A method of logistics management based on low-altitude flight

Abstract

The present invention discloses a low-altitude flight logistics management method. It proposes a multi-dimensional optimization solution for the problems of poor adaptability in complex scenarios, low resource utilization, and formation coordination. By constructing a dynamic-static collaborative impedance model, combining robust optimization to generate an anti-interference path, and reinforcing learning to predict a dynamic adjustment strategy, path adaptive optimization is achieved. A composite impedance model that couples energy consumption and task density is introduced to quantify the impact of regional load on energy consumption and optimize distribution efficiency in high-density areas. A spatiotemporal clustering algorithm is used to stratify tasks according to geographic proximity and time urgency. Through rapid response, standard, and flexible formation collaborative scheduling, urgent tasks are prioritized and regional coverage is optimized. Core and marginal areas are divided, and long-endurance drones are used to cover remote areas. High-load formations are used to deliver batches to the core area, dynamically balancing coverage and repeated delivery impedance. This effectively improves response speed, energy utilization, and global distribution efficiency in complex scenarios.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle (UAV) task allocation, and in particular to a low-altitude flight logistics management method. Background Art

[0002] The rapid growth of e-commerce has spurred a surge in demand for logistics and distribution, particularly in the "last mile" of delivery, which faces challenges such as inefficiency, traffic congestion, and rising labor costs. Against this backdrop, logistics drones, with their flexibility, rapid response, and ability to circumvent ground transportation restrictions, have become a cutting-edge technology for breaking through last-mile delivery bottlenecks. Globally, major companies have launched drone logistics pilots, demonstrating their potential in remote areas, emergency supplies, and time-sensitive applications.

[0003] The Chinese invention patent, publication number CN117634818B, provides a logistics drone task allocation method with priority classification. Based on the natural geographical environment, low-altitude airspace structure, and logistics drone performance, it extracts the three-dimensional operating space information of the logistics drone and obtains the logistics drone performance parameters. Combining the needs of logistics companies and customers, it obtains the logistics drone transportation type, flow, flow direction, and flow time data to construct a logistics drone operation scenario. Combining task priority with the fairness of logistics companies, it establishes a logistics drone service satisfaction index. With the goal of minimizing flight impedance and maximizing service satisfaction, it constructs a logistics drone task allocation model. The NSGA-II algorithm is used to solve the model and output the optimal logistics drone resource allocation plan.

[0004] However, most approaches rely on standalone optimization of individual aircraft, aiming to minimize flight impedance and maximize service satisfaction, and construct multi-objective optimization models. While these approaches can solve basic task allocation problems, they fail to consider adaptability to complex environmental scenarios, resource utilization, and formation coordination. Summary of the Invention

[0005] This application solves the problems of poor adaptability of logistics drones in complex environmental scenarios, low resource utilization and poor formation coordination in the existing technology by providing a low-altitude flight logistics management method.

[0006] The present application provides a low-altitude flight logistics management method, the method comprising:

[0007] S1. Dynamic task impedance and static planning impedance are associated with each other through a collaborative weight factor to form a dynamic-static collaborative impedance. A robust optimization model is used to generate multiple candidate static paths, each considering the maximum anti-interference capability. A prediction model based on reinforcement learning is used to predict the probability and cost of dynamic adjustments.

[0008] The dynamic-static synergistic impedance is a weighted combination of the static planning impedance and the dynamic task impedance, which is defined as:

[0009]

[0010] Where: α is the static planning weight α∈[0,1], reflecting the degree of dependence on the basic path; β is the dynamic adjustment penalty factor, β>0, the higher the dynamic impedance ratio, the greater the total collaborative impedance; T d is the dynamic task impedance; T s is the static planning impedance; T d / T s It represents the ratio of dynamic adjustment cost to static planning cost, which measures the robustness of static planning;

[0011] S2. Based on the dynamic-static synergistic impedance, energy management impedance and task density impedance are introduced to construct a composite impedance model. Task density-energy coupling modeling is performed. A composite impedance minimization decision combination is adopted to achieve dynamic self-adaptation-energy density synergistic optimization.

