromone is selected from the first non-dominated frontier, and the normalized distance to the ideal safe point determined by the risk of casualties among ground personnel and noise exposure is used as a co-selective criterion. The consensus deposition increases with the actual overlap length and the frontier quality of the benefit parent network and the safety parent network, and is subject to an independent upper limit constraint; consensus pheromones that do not continuously receive support from two frontiers decay with evaporation.

[0082] When the benefit file A E and security file A SAfter all feasible first non-dominated frontiers are formed, adjacent actual three-dimensional consensus segments are merged into consensus intervals. Within the consensus interval, common segments where the selected benefit parent network and the selected safety parent network actually overlap in three dimensions at the same origin-destination route and segment number are retained. In non-consensus intervals, the benefit parent network or safety parent network is inherited according to the interval, and switching only occurs at compatible anchor points. A compatible anchor point is a corresponding control point at the boundary between the consensus interval and the non-consensus inheritance interval within the same origin-destination route, and which has the same sequence number j in both the benefit parent network and the safety parent network. This embodiment requires that the horizontal distance between two corresponding control points is no greater than 10m, the height difference is no greater than 10m, and the connected adjacent segments satisfy the constraints of yaw angle, maximum climb angle, and maximum descent angle. The height difference between the control points corresponding to the benefit parent network and the control points corresponding to the safety parent network, as well as the changes in direction and height of adjacent flight segments after switching, are all satisfied by continuous height projection to meet the constraints of the planning area boundary, static and temporary no-fly zones, non-associated take-off and landing point protection zones, building clearance, permissible flight altitude, origin and destination endpoints, and horizontal and vertical safety separation between different origin and destination routes. After complete verification and calculation of propulsion energy consumption, three-dimensional maneuver costs, ground personnel casualty risks, and noise exposure, the negotiated offspring are only included in the corresponding file and participate in subsequent pheromone updates if they maintain competitiveness in the non-dominated ranking of the benefit or safety side.

[0083] Step 5: Identify the representative network and implement event-triggered replanning;

[0084] Step 5: Identify the representative network and implement event-triggered replanning. First, verify the benefit file A. E First Non-Domination Frontier and Security Archive A S Does the first non-dominated frontier consist of a single complete route network? If so, select the network from the common set that minimizes the larger of the normalized loss for the benefit side and the normalized loss for the safety side. If not, select the network from A that has already passed complete verification. E A S A B Among the A4 candidates, the joint non-dominated hierarchy is first obtained by ranking the candidates according to the four minimization objectives of propulsion energy consumption, three-dimensional maneuver cost, ground personnel casualty risk, and noise exposure. Then, the candidates are further filtered according to the non-dominated hierarchy of the candidate network on the corresponding objective groups in the benefit and safety archives to obtain the negotiated candidate set P. B And select a balanced representative network according to equation (3).

[0085] X*=argmax(X∈P B )min{s E (X), s S (X)}(3)

[0086] Among them, P BTo form a negotiation candidate set after screening through complete feasibility, joint non-dominant hierarchy, and benefit and security archives levels, s E (X) and s S (X) represents the normalized satisfaction of candidate complete route network X with respect to the benefit side and the safety side, respectively, which is obtained by transforming the normalized target loss of X relative to the first non-dominated frontier of the corresponding side; X* is the finally selected balanced representative network. During candidate screening, complete route networks with normalized losses exceeding 0.40 or corresponding non-dominated levels higher than the 3rd level are excluded from the benefit target group consisting of propulsion energy consumption and three-dimensional maneuver costs, or the safety target group consisting of ground personnel casualty risk and noise exposure. Equation (3) adopts the discrete Kalley-Smolotinski negotiation selection rule to select the candidate with the highest satisfaction of the weaker side; when the difference between the minimum satisfaction of the two candidates does not exceed 10 -6 The system sequentially compares the difference between the satisfaction levels of the benefit-seeking party and the safety-seeking party, the sum of the satisfaction levels of both parties, and the joint non-dominated hierarchy obtained by ranking the four minimization objectives: propulsion energy consumption, three-dimensional maneuvering costs, ground personnel casualty risk, and noise exposure. The final output includes the centerline of each origin-destination route, continuous altitude profile, route envelope, four target values, constraint verification results, and environmental snapshot version number.

[0087] The initial planning uses a cold-start ant count and evaluation budget, and after completion, saves the published balanced representative network and benefit archive A. E Security File A S Negotiation Candidate File A B Four target record files (A4) and their baseline environment snapshots. Upon arrival of new data during operation, a new environment snapshot is first generated. Under the current environment snapshot, the complete feasibility of the published network is reviewed, and the risk field change ΔR(t) and spatial domain change ΔZ(t) are calculated respectively. The event trigger judgment process is as follows: Figure 4 As shown:

[0088] trigger(t)=[ΔR(t)≥θ R ]∨[ΔZ(t)≥θ Z ]∨[V jw (t)>ε](4)

[0089] Where ΔR(t) and ΔZ(t) are the normalized change of the risk field and the geometric change of the spatial set at time t relative to the most recent actual replanning baseline, respectively, and θ R and θ Z V is the corresponding trigger threshold. jw (t) represents the constraint violation amount of the published network under the current environmental snapshot, and ε represents the numerical tolerance for complete feasibility.

