A hierarchical three-dimensional UAV logistics trajectory planning method and system

By constructing a hierarchical three-dimensional directed graph and a three-dimensional communication risk graph, a trajectory attribute matrix between task nodes corresponding to multiple discrete load levels is generated. This solves the problems of energy consumption assessment deviation and insufficient communication reliability in the existing UAV logistics trajectory planning, and realizes efficient and safe logistics delivery in low-altitude logistics scenarios.

CN123111301APending Publication Date: 2026-11-17NANCHANG UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611615313.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-19
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing drone logistics trajectory planning methods fail to effectively coordinate the three-dimensional trajectory planning of multiple pick-up and delivery nodes, resulting in biased energy consumption assessments and insufficient communication reliability, making them difficult to implement in practical engineering.

Method used

A hierarchical 3D UAV logistics trajectory planning method is adopted. By acquiring logistics task information, a hierarchical 3D directed graph and a 3D communication risk graph are constructed. A trajectory attribute matrix between task nodes corresponding to multiple discrete load levels is generated, and feasibility conditions are verified and corrected to generate a 3D logistics trajectory that meets the constraints.

Benefits of technology

It ensures communication reliability and flight safety in low-altitude logistics scenarios, significantly improves logistics and distribution efficiency, and reduces energy costs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123111301A_ABST
    Figure CN123111301A_ABST
Patent Text Reader

Abstract

This invention relates to a hierarchical three-dimensional UAV logistics trajectory planning method and system. The method includes: acquiring logistics task information of a task area; discretizing the task area to construct a hierarchical three-dimensional directed graph; constructing a three-dimensional communication risk graph based on the communication risk values ​​of different spatial location units; performing a three-dimensional path search in the hierarchical three-dimensional directed graph to obtain candidate three-dimensional trajectories between task nodes; generating trajectory attribute matrices corresponding to multiple discrete load levels based on the three-dimensional communication risk graph; generating a set of candidate task access sequences based on the trajectory attribute matrices under the condition of satisfying task execution constraints; generating candidate three-dimensional logistics trajectories based on the candidate task access sequences; and performing feasibility condition verification and correction operations on each flight segment of the candidate three-dimensional logistics trajectories to obtain the final three-dimensional logistics trajectory. This invention can significantly improve logistics delivery efficiency and reduce energy consumption costs while ensuring communication reliability and flight safety.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) logistics technology, specifically to a hierarchical three-dimensional UAV logistics trajectory planning method and system. Background Technology

[0002] With the development of building delivery, industrial park delivery, campus logistics, factory material transfer, and urban low-altitude instant logistics, drones are gradually transforming from simple point-to-point flight tools into aerial logistics execution platforms oriented towards multiple orders, multiple nodes, and multiple constraints. In a single logistics mission, a drone often needs to sequentially visit multiple pickup nodes and multiple delivery nodes. Each item typically corresponds to one pickup node and one delivery node, and the delivery node should be visited after its corresponding pickup node. Therefore, drone logistics trajectory planning involves not only spatial flight paths but also the access order between multiple sets of pickup and delivery nodes, cargo loading status, and mission fulfillment relationships.

[0003] Unlike ground vehicle delivery or fixed-altitude flight, low-altitude drone logistics missions have distinct three-dimensional spatial attributes. Pickup nodes may be located on warehouse rooftops, centralized sorting platforms, rooftop take-off and landing platforms, or at the same pickup altitude, while delivery nodes may be distributed across different floors, rooftops, balconies, platforms, temporary landing points, or hovering delivery locations. When drones move between mission nodes, they need to perform necessary climb, descent, and altitude-level switching maneuvers beyond horizontal displacement. Their flight costs are influenced by factors such as horizontal distance, delivery altitude, cumulative climb altitude, cumulative descent altitude, airspace obstacles, no-fly zones, and available battery power.