[0012] The composite impedance model is defined as:

[0013]

[0014] Where: T DA-EDCO represents the composite impedance; T e is the energy management impedance; T d is the task density impedance; γ is the dynamic-static collaborative impedance weight, γ∈[0,1], which inherits the balance between path robustness and flexibility of the original scheme; δ is the energy-density coupling weight, δ>0, reflecting the amplification effect of task density on energy consumption; T e ·T d / T s Indicates the energy consumption efficiency of high task density areas. The higher the density, the greater the energy consumption per unit static path. Two new types of impedances are introduced, which are related to T DA-EDCO The impedances together constitute the global optimization objective;

[0015] S3. Through UAV formation collaboration, combined with spatiotemporal coupled clustering and formation collaborative path planning, global task allocation optimization is achieved, improving emergency task response speed and regional coverage efficiency;

[0016] S4. By balancing the regional coverage impedance and repeated delivery impedance through grouping strategies, the regional coverage and repeated delivery efficiency are further optimized to achieve efficient delivery across the entire region.

[0017] Furthermore, the static planning impedance is:

[0018] T s =T fly+T tl +T e ,

[0019] Among them, T s is the static planning impedance, T fly is the flight time, T tl is the take-off and landing time, T e is the charging time;

[0020] The dynamic task impedance is:

[0021] T d =T w +T c +T e ,

[0022] Among them, T d is the dynamic task impedance, T w For weather, T c For congestion, T e For emergency tasks.

[0023] Furthermore, a robust optimization model is used to generate multiple candidate static paths by inputting geographic environment data, UAV performance parameters, and historical mission data. Each path considers the maximum anti-interference capability (such as reserved detour margin and charging node redundancy) to obtain a static path set. represents the nth static path, each path is accompanied by a robustness score.

[0024] Furthermore, based on the prediction model of reinforcement learning, the probability and cost of dynamic adjustment are predicted, and the adjustment strategy is selected according to the real-time environmental data, sudden task requests and the robustness score of the current static path: Strategy 1: Fine-tune the path, such as local obstacle avoidance, dynamic impedance T d Lower; Strategy 2: Switch to the backup static path, dynamic impedance T d Higher but avoid task interruption; output dynamically adjusted path P f and the corresponding T d , P f The optimal feasible path generated by the system through adjustment strategy. Make collaborative impedance minimization decisions with the following constraints: task completion rate ≥ 99%; service satisfaction A total ≥0.95; the objective function is:

[0025]

[0026] Furthermore, based on the original dynamic-static collaborative impedance, energy management impedance and task density impedance are introduced to construct a composite impedance model:

[0027]

[0028] Where: T DA-EDCO represents the composite impedance; T e is the energy management impedance; T d is the task density impedance; γ is the dynamic-static collaborative impedance weight, γ∈[0,1], which inherits the balance between path robustness and flexibility of the original scheme; δ is the energy-density coupling weight, δ>0, reflecting the amplification effect of task density on energy consumption; T e ·T d / T s It indicates the energy consumption efficiency in the high task density area. The higher the density, the greater the energy consumption per unit static path.

[0029] Furthermore, the task density is quantified: the regional levels are divided according to the density; the energy management impedance calculation formula is:

[0030]

[0031] Where: T e is the energy management impedance; Qu is the power consumption of the UAV; Q ful is the full power; u∈U represents a single drone u in the drone set U; D s represents the task density in area s, and the calculation formula is: N is the number of demand points, A is the area of ​​the region, η(D s ) is the density-energy consumption coefficient, low density η = 1, medium density η = 1.2, high density η = 1.5;

[0032]

[0033] Where: T d is the task density impedance, RT is the path congestion time, ST is the standard flight time, and s∈S represents a single region s in the region set S;

[0034] Formulate the composite impedance minimization decision, and the objective function is:

[0035]

[0036] Among them, minT DA-EDCO The value at which the composite impedance is minimized.

[0037] Furthermore, the formations are classified as follows: a rapid response formation composed of high-endurance, high-speed drones, dedicated to emergency time window tasks; a standard formation composed of ordinary drones, handling standard time window tasks; a flexible formation composed of rechargeable drones, responsible for loose time window tasks, and supporting dynamic task insertion; emergency tasks are assigned to the rapid response formation first to ensure the shortest path and least interference; standard tasks are assigned to the standard formation according to regional clustering results, and a single formation covers multiple demand points within the same cluster; the flexible formation is dynamically adjusted according to the real-time task density to fill high-load areas or replace low-battery drones.

[0038] Furthermore, based on the traditional K-means spatial clustering, the time window urgency is integrated and spatiotemporal coupling clustering is introduced:

[0039] CW=λ·SD+(1-λ)·WU,

[0040] Among them, CW is the spatiotemporal coupling clustering weight, λ is the space and time weight, which defaults to 0.7 and ranges from [0,1], SD is the spatial distance; WU is the time window urgency, which is defined as RT is the remaining time, ST is the total time window, and the larger the time window urgency value is, the more urgent it is.