[0090] In this embodiment, θ is taken. R =0.30、θZ =0.05, ε=10 -12 Any change that reaches a threshold or V jw Hot reprogramming is initiated when (t)>ε; otherwise, the complete feasibility of the published network and its propulsion energy consumption, 3D maneuver costs, ground personnel casualty risks, and noise exposure are only reviewed under the new environment snapshot. Hot reprogramming uses independently configurable ant numbers and evaluation budgets, retaining networks from the old archives that remain feasible after review under the new environment as seeds. Priority is given to perturbing origin-destination routes covered by changes in the risk field grid or the no-fly zone, as well as segments where new constraints are violated, and supplementing the reconstructed complete route network. The pheromone is reinitialized in this round to avoid locking new searches with old environment preferences. Upon completion, a new time stamp, network version, and verification report are generated.

[0091] The system implementing this method includes modules such as environmental snapshots, complete network construction, hard constraint detection, propulsion energy consumption and 3D maneuver cost evaluation, ground personnel casualty risk and noise exposure assessment, bilateral profiles and pheromones, actual 3D consensus negotiation, representative network selection, event monitoring and cold / hot budget replanning, and result storage interface. It can be deployed on a low-altitude route planning server or operations management platform. The system receives geographical, risk, and airspace status data through a data interface and outputs a versioned 3D route network to the review or simulation module, providing planning and update decision support for legally conducted airspace reviews, operational permits, and flight releases.

[0092] To verify the above technical solution, this embodiment uses a general-purpose computer to complete network construction and dynamic replanning. The computer is configured with an AMD Ryzen 7 7840H eight-core processor, 32GB of memory, running a 64-bit operating system and MATLAB R2024b, and enabling a parallel pool of eight worker processes. This computer, memory, and software operating environment are all existing general-purpose devices and do not constitute a limitation of this invention on dedicated hardware structures. The technical improvements of this invention are reflected in the complete network coding, independent files for both parties, actual three-dimensional consensus negotiation, and event-triggered replanning data processing flow.

[0093] The implementation area is a 5km × 5km urban low-altitude area, divided into 50 × 50 planar grids, each with a side length of 100m. The environmental snapshot shows 218 buildings occupying the grid, with building heights ranging from 11 to 91m and a median of 37m. The grid coordinates and altitudes of the five take-off and landing points N1–N5 are (9, 11, 50m), (29, 14, 50m), (15, 43, 50m), (28, 39, 50m), and (45, 43, 50m), respectively. Ten take-off and landing routes are formed from all the disordered take-off and landing point pairs. The center of the non-associated vertical take-off and landing point protection zone is set at (14, 10), and the center of the temporary no-fly zone is set at (27, 26). The aircraft mass is taken as 1.38 kg, the cruising speed as 20 m / s, the permissible flight altitude as 50–120 m, the maximum climb angle and maximum descent angle as 20°, the route envelope width as 10 m, and the horizontal and vertical safety separations for different origin and destination routes as 30 m and 20 m, respectively. The complete feasibility numerical tolerance is taken as 10. -12 The projection convergence accuracy is set to 10. -9 m, the precision for comparing schemes side-by-side is taken as 10. -6 The details are shown in Table 1.

[0094] Table 1

[0095]

[0096] Step 1 yields an environmental snapshot with time and version number, building and confined airspace spatial indexes, population exposure fields, and road and vehicle exposure fields. For example... Figure 5 , Figure 6 As shown, Figure 5 and Figure 6 The horizontal and vertical axes represent grid coordinates, with each grid cell being 100m. The two figures show the spatial distribution of road vehicle exposure and population exposure from 00:00 to 22:00, with snapshots taken every 2 hours. Figure 7 The spatial integrals of commercial area population, residential area population, and road vehicle exposure are further presented over time. Commercial area population exposure reaches approximately 1.057 × 10⁻⁶ at 14:00. 6 The residential population exposure at 00:00 reached approximately 2.317 × 10⁻⁶. 5 Road vehicle exposure reached approximately 4.219 × 10 at 18:00. 4 The peak times and directions of change of the three types of inputs are not the same, thus forming a dynamic environmental basis for determining whether to review and retain the old network or trigger replanning.