[0004] In existing technologies, common UAV logistics trajectory planning methods typically simplify the cost between task nodes to planar distance or two-dimensional shortest path cost, and then optimize the access order based on this. Some methods introduce three-dimensional Euclidean distance or a unified altitude penalty term to express the vertical motion cost, while a few methods use communication reliability as a constraint condition for path evaluation.

[0005] However, existing technologies have significant shortcomings in handling three-dimensional trajectory planning for multiple sets of pick-up and drop-off nodes: the two-dimensional distance matrix method cannot express the height differences and vertical maneuver costs between task nodes; existing three-dimensional planning methods fail to effectively coordinate the relationship between access sequence optimization and energy consumption and communication constraints in three-dimensional space, resulting in the generated trajectory being difficult to execute in engineering practice due to energy consumption assessment bias or insufficient communication reliability. Summary of the Invention

[0006] To address the aforementioned problems in the existing technology, this invention provides a hierarchical three-dimensional UAV logistics trajectory planning method and system. The technical problem to be solved by this invention is achieved through the following technical solution: In a first aspect, the present invention provides a hierarchical three-dimensional unmanned aerial vehicle (UAV) logistics trajectory planning method, comprising: Obtain logistics task information for the task area. The logistics task information includes task node information and environmental constraint information. The task node information includes a set of task nodes consisting of a take-off and landing point, several pickup nodes, and several delivery nodes. The task region is discretized in the height direction to construct a hierarchical three-dimensional directed graph; A three-dimensional communication risk map is constructed based on the communication risk values ​​of different spatial location units in the task area; A three-dimensional path search is performed in the layered three-dimensional directed graph to obtain candidate three-dimensional trajectories between task nodes, so as to generate a trajectory attribute matrix between task nodes corresponding to multiple discrete load levels based on the three-dimensional communication risk graph. Based on the trajectory attribute matrix between task nodes corresponding to multiple discrete load levels, a set of candidate task access sequences is generated under the condition of satisfying task execution constraints. Candidate 3D logistics trajectories are generated based on the candidate task access sequence set, and feasibility condition verification and correction operations are performed on each segment of the candidate 3D logistics trajectory to obtain a 3D logistics trajectory that meets the feasibility conditions.

[0007] Secondly, the present invention provides a hierarchical three-dimensional unmanned aerial vehicle (UAV) logistics trajectory planning system, comprising: The data acquisition module is used to acquire logistics task information in the task area. The logistics task information includes task node information and environmental constraint information. The task node information includes a set of task nodes consisting of a take-off and landing point, several pickup nodes, and several delivery nodes. A hierarchical directed graph construction module is used to discretize the task region in the height direction and construct a hierarchical three-dimensional directed graph; The communication risk modeling module is used to construct a three-dimensional communication risk map based on the communication risk values ​​of different spatial location units in the task area. The trajectory matrix generation module is used to perform a three-dimensional path search in the layered three-dimensional directed graph to obtain candidate three-dimensional trajectories between task nodes, so as to generate a trajectory attribute matrix between task nodes corresponding to multiple discrete load levels based on the three-dimensional communication risk graph. The access sequence optimization module is used to generate a set of candidate task access sequences based on the trajectory attribute matrix between task nodes corresponding to multiple discrete load levels, under the condition of satisfying task execution constraints. The verification and correction module is used to generate candidate three-dimensional logistics trajectories based on the candidate task access sequence in the candidate task access sequence set, and to perform feasibility condition verification and correction operations on the candidate three-dimensional logistics trajectories segment by segment to obtain three-dimensional logistics trajectories that meet the feasibility conditions.