[0041] Furthermore, long-flight drones will be set as coverage formations and equipped with large-capacity batteries. They will be responsible for delivery in remote and low-density areas, mainly covering scattered demand points and minimizing regional coverage impedance; high-load drones will be set as efficient formations to support batch task packaging and reduce repeated delivery. They will mainly serve high-density core areas and reduce repeated impedance through single multi-task delivery; the original formations will be retained: quick response formation, standard formation, and flexible formation.

[0042] Furthermore, based on the task density, the area is divided into: core area, edge area, and buffer area; matching rule: coverage formations are only assigned to the edge area, and efficient formations are only assigned to the core area to avoid cross-region task overlap;

[0043] The flexible formation monitors the task density in the buffer zone in real time and dynamically switches the formation type (such as from a standard formation to a covering formation).

[0044] Two new types of impedance are further introduced, DA-EDCO The impedances together constitute the global optimization objective;

[0045] min(γ1T cg +γ2T rp +γ3T DA-EDCO ),

[0046] Among them, γ1 is the repeated distribution impedance weight; γ2 is the repeated distribution impedance weight, and γ3 is the composite impedance weight. The sum of γ1γ2γ3 is equal to 1, which is dynamically adjusted according to the regional load.

[0047] Regional coverage impedance: measures the delivery efficiency in edge areas. It is positively correlated with distance and introduces a distance penalty coefficient:

[0048]

[0049] Among them, T cg is the regional coverage impedance; s∈EZ represents a single area s in the edge zone; WH is the direct straight-line distance between the demand point and the warehouse, in kilometers; η e is the edge area distance penalty coefficient, which is used to amplify the time cost of edge area delivery and give priority to long-flight UAVs, such as η e =1.2;

[0050]

[0051] Among them, T rp is the repeated delivery impedance; s∈CZ represents a single area s in the core area; AC represents the number of repeated deliveries; C rp Represents the cost coefficient of a single repeated delivery, such as C rp =0.5, used to quantify the resource consumption of repeated word delivery.

[0052] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0053] By providing a low-altitude flight logistics management method, the problems of poor adaptability of logistics drones in complex environmental scenarios, low resource utilization and poor formation coordination in the existing technology are solved. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 This is a flow chart of a low-altitude flight logistics management method in an embodiment of the present invention;

[0055] Figure 2 A flowchart for robust path generation in an embodiment of the present invention; DETAILED DESCRIPTION

[0056] To facilitate understanding of the present invention, the present application will be described more comprehensively below with reference to the relevant drawings; the drawings show preferred embodiments of the present invention, but the present invention can be implemented in many different forms and is not limited to the embodiments described herein; on the contrary, the purpose of providing these embodiments is to enable a more thorough and comprehensive understanding of the disclosed content of the present invention.

[0057] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this invention pertains; the terms used herein in the specification of the present invention are for the purpose of describing specific embodiments only and are not intended to limit the present invention; the term "and / or" used herein includes any and all combinations of one or more of the associated listed items.

[0058] Example 1: This application provides a low-altitude flight logistics management method, the method comprising:

[0059] When a logistics drone makes a delivery, it encounters a static planning impedance generated by a baseline cost calculated based on the preset route network, drone performance, and historical mission data. The definition formula is:

[0060] T s =T fly +T tl +T e ;

[0061] Among them, T fly is the flight time, T tl is the take-off and landing time, T e is the charging time; correspondingly, the additional cost caused by real-time environmental interference and sudden tasks is called dynamic task impedance; the definition formula is:

[0062] T d =T w +T c +T e ,

[0063] Among them, T w For weather, T c For congestion, T e For emergency tasks;

[0064] The weighted combination of static planning impedance and dynamic task impedance is called dynamic-static collaborative impedance, which aims to balance the efficiency of the basic path and the ability to adapt to the real-time environment. It is defined as:

[0065]

[0066] Where: α is the static planning weight α∈[0,1], reflecting the degree of dependence on the basic path; β is the dynamic adjustment penalty factor β>0, the higher the dynamic impedance ratio, the greater the total collaborative impedance; T d is the dynamic task impedance; T s is the static planning impedance; T d / T s It represents the ratio of dynamic adjustment cost to static planning cost, and measures the robustness of static planning.

[0067] Robust static planning pre-optimization, such as Figure 2 As shown: By inputting geographic environment data, UAV performance parameters, and historical mission data, multiple candidate static paths are generated. Each path considers the maximum anti-interference ability to obtain a static path set. Represents the nth static path, and each path is accompanied by a robustness score. Each path is accompanied by a robustness score.