[0097] Step 2 involves simultaneously advancing 10 origin-destination routes under the same environmental snapshot, and uniformly implementing checks on building clearance, no-fly zones, non-associated take-off and landing point protection zones, flight altitude, yaw angle, climb angle, descent angle, and the intervals between different origin-destination routes after forming a complete network. The cold start uses 100 complete network ants and 50,000 evaluation budgets, with both benefit and safety archives having a capacity of 100. Figure 8 In the static initialization results shown, all 10 origin and destination routes are within the allowable flight altitude range, and the complete network constraint violation is 0; the propulsion energy consumption f1 of the selected representative network is 8.03 × 10⁻⁶. 4 J, the cost of three-dimensional maneuvering f2 is 0.0615, and the risk of ground personnel casualties f3 is 4.11 × 10 -3 The noise exposure f4 is 1.71 × 10⁴. 6 This result indicates that the output object is not 10 independent single lines, but a unified network that has undergone a complete network verification and is available for auditing and simulation verification.

[0098] Steps 3 and 4 respectively generate benefit files A. E Security File A S And the actual three-dimensional consensus pheromone that is independently updated and evaporated. In a set of dynamic cold-start experiments using the same spatial settings, the feasible profile reached 100 complete networks, the first non-dominated frontier of the benefit profile contained 6 networks, and the first non-dominated frontier of the safety profile contained 18 networks; the actual three-dimensional consensus detection recorded 66 overlapping segments that received common support from both frontiers in 148 segments to be compared, with an overlap length of approximately 14.131 km, accounting for part of the total length of 29.920 km to be compared, and an average support across generations of 0.352. Thus, the benefit side and the safety side do not perform a weighted compromise only after the calculation is completed, but instead, while maintaining their respective dual-objective ranking, they convert common segments with the same origin and destination routes, the same segment number, and compatible actual positions, altitudes, and directions into consensus pheromones and feed them back to the subsequent construction; accidental overlaps that do not continue to receive support evaporate with the consensus pheromone.

[0099] Step 5 according to Figure 4The process shown performs event determination at 12 monitoring times. A cold start is performed at 00:00; in the subsequent 11 times, hot replanning is triggered at 02:00, 06:00, 08:00, 10:00, 12:00, 18:00, and 20:00, while only the old network is reviewed and retained at 04:00, 14:00, 16:00, and 22:00. After the addition of a temporary no-fly zone at 02:00, the constraint violation of the old network in the new environment is 370.588, triggering replanning. After replanning, the constraint violation of the selected network returns to 0. At 04:00, the relative change in the risk field is only 0.0214, the airspace set remains unchanged, and the old network is feasible, therefore no new ant colony search is performed. At 12:00, although the relative change in the risk field is 0.2017, lower than 0.30, the old network loses feasibility due to local changes along the published flight path, therefore V... jw (t)>ε condition triggers hot reprogramming. Figure 9 , Figure 10 and Figure 11 The dynamic risk field and the actual deployed route network after review, retention or replanning are shown in chronological order for twelve monitoring moments; the horizontal and vertical axes represent grid coordinates, each grid is 100m, the background color represents the normalized comprehensive risk, and the results of the main event judgments are listed in Table 2.

[0100] Table 2

[0101]

[0102] In the 12 snapshots of this embodiment, if a cold start method of 50,000 evaluations is used at each time point, a total of 600,000 network evaluations would be required. After adopting the event triggering and cold / hot budgeting of this invention, the actual number of network evaluations used is 50,000 + 7 × 25,000 = 225,000, reducing the cumulative evaluation budget by 62.5%, and avoiding meaningless re-optimization at the four time points where no substantial changes occurred. The complete constraint violation of the network deployment at all 12 time points is 0. The technical advantages reflected in this result are: the complete network-level construction enables the unified discovery of OD conflicts and single-line constraints in the same candidate solution; the independent archives of both parties retain the respective comparative basis for operational efficiency and public safety; the actual three-dimensional consensus enables jointly selected flight segments to participate in subsequent searches rather than just for ex-post interpretation; and the event-triggered hot replanning reduces duplicate evaluations while maintaining network feasibility, and focuses the search on OD routes and flight segments affected by risk fields or airspace changes.

[0103] The final technical result obtained in this embodiment is a set of versioned 3D common route networks that can be directly entered into the review and simulation stages, rather than just a few independent single paths or a set of objective function values. After a cold start at 00:00, the system outputs a representative network containing 10 origin-destination routes; for each origin-destination route, it simultaneously provides a 3D centerline, continuous height profile, and a 10m wide route envelope, and saves propulsion energy consumption, 3D maneuver costs, ground personnel casualty risk, noise exposure, complete constraint verification results, environmental snapshot number, and negotiation selection record along with the network version. Figure 8 The first publishable network is shown in three dimensions: centerline, route envelope, and urban built environment. Route projection on the corresponding integrated risk field is also given. Figure 9 , Figure 10 and Figure 11 The network sequences that were actually deployed after review and retention or replanning are displayed for twelve monitoring times. The final deployed networks at all 12 times meet the requirements for building clearance, no-fly zone, protection of unassociated take-off and landing points, flight altitude, climb angle, descent angle, and different take-off and destination route separations, with zero violations of integrity constraints.