[0008] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention provides a hierarchical 3D UAV logistics trajectory planning method. This method acquires logistics task information and constructs a hierarchical 3D directed graph and a 3D communication risk graph, achieving digital modeling of the low-altitude logistics UAV flight environment. Based on this, a trajectory attribute matrix is ​​generated through 3D path search, corresponding to task nodes with multiple discrete payload levels. This provides a readily available data foundation for subsequent sequence optimization, avoiding the computational overhead of repeated pathfinding. Furthermore, based on the trajectory attribute matrix, a set of candidate task access sequences is generated under the condition of satisfying task execution constraints. Feasibility conditions are verified and corrected segment by segment for the candidate 3D logistics trajectories generated from the candidate task access sequences, ensuring that the final output task access sequences are practically executable in terms of battery power consumption, flight speed, climb and descent rates, flight altitude, and no-fly zone avoidance. Compared with existing technologies, this invention fully considers the dynamic changes in the communication environment, the coupling of UAV payload and energy consumption, and the complex characteristics of interwoven multi-layered spatial constraints in low-altitude logistics scenarios. It can significantly improve logistics delivery efficiency and reduce energy costs while ensuring communication reliability and flight safety.

[0009] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0010] Figure 1 This is a flowchart illustrating a hierarchical three-dimensional UAV logistics trajectory planning method provided by the present invention; Figure 2 This is a schematic diagram of a hierarchical three-dimensional UAV logistics trajectory planning system provided by the present invention. Detailed Implementation

[0011] The present invention will be further described in detail below with reference to specific embodiments, but the implementation of the present invention is not limited thereto.

[0012] Example 1

[0013] In pickup and delivery logistics scenarios, the drone's payload dynamically changes depending on the order of visits. After picking up the goods, the drone is in a loaded state. If the subsequent delivery point is located at a higher altitude, the climb segment needs to overcome the additional power requirements brought by the increased altitude and cargo weight, typically consuming more energy than a level flight segment of the same horizontal distance. After the drone completes the delivery, its payload decreases, and the energy consumption characteristics of the subsequent descent, return, or re-pickup segments change again. Therefore, the mission visit sequence changes the location, number, and cumulative altitude of heavy-load climbs, thus affecting the endurance margin, mission duration, and executability of the entire three-dimensional logistics trajectory.

[0014] Low-altitude flight also relies on wireless communication links to support the issuance of control commands, flight status feedback, positioning information synchronization, mission progress monitoring, anomaly alarms, and manual takeover when necessary. If the quality of the wireless link in the low-altitude environment is insufficient, problems such as increased control command latency, telemetry data packet loss, discontinuous video or status feedback, link interruption, frequent base station switching, and expanded mission monitoring blind spots may occur. In severe cases, it may also affect the UAV's flight path tracking, obstacle avoidance decisions, return-to-home strategies, and fault handling, increasing the risk of mission interruption, replanning, return-to-home, or safety issues.

[0015] In urban buildings, park roads, campus complexes, and complex terrain environments, factors such as building obstruction, terrain undulations, base station deployment locations, antenna downtilt angles, low-altitude multipath propagation, and edge coverage can cause significant spatial non-uniformity in wireless communication quality. The same horizontal location may exhibit different received signal strength, signal-to-noise ratio, signal-to-interference-plus-noise ratio, packet error rate, link availability, or communication outage probability at different altitudes. Therefore, 3D logistics trajectory planning, while considering pickup and delivery relationships and load energy consumption, also needs to pay attention to the differences in communication reliability at different altitudes and on different flight segments.

[0016] Against this backdrop, for multiple pickup and delivery tasks, incorporating the three-dimensional altitude attributes of task nodes, pickup and delivery sequence, load changes, horizontal flight energy consumption, load climb energy consumption, descent maneuver energy consumption, and low-altitude communication link reliability into a unified hierarchical three-dimensional logistics trajectory planning framework has become an important research and application direction for improving the feasibility, safety, and fulfillment efficiency of UAV logistics tasks.

[0017] For the reasons stated above, this invention provides a hierarchical three-dimensional UAV logistics trajectory planning method. Please refer to [link to relevant documentation]. Figure 1 , Figure 1 This is a flowchart illustrating a hierarchical three-dimensional UAV logistics trajectory planning method provided by the present invention. The hierarchical three-dimensional UAV logistics trajectory planning method provided in this embodiment includes: Step 1: Obtain logistics task information for the task area. The logistics task information includes task node information and environmental constraint information. The task node information includes take-off and landing points as well as several pairs of pick-up and delivery nodes consisting of pick-up nodes and delivery nodes.