[0068] Redundant paths are generated through Monte Carlo simulation, and detour margins and robustness scores are reserved to screen highly anti-interference paths, reducing the need for dynamic adjustments.

[0069] Based on the prediction model of reinforcement learning, the probability and cost of dynamic adjustment are predicted. According to the real-time environmental data, sudden task requests and the robustness score of the current static path, the adjustment strategy is selected: Strategy 1: Fine-tune the path, such as local obstacle avoidance, dynamic impedance T d Lower; Strategy 2: Switch to the backup static path, dynamic impedance T d Higher but avoids task interruption; outputs the dynamically adjusted path and the corresponding T d , P f It is the optimal feasible path generated by the system through adjustment strategy.

[0070] The RL model can predict the probability of path congestion in the next 30 minutes, give priority to local adjustment strategies, reduce the frequency of path switching, and reduce the dynamic adjustment cost T d , the task interruption rate is reduced and the emergency task response time is shortened.

[0071] Make a decision to minimize the collaborative impedance, with the following constraints: task completion rate ≥ 99%; service satisfaction A total ≥0.95; the objective function is:

[0072]

[0073] Improved NSGA-II algorithm, the drone path is encoded as an integer sequence; weights α and β are encoded as floating point numbers with a dynamic range of [0,1]×[0,10]; α and β are adjusted according to the real-time environmental risk level:

[0074] β=2·L,

[0075] Where L is the risk level, and its value is [0,10];

[0076] Map risk levels to α, β: Peak hours: α = 0.2, β = 16;

[0077] Low period (risk level 2): ​​α=0.8, β=4.

[0078] Through dynamic optimization of collaborative weights, the dynamic adjustment ratio and task punctuality rate during peak periods can be improved; the static path utilization rate during off-peak periods can be improved and energy consumption can be reduced.

[0079] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages:

[0080] This application achieves a triple breakthrough in efficiency, robustness, and satisfaction for logistics drone task allocation through a dynamic-static impedance coupling model, robust path pre-optimization, and data-driven weight adjustment, while simultaneously improving service satisfaction and task reliability. This allows for a seamless transition from static planning to dynamic response. Static planning pre-optimization reduces the need for dynamic adjustments, while dynamic prediction and strategy selection improve response efficiency. Multiple strategies collaborate to address sudden disruptions, achieving more efficient and stable logistics drone task allocation while improving satisfaction.

[0081] Existing technologies rely solely on static route optimization and fail to account for real-time dynamic disruptions, such as weather changes and unexpected tasks, which can cause order delays. For example, in a city center, daily delivery volume is stable, perhaps 200 orders per day, with clear skies and clear skies. However, temporary traffic control during the morning rush hour in a residential area can cause order delays. The aforementioned technical solution directly invokes the pre-generated robust static route, detects the restricted area through the RL model, and triggers Strategy 1, reducing the sudden detour distance by 68% and the average detour time from 15 minutes to 4 minutes. By dynamically adjusting the weights α = 0.6 and β = 8, the probability of task interruption is reduced by 73%.

[0082] Example 2: While the above example optimizes the synergy between dynamic routing and static planning, it does not fully consider the amplifying effect of task space density on energy efficiency. For example, when delivery points are highly concentrated in the same area, the frequent takeoffs and landings, obstacle avoidance, and charging waits of drones can lead to a nonlinear increase in energy consumption, thus affecting overall delivery efficiency. This example further improves on the above example.

[0083] Based on the original dynamic-static collaborative impedance, energy management impedance and task density impedance are introduced to construct a composite impedance model:

[0084]

[0085] Where: T DA-EDCO represents the composite impedance; T e is the energy management impedance; T d is the task density impedance; γ is the dynamic-static collaborative impedance weight, γ∈[0,1], which inherits the balance between path robustness and flexibility of the original scheme; δ is the energy-density coupling weight, δ>0, reflecting the amplification effect of task density on energy consumption; T e ·T d / T s Indicates the energy consumption efficiency in high task density areas (the higher the density, the greater the energy consumption per unit static path).

[0086] Identify high-density grids and automatically trigger the following strategies: assign fully charged drones or pre-placed mobile charging stations; adjust flight altitude to the low turbulence layer to reduce headwind energy consumption; reduce unit mission energy consumption in high-density areas and increase average daily flight time.