[0104] The practical technical problem addressed in this embodiment is: when 10 public air routes between 5 take-off and landing points jointly occupy urban low-altitude airspace, the planning platform must avoid inter-route conflicts caused by the superposition of separately generated routes, provide a complete network with traceable negotiation basis when operational efficiency and public safety judgments are inconsistent, and determine whether the old network can continue to be used after changes in traffic flow, population exposure, or temporary no-fly zones. For example, if a new temporary no-fly zone is added at 02:00, increasing the constraint violation amount of the old network to 370.588, the system identifies the affected take-off and destination routes and segments and initiates hot replanning, restoring the constraint violation amount of the new version network to 0; if the environmental changes at 04:00 do not reach the risk field and airspace change thresholds and the old network is feasible after review, the existing network is directly retained. Therefore, the low-altitude operation management platform obtains a closed-loop result consisting of "currently deployable network, trigger or retention reason, affected scope, four target values, and complete constraint verification conclusion," which can be used for subsequent airspace review, operation simulation, and network version maintenance.

[0105] Compared to existing technologies, the superiority of this embodiment is demonstrated by the verifiable results within the same instance. First, instead of generating origin-destination routes one by one and then superimposing and checking them, this method encodes 10 origin-destination routes into a complete network candidate and uniformly checks the route constraints and route spacing in a single network-level evaluation. The constraint violations of the final deployed network at all 12 time points are 0, ensuring that the deployment object itself is a complete and feasible network. Second, instead of pre-weighting into a single objective or only selecting a compromise solution after calculation, this method retains both benefit and safety archives and transforms 66 actual three-dimensional overlapping routes supported by two first non-dominated frontiers into volatile consensus pheromones, making the source of the common routes representing the final network and the concessions made by both parties traceable data. Furthermore, compared to the approach of performing 50,000 evaluations from scratch at each of the 12 time points, this method employs one cold start, seven hot replanning iterations, and four verification-only iterations, reducing the total evaluation budget from 600,000 to 225,000, a reduction of 62.5%, while ensuring that the deployment network at all 12 time points passes the complete constraint verification. These results demonstrate that the present invention achieves the joint feasible construction of a complete route network, the traceable preservation of the basis for benefit and safety negotiations, and the reduction of redundant evaluations in dynamic environments.

[0106] This invention provides a method for constructing a dynamic low-altitude public airway network based on multi-target ant colony negotiation between two parties. Many methods and approaches exist for implementing this technical solution; the above description is merely a preferred embodiment of the invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of this invention, and these improvements and modifications should also be considered within the scope of protection of this invention. All components not explicitly stated in this embodiment can be implemented using existing technologies.

Claims

1. A method for constructing a dynamic low-altitude public airway network based on multi-objective ant colony negotiation between two parties, characterized in that, Includes the following steps: Step 1: Create an environmental snapshot and represent the complete route network; Step 2: Simultaneously construct the complete network and implement hard constraint feasibility. Step 3: Create non-dominant files for both parties and update pheromones; Step 4: Conduct in-process negotiations using the three-dimensional common route; Step 5: Identify the representative network and implement event-triggered replanning.

2. The method according to claim 1, characterized in that, Step 1 includes: acquiring the planning area boundary, building occupancy and height, population and road / vehicle exposure, important facilities, static and temporary no-fly zones, take-off and landing points, and protection zones under a unified coordinate reference, and recording the permitted flight altitude, aircraft speed and mass, maximum climb angle, maximum descent angle, airway envelope width, and safety separation; wherein, the take-off and landing points include the take-off and landing points of vertical take-off and landing manned aircraft, and the protection zone is a restricted airspace that unrelated UAVs are not allowed to intrude into; building and geographic data are formed from urban 3D geographic information or surveying data, population, road, and vehicle exposure are formed from population grids, road networks, and traffic status data, no-fly zones are formed from airspace management or operational management status, and take-off and landing points are formed from confirmed site data; Data from the same sampling time and under the same spatial coordinate reference are combined into an environmental snapshot with time and version identifiers; based on the environmental snapshot, inputs for ground personnel casualty risk assessment, noise assessment, and spatial occupancy indexes for buildings, no-fly zones, and take-off and landing point protection zones are generated; the spatial occupancy indexes are used to query buildings and restricted airspace near the flight segment envelope without changing the judgment criteria for the corresponding constraints; All complete network ants in the same planning round read the same environmental snapshot to ensure that candidate complete route networks are constructed and evaluated under the same environmental conditions; when subsequent monitoring data arrives, the environmental snapshot used to generate the currently published network or the most recent actual replanning is used as the benchmark for judging whether the environment has undergone substantial changes.