[0018] Specifically, the logistics task information acquired in the task area is the input data required for the UAV to perform a logistics task. The logistics task information includes at least two parts: task node information and environmental constraint information.

[0019] In this embodiment, the task node information refers to all the spatial locations that the UAV needs to access during the execution of a logistics task and their interrelationships. The task node information includes a set of task nodes consisting of multiple task nodes, including take-off and landing points, several pickup nodes, and several delivery nodes. The task node set is represented as follows: ; Where 0 represents the take-off and landing point, This represents the nth pickup node. Indicates and Corresponding delivery nodes This represents the number of node pairs for sending and receiving data, n=1,2,...N. This represents the set of task nodes consisting of the take-off and landing points, all pickup nodes, and all delivery nodes. The complete set of nodes required to determine the subsequent 3D trajectory matrix and access sequence optimization.

[0020] Furthermore, the set of task nodes One of the pickup nodes and a corresponding delivery node This constitutes a pickup / delivery node pair, and the access order of each pickup / delivery node pair is as follows: pickup node... and delivery nodes That is, any delivery node Must be picked up at the corresponding pickup point This is then accessed to describe the pickup and delivery operations that the drone needs to complete sequentially. The pickup / delivery node pairs and the pickup / delivery order constraints can be represented as: ; ; in, Represents the set of fetch and send node pairs. Indicates the first Group retrieval and delivery node pairs, This represents the total number of node pairs to be retrieved and sent. Indicates the task access sequence. Indicates the task node in the task access sequence The two formulas above are used to define the goods pickup and delivery pairing relationship and to constrain the delivery node to access the corresponding pickup node later than the pickup node.

[0021] In addition, to highlight the height difference between pickup and delivery nodes, a height attribute and height layer mapping are configured for each task node. Let the task node... The three-dimensional coordinates are Then, its corresponding candidate height layer can be determined by the nearest height layer or the reachable height layer that meets the task requirements, that is: ; in, Represents task node The three-dimensional coordinates and Represents the horizontal and vertical coordinates. Represents task node height, Indicates the total number of discrete height layers. Indicates the first There are discrete height layers, k=1,2,...,K. Represents task node The height layer number obtained from the mapping, Indicates selection Minimum height layer. This formula is used to map continuous 3D task nodes to discrete height layers in a hierarchical 3D motion space.

[0022] For example, in an embodiment with a unified height layer for pickup nodes, each pickup node can be configured to meet the following requirements: For delivery nodes, different delivery nodes can be mapped to different delivery height levels to represent differences in floor, roof, or platform heights, i.e.: ; in, and These represent the task heights of the pickup node and the delivery node, respectively. Indicates the uniform pickup height level. Indicates the first Delivery height level, This indicates the number of delivery height levels, s=1,2,...,S. This formula is used to express the idea that pickup points can be concentrated at a single height level, while delivery nodes can be distributed across different height levels according to floor, roof, or platform height.

[0023] When a mission node has multiple optional landing altitudes or hovering delivery altitudes, multiple candidate 3D nodes are created for the same mission node. When performing a 3D graph search on any two mission nodes subsequently, a combination of start and end altitudes that satisfies communication risk, obstacle avoidance, climb and descent capabilities, and energy consumption constraints is selected from the candidate 3D nodes, and a corresponding 3D waypoint sequence is generated.

[0024] In this embodiment, environmental constraints are external limitations that the UAV must adhere to during logistics missions. These constraints include at least one of the following: no-fly zone distribution, obstacle distribution, flight altitude range, horizontal speed range, vertical speed range, maximum rate of climb, maximum rate of descent, communication station location, and communication reliability threshold. A communication station can be a base station or a communication access point. The communication reliability threshold is the minimum communication quality requirement to ensure normal communication between the UAV and the ground communication station. When the actual communication quality is lower than the communication reliability threshold, communication at that location or flight segment is deemed unreliable. Specifically, the no-fly zone distribution and obstacle distribution define areas the UAV cannot access; the flight altitude range, horizontal speed range, vertical speed range, maximum rate of climb, and maximum rate of descent limit the UAV's flight capability boundaries; and the communication station location and communication reliability threshold describe the coverage conditions and service quality requirements of the communication environment. These environmental constraints collectively form the basis for subsequent three-dimensional path search and feasibility verification.