[0087] The dynamic-static collaborative impedance is optimized, using a static pre-generated robust path set and real-time environmental data (weather, airspace status) as input data. Based on reinforcement learning, high-probability interference events (such as sudden congestion) are predicted, and local detours or switching to alternative paths are selected as dynamic path adjustments. The γ value is dynamically adjusted according to the time period (peak / valley). For example, during peak periods, γ is reduced to focus on dynamic adjustment, while during vale periods, γ is increased to prioritize static path stability as weighted adaptability, and the output dynamic path P is optimized. f And the corresponding T DS value.

[0088] The composite impedance model is defined as:

[0089]

[0090] Where: T DA-EDCO represents the composite impedance; T e is the energy management impedance; T d is the task density impedance; γ is the dynamic-static collaborative impedance weight, γ∈[0,1], which inherits the balance between path robustness and flexibility of the original scheme; δ is the energy-density coupling weight, δ>0, reflecting the amplification effect of task density on energy consumption; T e ·T d / T s Indicates the energy consumption efficiency in high task density areas. The higher the density, the greater the unit static path energy consumption.

[0091]

[0092] Where: T e is the energy management impedance, Qu is the power consumption of the UAV, Q ful is the full power; u∈U represents a single drone u in the drone set U; D s represents the task density in area s, and the calculation formula is: N is the number of demand points, A is the area of ​​the region, η(D s ) is the density-energy consumption coefficient, low density η = 1, medium density η = 1.2, high density η = 1.5;

[0093]

[0094] Where: T d is the task density impedance, RT is the path congestion time, ST is the standard flight time, and s∈S represents a single region s in the region set S;

[0095]

[0096] Among them, minT DA-EDCO is the value that minimizes the composite impedance.

[0097] Constraints were also added, including regional load balancing: a single drone could only perform a maximum of three missions in high-density areas to avoid concentrated power consumption; an energy security threshold: a drone's remaining battery power must not fall below 20% to prevent emergency landings. Through collaborative optimization, a "low-energy path + balanced load" combination was prioritized; in low-density areas, straight routes were used to maximize static route efficiency; in high-density areas, a serpentine patrol mode was enabled to complete multi-point deliveries in batches, reducing composite impedance.

[0098] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages:

[0099] This application introduces the energy management-task density coupling relationship. The DA-EDCO solution further optimizes the energy consumption and task allocation balance in high-density areas based on dynamic adaptive path planning. Its core lies in: accurately quantifying the impact of density on energy consumption: by η(D s ) coefficient, dynamically adjusting energy allocation strategies; load balancing and composite impedance minimization. For example, during the "Double Eleven" shopping festival, order volume surged, creating intensive delivery demands in core business districts. Traditional solutions faced a range crisis, with drones running out of battery and forced to land due to frequent takeoffs and landings. The aforementioned technical solution activates the composite impedance model, identifies high-density areas, automatically deploys mobile charging stations, adjusts flight altitude to the low turbulence layer, activates a serpentine patrol mode, and completes multi-point deliveries in batches. This density-aware energy allocation increased single-machine mission capacity in high-density areas by 40% and extended flight time by 2.3 hours. This reduces composite impedance, increases mission capacity, and enables more efficient and stable logistics drone collaborative delivery.

[0100] Example 3: While the above examples optimize energy consumption and task allocation, in large-scale deployment scenarios for logistics drones, the single-machine task allocation model faces bottlenecks such as delayed emergency task response, low regional coverage efficiency, and poor adaptability to dynamic environments. This example further improves upon the above examples.

[0101] The formations are classified as follows: a rapid response formation composed of high-endurance, high-speed drones is dedicated to emergency time window tasks; a standard formation composed of ordinary drones handles standard time window tasks; a flexible formation composed of rechargeable drones is responsible for loose time window tasks and supports dynamic task insertion; emergency tasks are assigned to the rapid response formation first to ensure the shortest path and least interference; standard tasks are assigned to the standard formation according to regional clustering results, and a single formation covers multiple demand points within the same cluster; the flexible formation is dynamically adjusted according to the real-time task density to fill high-load areas or replace low-battery drones.

[0102] Based on the traditional K-means spatial clustering, the time window urgency is integrated and spatiotemporal coupling clustering is introduced:

[0103] CW=λ·SD+(1-λ)·WU,

[0104] Among them, CW is the spatiotemporal coupling clustering weight, λ is the space and time weight, which defaults to 0.7 and ranges from [0,1], SD is the spatial distance; WU is the time window urgency, which is defined as RT is the remaining time, ST is the total time window, and the larger the time window urgency value is, the more urgent it is.