3. The method according to claim 2, characterized in that, Step 1 further includes: assuming the complete route network contains K origin-destination routes, these origin-destination routes are called OD routes, where OD is an abbreviation for Origin-Destination and represents a pair of origins to destinations; the kth origin-destination route contains B k Each network ant's decision variables simultaneously store the horizontal state and continuous altitude state of K origin-end routes; both beneficial ants and safe ants belong to the complete network ant family, using the same network coding and hard constraint verification, differing only in their dominant target, pheromone, heuristic information, and the file they enter; the horizontal state is accumulated segment by segment in the local coordinate system of each origin-end route's origin and destination using five horizontal increments: shifting two squares to the left, shifting one square to the left, maintaining, shifting one square to the right, and shifting two squares to the right; the pheromone address includes the origin-end route number, segment number, current actual horizontal-altitude state, and next actual horizontal-altitude state, to distinguish different actual segments corresponding to the same action number in different preceding states; Highly employing continuously normalized variables Encode the data and decode it into the actual physical height according to equation (1): (1), in This represents the continuously normalized altitude variable of the i-th control point on the k-th OD route. ∈[0,1]; The actual altitude of the i-th control point on the k-th OD route. and These are the lower and upper limits of the permitted flight altitude, respectively.

4. The method according to claim 3, characterized in that, Step 2 includes: In the joint construction round of the s-th segment, all OD routes that have not yet reached their destination form a legal next transfer set for the s-th segment based on the same initial frozen pheromone snapshot, the same environmental snapshot, and the actual state at the end of their respective s-1 segments; after each OD route completes its transfer selection in this round, it submits its selection and then enters the joint construction round of the s+1 segment. The legal next transfer set is directly screened from the planning area boundary, destination accessibility, static and temporary no-fly zones, non-associated take-off and landing point protection zones, and the envelope requirements of this OD route; among them, destination accessibility refers to the ability to reach a fixed destination within the remaining number of segments and the allowable lateral increment. For each transfer within the set of legitimate next transfers, the benefit ant forms its own heuristic information based on the incremental evaluation of propulsion energy consumption and 3D maneuver costs, and reads the benefit ant's pheromone; the safety ant forms its own heuristic information based on the incremental evaluation of ground personnel casualty risk and noise exposure, and reads the safety ant's pheromone; the opponent's heuristic correction is formed by the normalized incremental cost of the other party's corresponding target group in the same set of legitimate next transfers, and participates with a bounded factor with a preset upper limit; the 3D consensus pheromone is formed by the actual 3D overlapping segments with the same origin-destination route and the same segment number in the first non-dominated frontier of the benefit archive and the first non-dominated frontier of the safety archive in the previous iteration; then, the pseudo-random proportional transfer rule of the ant colony system is adopted: in the utilization branch, the transfer with the greatest comprehensive attractiveness is selected, and in the exploration branch, the transfer is selected from the remaining legitimate next transfers other than the transfer with the greatest comprehensive attractiveness according to the normalized attractiveness probability; two or more complete network ants can be constructed independently in parallel working process; after a single complete network ant is completed, all origin-destination routes and the spatial relationships between different origin-destination routes included in the complete network ant are evaluated at once; For each horizontal segment, a local forward axis is established based on the start and end points of each route. A local lateral axis orthogonal to this local forward axis is then established. Reference control points are set along the local forward axis according to the target segment length. The lateral offset of the internal control points is accumulated segment by segment using five increments {-2, -1, 0, 1, 2} and multiplied by the lateral offset unit. The starting and ending control points are fixed at the starting point S. k and endpoint E k Adjacent control points are connected to form a horizontal flight segment; based on the space occupancy index established by the environmental snapshot, only the buildings and restricted airspace units covered by the bounding box after the current flight segment envelope expansion are queried to calculate the minimum allowable altitude of each control point; the continuous altitude sequence is obtained by decoding the continuous normalized altitude variables of each control point according to formula (1); under the constraints of fixed origin and destination altitude, allowable flight altitude range, building height and vertical clearance, maximum climb angle and maximum descent angle, forward projection and reverse projection are alternately performed on the continuous altitude sequence, under the condition that the upper and lower limits of the altitude of each control point and the adjacent altitude difference limit are met. This minimizes the total adjustment of the projected altitude relative to the original decoded altitude; the range of consecutive segments where different origin-destination route envelopes simultaneously fail to meet both horizontal and vertical safety intervals constitutes a conflict zone; if there are two or more consecutive altitude adjustment schemes q that can eliminate the conflict, then the additional propulsion energy consumption Δf1(q) and the additional three-dimensional maneuver cost Δf2(q) are calculated separately, schemes whose increments Δf1(q) and Δf2(q) are simultaneously dominated by other schemes are deleted, and the two increments Δf1(q) and Δf2(q) of the remaining schemes are normalized according to the value range of this candidate set, and then selected. The minimum solution; among which This represents the normalized increase in propulsion energy consumption under scheme q. This represents the normalized increment of the cost of adding new three-dimensional maneuvers under scheme q. If the difference between the maximum normalized increments of the two schemes does not exceed the preset numerical precision, then calculate the total height change H(q) = Σ k Σ i |z' k,i (q)-z k,i |And select the scheme with smaller H(q) from the two schemes, where z' k,i (q) represents the altitude of the i-th control point on the k-th origin-destination route after the projection of scheme q, z k,i Before adjusting the altitude; if the difference in total altitude change does not exceed the preset altitude accuracy, select according to the pre-fixed scheme number; if the conflict cannot be eliminated within the allowable altitude and climb and descent constraints, mark the corresponding complete route network as infeasible; The projected complete route network is recalculated based on the actual three-dimensional route segments to include four target values: propulsion energy consumption, three-dimensional maneuvering cost, ground personnel casualty risk, and noise exposure. At the same time, positive residuals are calculated for altitude overruns, insufficient clearance in buildings or no-fly zones, intrusion into non-associated take-off and landing point protection zones, yaw angle overruns, climb or descent angle overruns, and insufficient intervals between different OD routes. The sum of the positive residuals is recorded as the constraint violation amount, and the target values ​​and constraint violation amounts are recorded independently.