[0025] In addition, this step also obtains the cargo weight of each set of pick-up and delivery nodes, the maximum total cargo weight (i.e., maximum payload) that the drone is allowed to carry during flight, and the total available power reserves of the drone.

[0026] Through the above processing, this step transforms the original logistics task data into a standardized set of task nodes with height attributes, pick-up and delivery pairing relationships, load constraints, and spatial constraints, providing a unified input for subsequent communication risk modeling, hierarchical 3D pathfinding, and access sequence optimization.

[0027] Step 2: Discretize the task region in the height direction to construct a hierarchical 3D directed graph.

[0028] Specifically, the vertical space in which the UAV can fly is divided into several discrete height layers according to a preset interval. Horizontal movement edges are established within each height layer, and vertical switching edges are established between different height layers, thus forming a three-dimensional directed graph with discrete height layers as the hierarchical structure and horizontal movement edges and vertical switching edges as the connection relationships. This hierarchical three-dimensional directed graph is used to represent the set of traversable states and executable actions of the UAV in three-dimensional space, providing a structured search space for subsequent three-dimensional path search.

[0029] In an optional embodiment, step 2 includes: Step 2.1: Discretize the task area into multiple discrete height layers in the height direction.

[0030] Specifically, the vertical space in which the drone can fly is divided into several parallel discrete height layers according to a preset spacing or rule. Each discrete height layer corresponds to a fixed height value or height range. The set of discrete height layers is represented as follows: ; in, This represents a set of discrete altitude layers. The number and spacing of discrete altitude layers can be flexibly determined based on factors such as flight altitude range, obstacle distribution, communication environment characteristics, or the distribution altitude of mission nodes. For example, a discrete altitude layer can be 3m.

[0031] Step 2.2: Establish a horizontal motion network within each discrete height layer, wherein the horizontal movement edges in the horizontal motion network are used to connect different horizontal positions within the same discrete height layer.

[0032] Specifically, within each discrete altitude layer, based on the corresponding horizontal spatial range and environmental constraints such as obstacles and no-fly zones, a network structure is constructed consisting of several horizontal nodes and horizontal movement edges connecting these nodes. This mesh structure is the horizontal motion network. The horizontal motion network is used to enable the UAV to move from any position to another within the same discrete altitude layer, and it serves as the basic path framework for the UAV to perform horizontal flight in three-dimensional space.

[0033] Step 2.3: Establish vertical switching edges between different discrete height layers. The vertical switching edges are used to connect the same horizontal position or adjacent horizontal positions between different discrete height layers.

[0034] Specifically, between adjacent or non-adjacent discrete altitude layers, vertical connecting channels are established for passable nodes at the same or adjacent horizontal positions. These connecting channels are called vertical switching edges. Vertical switching edges are used to enable UAVs to climb or descend between different discrete altitude layers and are necessary path elements for UAVs to switch from the current discrete altitude layer to the target discrete altitude layer.

[0035] Step 2.4: Perform feasibility screening on the horizontal motion network and the vertical switching edge, remove horizontal movement edges and vertical switching edges that do not meet the preset conditions, and construct a layered three-dimensional directed graph. The layered three-dimensional directed graph includes several discrete height layers, passable horizontal movement edges in each discrete height layer, and passable vertical switching edges connecting different discrete height layers.