[0105] Different cluster matching strategies are adopted for different formations; Quick response formation: high-urgency cluster, time window urgency ≥ 0.8, using a straight path to reach the target; Standard formation: medium-low urgency cluster, time window urgency < 0.8, serving all points in the cluster in closed path order; Flexible formation: dynamically fill uncovered clusters or take over overflow tasks.

[0106] Formation collaborative path planning: The ant colony algorithm is used to generate the shortest closed path for each formation, covering all demand points in the cluster. The objective function is:

[0107] min∑(T fly +T tl )+μ·max(T tw ),

[0108] Among them, T fly is the flight time; T tl is the take-off and landing time; μ is the time window delay penalty coefficient, which is used to amplify the cost of emergency task overtime and ensure delivery punctuality, such as μ = 2; T tw The maximum time window delay value of all tasks in the formation, in minutes, is calculated as follows:

[0109] T tw =max(TT-TM, 0),

[0110] Among them, TT is the actual arrival time, TM is the upper limit of the time window; if the time window of a task is [09:00,10:00] and the actual arrival time is 10:15, then the delay is T tw The delay time is set to 15 minutes, with 0 representing the lower limit of the delay time. This means that if the actual arrival time does not exceed the upper limit of the time window, the delay time is forcibly set to 0. Inter-formation coordination rules are established: Rapid response formations enjoy airspace priority, and standard formations must detour their paths by at least 500 meters. Flexible formation drones can provide emergency charging support to other formations, extending mission cycles through wireless charging modules. Based on real-time location sharing, formation altitudes are automatically adjusted, with rapid formations flying at 100-150 meters and standard formations at 50-100 meters.

[0111] Dynamic task reallocation and formation reconstruction; triggering the insertion of emergency tasks when the following situations occur: such as temporary expedited orders; equipment failure: low drone battery or communication interruption; environmental interference: severe weather leading to flight bans in some areas. When encountering the above situations, the following reallocation strategies are adopted: Emergency task takeover: The rapid response formation is given priority. If it is fully loaded, the elastic formation expansion is triggered, temporarily adding 2-3 drones; Standard task delay processing: After re-clustering, it is assigned to the elastic formation and its time window weight is reduced to below 0.5; Formation reconstruction algorithm: A dynamic adjustment model based on deep reinforcement learning is used to minimize the global impedance T total For the goal, optimize the formation structure and path in real time.

[0112] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages:

[0113] This application improves emergency task response capabilities and resource utilization by introducing formation collaborative priority clustering optimization; optimizes the spatiotemporal matching of tasks within the formation to reduce conflicts; and reduces global impedance through priority passage and energy sharing. For example, if materials need to be delivered to 8 locations, the time window is compressed to 45 minutes. At the same time, encountering thunderstorms, 3 main routes are closed. The above technical solution is adopted for formation collaborative emergency response, and the rapid response formation breaks through in a straight line. Spatiotemporal clustering merges 8 points into 2 emergency clusters. At the same time, the flexible formation provides air charging support, extending the mission radius by 35%. Spatiotemporal coupling clustering increases the emergency task response speed by 112%, and formation collaboration avoids 62% of route conflicts. Ultimately, the formation solution significantly improves response speed, task volume and stability, and more efficient and stable logistics drone collaborative delivery.

[0114] Example 4: In the above embodiment, although the response speed, task volume and stability are improved, the regional coverage is uneven. In remote areas, due to the low task density and long distance, the repeated round trips of drones result in high energy consumption and poor timeliness. In high-density areas, due to the dispersion of orders, a single drone needs to travel back and forth to the same area multiple times, resulting in path redundancy. In order to solve the coverage and repeated delivery problems, this embodiment further improves the above embodiment.

[0115] Long-flight drones will be set as coverage formations and equipped with large-capacity batteries. They will be responsible for delivery in remote, low-density areas, mainly covering scattered demand points and minimizing regional coverage impedance; high-load drones will be set as efficient formations to support batch task packaging and reduce repeated delivery. They will mainly serve high-density core areas and reduce repeated impedance through single multi-task delivery; the original formations will be retained: quick response formation, standard formation, and flexible formation.

[0116] Based on task density, the region is divided into core area, edge area, and buffer area. Matching rules: Coverage formations are only assigned to edge areas, and efficient formations are only assigned to core areas to avoid cross-regional task overlap.

[0117] The flexible formation monitors the task density in the buffer zone in real time and dynamically switches the formation type (such as from a standard formation to a covering formation).