5. The method according to claim 4, characterized in that, Step 3 includes: For a complete airway network that has passed full verification and whose constraint violation amount is no greater than the preset numerical tolerance, four minimization objectives are calculated: propulsion energy consumption, three-dimensional maneuvering cost, ground personnel casualty risk, and noise exposure. Propulsion energy consumption is calculated based on the changes in flight segment length and altitude, as well as the aircraft performance model. Three-dimensional maneuvering cost is calculated based on changes in horizontal turning and climb and descent. Ground personnel casualty risk is calculated based on the exposure of population, road vehicles, and important facilities below the airway and the accident consequences model. Noise exposure is calculated based on the aircraft flight status, propagation distance, and ground-sensitive receptors. The efficient population and the safe population are two pre-defined complete network ant colonies; the efficient population is non-dominated and sorted according to minimizing propulsion energy consumption and minimizing three-dimensional maneuver costs, while the safe population is non-dominated and sorted according to minimizing the risk of casualties to ground personnel and minimizing noise exposure. Benefits File A E Formed by a complete route network that has undergone full verification and is non-dominatedly sorted according to propulsion energy consumption and three-dimensional maneuver costs, Safety File A S A complete route network, fully validated and sorted according to ground personnel casualty risk and noise exposure, is formed. Benefit profiles and safety profiles are updated after each ant colony iteration. When a profile exceeds its capacity, it is sorted according to the following rules: networks with smaller non-dominated level numbers have higher priority; networks with larger congestion distances within the same non-dominated level have higher priority, and the sorted network priorities are retained until the profile capacity limit is reached. Boundary networks have infinite congestion distances and are prioritized; when congestion distances are the same within the same level, networks with smaller normalized distances to the corresponding target group's ideal point are prioritized. The benefit ideal point is determined by the minimum propulsion energy consumption and the minimum three-dimensional maneuver cost in the first non-dominated front of the current benefit profile. The safety ideal point is determined by the minimum ground personnel casualty risk and the minimum noise exposure in the first non-dominated front of the current safety profile. The normalized distance from a candidate network to its corresponding ideal point is calculated by normalizing the corresponding two targets within the range of the current first non-dominated front. Simultaneously, a negotiation candidate profile A is maintained. B And four target record files A4, A B A4 is used to store complete feasible networks that can access both sides' comparisons and representative network selections. It records the joint non-dominated relationships and overall trade-offs among the four minimization objectives: propulsion energy consumption, three-dimensional maneuver costs, ground personnel casualty risk, and noise exposure. It does not replace A4. E and A S Their respective sorting criteria; Pheromones are stored separately according to the actual state transition e and the target category, and the update relationship is as follows: (2), in, Let p represent the pheromone intensity of searcher p at the actual state transition e before the end of the g-th ant colony iteration; p∈{E,S}, where E represents the beneficiary and S represents the safe. This represents the pheromone evaporation rate of searcher p; This represents the current generation deposition amount for the actual state transition e in the g-th ant colony iteration; the actual three-dimensional consensus pheromone uses independent pheromone memory, intensity upper limit, and volatility; After each ant colony iteration, a batch local update is first performed on the actual state transitions visited in this generation, and then evaporation is executed. Finally, the current generation's elite ants and benefit archive A are processed. E The sparse, high-quality network in the middle is responsible for the deposition of pheromones by the benefit side, which is composed of the current generation of elites and security archives A. S The sparse, high-quality networks in the data are used for pheromone deposition by the safe side; among them, the current-generation elite of the benefit side is selected from the contemporary feasible complete route network after non-dominance ranking according to propulsion energy consumption and three-dimensional maneuver costs; the current-generation elite of the safe side is selected from the contemporary feasible complete route network after non-dominance ranking according to ground personnel casualty risk and noise exposure; the sparse, high-quality networks in the historical archive are selected in order of non-dominance level number from small to large and congestion distance at the same level from large to small; when a complete feasible network has not yet been formed, only local origin-destination routes that have passed the checks of flight altitude, building clearance, climb and descent angles, yaw angles, endpoint requirements, and different origin-destination route intervals are allowed to provide structural clues at a preset proportion lower than the deposition amount of the complete feasible network; infeasible complete route networks are not allowed to enter A. E A S A B Or A4.