[0036] Specifically, all initially established horizontal movement edges and vertical switching edges are checked one by one, and movement edges that do not meet the preset conditions are eliminated. The preset conditions include at least one of the following: the movement edge does not cross obstacles, does not cross no-fly zones, does not exceed the flight altitude range, the horizontal speed does not exceed the maximum allowable value, the vertical speed does not exceed the maximum climb rate or maximum descent rate, the climb angle does not exceed the maximum allowable value, and the climb energy consumption under different loads does not exceed the preset threshold. After feasibility screening, only movement edges that meet the preset conditions are retained, thus forming a clean and executable hierarchical three-dimensional motion space, ensuring that every path subsequently searched in this space is feasible for actual flight.

[0037] In this embodiment, if the set of horizontally accessible nodes is Q, then the hierarchical set of accessible nodes can be represented as: ; in, This represents a hierarchical set of accessible nodes, obtained by combining horizontally accessible nodes and discrete height layers. This represents a horizontally accessible node.

[0038] Therefore, the hierarchical 3D directed graph constructed on the hierarchical set of walkable nodes is as follows: ; in, Represents a layered three-dimensional directed graph. This represents the set of horizontally accessible edges within the same layer, used to express horizontal flight between adjacent or horizontally accessible nodes within the same discrete height layer. This represents the set of vertical switching edges that can be traversed across layers, used to express ascent or descent between different discrete height layers at the same horizontal coordinate or within a preset horizontal neighborhood.

[0039] Therefore, to ensure the feasibility of actual flight for the 3D motion edges, speed, climb / descent rate, and climb angle constraints are set for each candidate motion edge. For the 3D motion edge e=(a,b) connecting node a and node b, its horizontal distance and height changes are as follows: ; ; in, Representing an edge Horizontal distance, Representing an edge Height difference between the two ends and Representing nodes respectively and nodes The three-dimensional coordinates Indicates passing through the edge Time required Indicates the maximum vertical velocity. Indicates the maximum angle of ascent or descent. To avoid extremely small positive numbers where the denominator is zero when the horizontal distance is zero, the first set of formulas is used to calculate the changes in horizontal distance and height of candidate moving edges, while the second set of inequalities is used to filter out unexecutable edges that exceed the vertical speed or climb angle capability.

[0040] If the edge If an edge crosses a no-fly zone or an area of ​​obstacles, that edge is removed. ; in, Indicates the connection node and nodes spatial line segments, Indicates the area of ​​obstacles. Indicates a no-fly zone. This indicates that there is no intersection. This expression means that candidate 3D edges must not cross obstacles or no-fly zones.

[0041] Through the above steps, a searchable 3D candidate space containing horizontal movement edges and vertical switching edges is formed, and the distance, height change, feasible speed range, communication risk, and vertical maneuver cost of each candidate movement edge are output as edge attributes.

[0042] Step 3: Construct a three-dimensional communication risk map based on the communication risk values ​​of different spatial location units in the task area.

[0043] Specifically, after dividing the mission area into several spatial location units, the communication risk value of each spatial location unit is determined based on communication-related data. Then, the communication risk values ​​of each spatial location unit are stored or associated according to horizontal position and height to form a three-dimensional data structure that reflects the communication quality level at different locations using spatial location as an index. This three-dimensional data structure is the three-dimensional communication risk map, which characterizes the spatial distribution of the reliability of communication quality at different spatial location units within the mission area.

[0044] In an optional embodiment, step 3 includes: Step 3.1: Divide the task area into several spatial units.

[0045] Specifically, the three-dimensional space covered by the drone performing the logistics task is divided into a series of discrete, non-overlapping sub-regions according to preset rules, with each sub-region being a spatial location unit. This division is typically carried out in both the horizontal and vertical directions: in the horizontal direction, the task area is divided into a regular horizontal grid. In the vertical direction, K discrete height layers are divided. The combination of horizontal grids and height layers constitutes a three-dimensional spatial location unit, which is formed by the first... Horizontal grid The spatial location unit formed by the k-th discrete height layer is represented as: ; in, Indicates the first Horizontal grid The spatial location unit formed by the k-th discrete height layer, This represents the total number of horizontal grid cells.