[0118] Two new types of impedance are introduced, DA-EDCO The impedances together constitute the global optimization objective;

[0119] min(γ1T cg +γ2T rp +γ3T DA-EDCO ),

[0120] Among them, γ1 is the repeated distribution impedance weight; γ2 is the repeated distribution impedance weight, and γ3 is the composite impedance weight. The sum of γ1γ2γ3 is equal to 1, which is dynamically adjusted according to the regional load.

[0121]

[0122] Among them, T cg is the regional coverage impedance; s∈EZ represents a single area s in the edge zone; WH is the direct straight-line distance between the demand point and the warehouse, in kilometers; η e is the edge area distance penalty coefficient, such as η e =1.2;

[0123]

[0124] Among them, T rp is the repeated delivery impedance; s∈CZ represents a single area s in the core area; AC represents the number of repeated deliveries; C rpRepresents the cost coefficient of a single repeated delivery, such as C rp =0.5, used to quantify the resource consumption of repeated word delivery.

[0125] Formations are matched according to regional clustering. The core area generates high-density task clusters based on the K-means++ algorithm, and each cluster is served by one efficient formation. The edge area is divided into grids according to administrative boundaries or natural terrain, and each grid is assigned one coverage formation. The buffer zone monitors task density in real time and triggers flexible formation mode switching, such as switching from a standard formation to a coverage formation.

[0126] Among them, the coverage formation adopts a "radiating straight line path" to go directly from the warehouse to the edge demand point, and can carry sporadic orders on the return trip; the efficient formation generates a "closed ring path" based on the improved ant colony algorithm to cover all demand points in the cluster; the flexible formation enables the "dynamic patrol mode" in the buffer zone to scan for unassigned tasks along the main road.

[0127] When the core area is overloaded, the cluster is automatically split and the elastic formation support is called to implement overload warning; when the coverage formation task volume exceeds the threshold, the elastic formation switches to coverage mode to take over the excess tasks and reinforce the edge area.

[0128] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages:

[0129] This application solves the problems of duplicate delivery in the core area and coverage in the edge area simultaneously through fleet grouping and regional exclusive service strategies, reducing the overall impedance; improving fleet utilization, and reducing redundant investment in the elastic resource pool; dynamically balancing coverage, duplication and the original trade-off goals; for example, there are long-term delivery pain points in remote mountainous areas: 5 villages are scattered 80km 2 In mountainous areas, there are only 8 orders per day, but the traditional solution requires drones to travel back and forth every day, which is very costly. The above technical solution is adopted to deploy a dedicated fleet, covering the fleet's radial straight line delivery, and shortening the 80km 2 Divide into 3 exclusive grids and enable distance penalty coefficient η e =1.2 Optimized routes increased edge coverage efficiency by 217%, shortening delivery time in mountainous areas from 48 hours to 8 hours. This achieved global optimization, resulting in more efficient and stable logistics drone-assisted delivery.