6. The method according to claim 5, characterized in that, Step 4 includes: Actual 3D consensus detection only compares benefit profile A E First Non-Domination Frontier and Security Archive A S In the first non-dominated frontier, two actual three-dimensional centerline segments corresponding to the same origin-destination route and the same segment number are considered. When the directional difference between the two actual three-dimensional centerline segments is no greater than the directional tolerance, the horizontal distance is no greater than the route envelope tolerance, the altitude difference is no greater than the altitude tolerance, and they have a positive overlap length along the line, the overlapping portion is determined to be the actual three-dimensional consensus. The benefit parent network starts from benefit file A. E Selected from the first non-dominated frontier, within the same frontier, networks with larger congestion distances are preferred. If congestion distances are the same, networks with smaller normalized distances to the ideal benefit point determined by propulsion energy consumption and three-dimensional maneuver costs are selected. The safety parent network is selected from safety file A. S In the first non-dominated frontier, networks with larger crowding distances are selected first within the same frontier. When the crowding distances are the same, networks with smaller normalized distances to the safe ideal point determined by the risk of injury or death to ground personnel and noise exposure are selected. Segments that intersect at a plane, have the same action code but different actual locations, are located at different altitudes, or belong to different OD routes do not constitute a consensus; The actual three-dimensional overlap length and the benefit parent network in benefit file A E The cutting-edge quality and security parent network in Security Archive A S The frontier quality in the ant colony jointly determines the amount of consensus pheromone deposition; the consensus pheromone adopts an independent intensity upper limit and decays at an independent volatility rate in each ant colony iteration, and the accidental overlap of the high-quality frontier that does not continuously obtain the benefit archive and the safety archive decays with iteration; When the benefit file A E and security file A S After forming the first non-dominated frontier consisting of a complete feasible network, the endpoints of the same origin-destination route, the same segment number, directional difference, horizontal spacing and height difference are all within the tolerance and have positive overlap length, and adjacent and continuous actual three-dimensional consensus segments are merged into consensus intervals; continuous segments in the same origin-destination route other than consensus intervals constitute non-consensus intervals. From the benefit file A E The first non-dominated frontier selects the parent network of benefits and the security archive A. S The first non-dominated frontier selects a safety parent network to generate negotiated offspring: In the consensus interval, common segments where the selected benefit parent network and the selected safety parent network actually overlap in three dimensions at the same origin-destination route and segment number are retained. In the non-consensus interval, the benefit parent network or safety parent network is inherited according to the interval, and switching only occurs at compatible anchor points. The compatible anchor point is a corresponding control point with the same sequence number j in both the benefit parent network and the safety parent network, located at the boundary between the consensus interval and the non-consensus inheritance interval in the same origin-destination route. The horizontal distance between these corresponding control points is no greater than the route envelope tolerance, the altitude difference is no greater than the altitude tolerance, and the altitude difference between the corresponding control points in the benefit parent network and the safety parent network, as well as the direction of the adjacent segments after switching, are also considered. All altitude changes, after continuous altitude projection, satisfy constraints on yaw angle, maximum climb angle, maximum descent angle, and horizontal and vertical safety separation between different take-off and destination routes, including planned area boundaries, static and temporary no-fly zones, non-associated take-off and landing protection zones, building clearance, permissible flight altitude, origin and destination endpoints, and different origin and destination routes. Benefit ants or safety ants constructed independently based on their own pheromones and heuristic information are called independently constructed ants. Negotiated offspring and independently constructed ants are subject to the same continuous altitude projection, planned area boundaries, no-fly zones, take-off and landing protection zones, building clearance, flight altitude, maneuverability, and route separation checks, and the four minimization objectives of propulsion energy consumption, three-dimensional maneuvering cost, ground personnel casualty risk, and noise exposure are calculated respectively.