[0046] Step 3.2: Based on at least one data source among the communication propagation model, historical link measurement data, signal-to-interference-plus-noise ratio prediction results, and real-time link quality data, determine the communication risk value for each spatial location unit. The communication risk value is used to characterize the possibility that the communication quality within the spatial location unit does not meet the preset reliability requirements.

[0047] Specifically, the communication propagation model is a theoretical model based on electromagnetic wave propagation theory and base station parameters, which calculates signal attenuation in space using mathematical formulas; historical link measurement data is communication link quality data actually collected and recorded during previous UAV flight missions; the signal-to-interference-plus-noise ratio (SIR / NOT) prediction result is obtained by using machine learning or statistical models to predict the SIR / NOT of unsampled locations based on input features; and real-time link quality data is the latest communication link quality data collected and transmitted back by the UAV during its current flight. Therefore, when a spatial location unit has at least one of the above data sources, the communication risk value of that spatial location unit can be determined. The larger the value, the higher the probability that the communication quality within that spatial location unit will not meet the preset reliability requirements.

[0048] In a specific embodiment, if only one of the following data sources exists within the current spatial location unit: communication propagation model, historical link measurement data, signal-to-interference-plus-noise ratio prediction results, and real-time link quality data, then the communication risk value of the current spatial location unit is determined based on the data source.

[0049] (1) If the data source of the current spatial location unit is a communication propagation model, let the spatial state of the UAV during flight be: ; in, Indicates the drone at a certain time The three-dimensional state, and Indicates the drone at a certain time Horizontal coordinates Indicates the drone at a certain time The flight altitude. This state variable serves as input for querying the 3D communication risk map and calculating link propagation loss.

[0050] For the first within the task area A communication station, which communicates with the drone in real time. 3D distance Represented as: ; in, Indicates the first The location of a communication site. For example, a communication site is a base station or a communication access point.

[0051] No. Each communication station at time The corresponding path loss can be expressed as: ; in, Indicates time No. Path loss for each communication station For reference distance Path loss at the location, This represents the path loss index. For a moment drones to the The distance between communication stations Indicates time Shadow decay, building occlusion, or experience modifiers.

[0052] Based on the transmit power, antenna gain, and path loss, the receive power can be obtained: ; in, Indicates time drones from the The signal power received by each communication station Indicates time No. The transmission power of each communication station, Indicates the elevation angle of the transmitting antenna. and azimuth Gain in direction, This indicates the gain of the drone's receiving antenna.

[0053] When the received power is expressed in dBm, Convert to linear power : ; Based on this, the first Each communication station at time Corresponding signal-to-noise ratio H-Signal Interference-to-Noise Ratio They are represented as follows: ; ; in, Indicates noise power. Indicates the first One interference source at time Interference power.

[0054] Therefore, the communication site with the highest signal-to-interference-plus-noise ratio is selected as the service site: ; in, Indicates time service stations, Indicates time The maximum value of communication quality indicators across all communication sites.

[0055] If the maximum signal-to-interference-plus-noise ratio (SINR) falling below the communication reliability threshold is taken as the communication interruption event, then when the UAV is at time... Located in spatial unit At that time, the spatial location unit obtained based on the communication propagation model At any moment Communication risk value for: ; in, Represents probability operators, This indicates the communication reliability threshold.

[0056] Since shadow fading, interference power, and small-scale fading are random, when the above probabilities are difficult to calculate directly analytically, we can use time... random channel state Second sampling, and estimation of communication risk value based on the proportion of interruption events: ; in, Indicates time No. Maximum signal-to-interference-plus-noise ratio under secondary channel sampling This indicates the indicator function. This method is suitable for grid cells with communication site parameters, propagation environment parameters, and interference models, but with a limited number of measured samples.

[0057] (2) If the data source for the current spatial location unit is historical link measurement data, first map the historical samples to the corresponding spatial location units according to the measurement location and altitude. Let the first... The location of each historical sample is Its grid attribution variable Represented as: ; in, Indicates the first The height range corresponding to each discrete height layer.