Claims

1. A low-altitude flight logistics management method, characterized in that: The following steps are involved: S1. Dynamic task impedance and static planning impedance are associated with each other through a collaborative weight factor to form a dynamic-static collaborative impedance. A robust optimization model is used to generate multiple candidate static paths, each considering the maximum anti-interference capability. A prediction model based on reinforcement learning is used to predict the probability and cost of dynamic adjustments. The dynamic-static synergistic impedance is a weighted combination of the static planning impedance and the dynamic task impedance, which is defined as: ； in: For static planning weights , reflecting the degree of dependence on the basic path; To dynamically adjust the penalty factor, , the higher the proportion of dynamic impedance, the greater the total cooperative impedance; is the dynamic task impedance; is the static planning impedance; It represents the ratio of dynamic adjustment cost to static planning cost, which measures the robustness of static planning; The static planning impedance is: ， in, is the static planning impedance, is the flight time, is the take-off and landing time, is the charging time; The dynamic task impedance is: ； in, is the dynamic task impedance, Additional costs caused by weather, For the additional costs caused by congestion, Additional costs incurred for unexpected tasks; S2. Based on the dynamic-static synergistic impedance, energy management impedance and task density impedance are introduced to construct a composite impedance model; task energy-density coupling modeling is performed; and composite impedance minimization decisions are made to improve delivery efficiency. The composite impedance model is defined as: ; in: represents the composite impedance; Impedance for energy management; is the task density impedance; is the dynamic-static cooperative impedance weight, ,Inherit the original solution’s balance between path robustness and flexibility; is the energy-density coupling weight, , reflecting the amplifying effect of task density on energy consumption; Indicates the energy consumption efficiency of high task density areas. The higher the density, the greater the unit static path energy consumption. The introduction of regional coverage impedance and repeated delivery impedance is related to The impedances together constitute the global optimization objective; The energy management impedance is: ， in: Impedance for energy management, Power consumption for drones, Fully charged capacity; Represents a single drone u in the drone set U; represents the task density in area s, and the calculation formula is: , N is the number of demand points, A is the area of ​​the region, is the density-energy consumption coefficient, low density =1, medium density =1.2, high density =1.5; The task density impedance is: ， in: is the task density impedance, RT is the path congestion time, ST is the standard flight time, Represents a single region s in the region set S; The regional coverage impedance measures the delivery efficiency of the edge area, is positively correlated with the distance, and introduces a distance penalty coefficient; The repeat delivery impedance is defined as: ， in, For repeated distribution impedance; represents a single area s within the core area; AC represents the number of repeated deliveries; Represents the cost coefficient of a single repeated delivery, which is used to quantify the resource consumption of a single repeated delivery; S3. Through UAV formation collaboration, combined with spatiotemporal coupled clustering and formation collaborative path planning, global task allocation optimization is achieved, improving emergency task response speed and regional coverage efficiency; The spatiotemporal coupled clustering is defined as: Among them, CW is the spatiotemporal coupling clustering weight, is the spatial and temporal weight, SD is the spatial distance; WU is the time window urgency, which is defined as ,RT is the remaining time, ST is the total time window, and the larger the time window urgency value is, the more urgent it is; S4. By balancing the regional coverage impedance and repeated delivery impedance through grouping strategies, the regional coverage and repeated delivery efficiency are further optimized to achieve efficient delivery across the entire region.

2. The low-altitude flight logistics management method according to claim 1, characterized in that: The robust optimization model is defined as: By inputting geographic environment data, UAV performance parameters, and historical mission data, multiple candidate static paths are generated. Each path considers the maximum anti-interference capability to obtain a static path set. , represents the nth static path, with a robustness score attached to each path.

3. The low-altitude flight logistics management method according to claim 1, characterized in that: The composite impedance minimization decision is: ， in, The value at which the composite impedance is minimized.

4. The low-altitude flight logistics management method according to claim 1, characterized in that: When the formation collaborative path planning generates the optimal path, the objective function is: ， in, is the flight time; Takes time for takeoff and landing; is the time window delay penalty coefficient, which is used to amplify the cost of urgent task overtime; The maximum time window delay value of all tasks in the formation, in minutes, is calculated as follows: ， Where TT is the actual arrival time, TM is the upper limit of the time window, and 0 represents the lower limit of the delay time. That is, when the actual arrival time does not exceed the upper limit of the time window, the delay time is forcibly set to 0.

5. The low-altitude flight logistics management method according to claim 1, characterized in that: The formations include: rapid response formation, standard formation and flexible formation; The rapid response formation refers to a formation composed of high-endurance, high-speed drones, which are dedicated to emergency tasks within 1 hour. Among them, high endurance refers to a full-load flight distance of ≥ 50 kilometers after a single charge, and high speed refers to a cruising speed of ≥ 90 km / h and a maximum speed of ≥ 120 km / h; the standard formation refers to a formation that can complete a standard task within 3 hours and is allocated according to the results of spatiotemporal clustering; the flexible formation refers to a formation composed of rechargeable drones, which dynamically fill high-load areas and take over overflow tasks.

6. The low-altitude flight logistics management method according to claim 1, characterized in that: The global optimization objectives include: ， in, is the repeated distribution impedance weight; is the repeated distribution impedance weight, is the composite impedance weight, The sum is equal to 1, and is adjusted dynamically according to the regional load.

7. The low-altitude flight logistics management method according to claim 1, characterized in that: When the core area is overloaded with tasks, the elastic formation is triggered to expand or split the cluster; when the task volume exceeds the threshold, the elastic formation switches to overlay mode to take over the tasks.

8. The low-altitude flight logistics management method according to claim 1, characterized in that: The robustness score of the candidate static path is calculated based on the detour margin, charging node redundancy and historical anti-interference data. The dynamic adjustment strategy includes local path fine-tuning or switching to an alternative path.

9. The low-altitude flight logistics management method according to claim 1, characterized in that: Formation coordination includes: the rapid response formation enjoys airspace priority, and the standard formation must detour its path by at least 500 meters; the flexible formation provides cross-formation emergency charging support.

Images