7. The method according to claim 6, characterized in that, Step 5 includes: after reaching the preset number of iterations, evaluating the budget or the convergence stopping condition, and completing the last update of the benefit file and safety file, first searching for benefit file A in the decision space of the complete network ant and the decoded actual three-dimensional path space. E First Non-Domination Frontier and Security Archive A S The first non-dominated front points to a common set of the same complete route network; if the common set is not empty, then the network with the smallest larger value of the benefit-side normalized loss and the safety-side normalized loss is selected from the common set; if the common set is empty, then the network that has passed complete verification and comes from A... E A S A B In the candidate complete route network of A4, the joint non-dominated hierarchy is obtained by sequentially performing non-dominated ranking based on four minimization objectives: propulsion energy consumption, three-dimensional maneuvering cost, ground personnel casualty risk, and noise exposure. The candidate network is then ranked within the A4 range. E and A S The non-dominated levels on the corresponding target groups are screened; the complete route networks whose normalized losses exceed the preset threshold or whose corresponding non-dominated levels are higher than the preset levels are excluded from the benefit target groups consisting of propulsion energy consumption and three-dimensional maneuver costs, or the safety target groups consisting of ground personnel casualty risks and noise exposure. The relative satisfaction of the remaining candidate files is calculated, and the balanced representative network is determined according to formula (3): X*=argmax(X∈P B )min{s E (X),s S (X)}(3), Among them, P B To form a negotiated candidate set after screening through complete feasibility, joint non-dominant hierarchy, and benefit and security archive hierarchy, s E (X) and s S (X) represents the normalized satisfaction of candidate network X with the benefit side and the normalized satisfaction with the safety side, respectively; X* represents the final selected balanced representative network; complete feasibility means that all network-level hard constraints are passed; joint non-dominated hierarchy refers to the hierarchy obtained by jointly performing non-dominated ranking according to the four minimization objectives of propulsion energy consumption, three-dimensional maneuver cost, ground personnel casualty risk and noise exposure; Equation (3) adopts the discrete Kalley-Smolotinski negotiation selection rule to select min{s E (X), s S The largest candidate network for (X)}; The initial planning uses a cold-start ant population and evaluation budget, and saves the balanced representative network selected in step 5 for review, simulation, or release as the released network, and saves the benefit file A generated from the initial planning. E Security File A S Negotiation Candidate File A B The four target record files A4, and the environmental snapshot used when generating the published network are used as the baseline environmental snapshot; during operation, new risk fields and no-fly zone statuses are periodically received, and the complete feasibility of the published network is first reviewed under the current environmental snapshot, and then it is determined whether replanning is triggered according to formula (4): trigger(t)=[ΔR(t)≥θ R ]∨[ΔZ(t)≥θ Z ]∨[V jw (t)>ε](4), Where t represents the current monitoring time; trigger(t) indicates whether replanning is triggered at time t; ΔR(t) represents the normalized change of the current risk field relative to the risk field in the baseline environmental snapshot; ΔZ(t) represents the geometric change of the current no-fly zone or airspace restrictions relative to the no-fly zone or airspace restrictions in the baseline environmental snapshot; θ R Indicates the threshold for triggering changes in the risk field; θ Z Indicates the threshold for triggering changes in no-fly zones or airspace; V jw (t) represents the number of constraint violations by the published network under the current environmental snapshot; ε represents the feasibility numerical tolerance.

8. The method according to claim 7, characterized in that, In step 5, when ΔR(t) ≥ θ R ΔZ(t)≥θ Z V jw Hot reprogramming is initiated when any of the conditions in (t)>ε are met; otherwise, the feasibility and target value of the existing network are only reviewed in the new environment. Hot reprogramming uses an independently configurable number of ants and an evaluation budget, with the most recently effective planning resulting in A. E A S A B The old file is formed by combining A4 and the network that is still feasible after being reviewed in the new environment. The network is retained as a seed, and the affected origin and destination routes or segments are disturbed and the reconstructed network is supplemented. The pheromone is reinitialized in this round. After completion, a new time stamp, network version and verification report are generated. The hot replanning is a replanning process relative to the cold start initial planning.

9. An electronic device, characterized in that, It includes a processor and a memory, the memory storing program code that, when executed by the processor, causes the processor to perform the steps of the method as described in any one of claims 1 to 8.

10. A storage medium, characterized in that, It stores a computer program or instructions that, when run on a computer, perform the steps of the method as described in any one of claims 1 to 8.

Citation Information

Patent Citations

  • Methodology for Designing Low-Altitude Public Flight Routes for Unmanned Aerial Vehicles

    CN109948834B

  • Remote planning method for unmanned aerial vehicle air safety corridor paths

    CN110597276B