[0058] When the maximum signal-to-interference-plus-noise ratio is lower than the communication reliability threshold as the interruption criterion, the first... Disruption markers for historical samples for: ; in, A value of 0 indicates the first... The historical samples communicated normally. A value of 1 indicates the first... Communication was interrupted for one historical sample. Indicates the first The maximum signal-to-interference-plus-noise ratio of a historical sample.

[0059] When it is necessary to consider packet loss rate, latency, or duration of continuous interruption simultaneously, communication quality exceeding limits events can also be defined as: ; in, This represents the event where the overall communication quality exceeds the limit for the d-th historical sample. This represents the actual packet loss rate of the d-th historical sample. This indicates the preset maximum allowable packet loss rate threshold. This represents the actual communication latency of the d-th historical sample. This indicates the preset maximum allowable delay threshold. This represents the duration of consecutive communication interruptions in the d-th historical sample. This indicates the preset maximum allowed continuous interruption duration threshold.

[0060] And order To enhance the effectiveness of recent and high-reliability data, the statistical weight of the d-th historical sample is... It can be represented as: ; in, This represents the credibility of the d-th historical sample. This represents the collection time of the d-th historical sample. This represents the time decay coefficient of historical data.

[0061] Therefore, the first The communication risk value for each spatial location unit is: ; in, Indicates the first Historical communication risk values ​​for each spatial location unit This represents the total number of historical samples. The formula represents the weighted proportion of communication quality exceedance events within the corresponding spatial location unit, and the result is... lie in to between.

[0062] (3) If the data source of the current spatial location unit is the signal-to-interference-plus-noise ratio prediction result, the input features are constructed based on the horizontal grid position, height, relative position of communication stations, obstacle occlusion characteristics, and existing link measurements. And the prediction model is used to obtain the predicted value of the maximum signal-to-interference-plus-noise ratio of the current spatial location unit: ; in, Indicates the first The maximum signal-to-interference-plus-noise ratio prediction value for each spatial location cell. The parameter is The signal-to-interference-plus-noise ratio prediction model.

[0063] Model parameters can be determined based on the prediction error of the training samples: ; in, Represents the optimal model parameters. Let f represent the total number of training samples, where f = 1, 2, ... , This represents the true maximum signal-to-interference-plus-noise ratio (SIR) of the f-th training sample. For using parameters and input features The predicted maximum signal-to-interference-plus-noise ratio.

[0064] Let the standard deviation of the predicted residuals be... for: ; When using the Gaussian prediction error approximation, the first Predicted communication risk value for each spatial location unit It can be represented as: ; in, This represents the cumulative distribution function of the standard normal distribution. This formula indicates the probability that the actual maximum signal-to-interference-plus-noise ratio (SIR) is lower than the communication threshold, after considering prediction errors.

[0065] If the prediction model directly outputs the conditional probability distribution of the maximum signal-to-interference-plus-noise ratio, then the... Predicted communication risk value for each spatial location unit It can also be expressed as: ; in, This represents the conditional probability distribution for the maximum signal-to-interference-plus-noise ratio.

[0066] (4) If the data source of the current spatial location unit is real-time link quality data, set the length to be before the current time t. A sliding time window is used, and the data within the window is mapped to the corresponding spatial location unit. For received power, signal-to-interference-plus-noise ratio, packet loss rate, latency, duration of continuous interruption, and number of communication site handovers, these can be converted to values ​​ranging from [value range missing]. to Furthermore, a higher numerical value indicates a higher level of risk. For example: ; ; ; ; ; ; in, , , , , , These represent the power risk component, signal-to-interference-plus-noise ratio risk component, packet loss rate risk component, delay risk component, continuous interruption duration risk component, and handover risk at the s-th sampling point within the sliding time window, respectively. , , , , , These represent the power value, signal-to-interference-plus-noise ratio, packet loss rate, latency, duration of continuous interruption, and number of handovers at the s-th sampling point within the sliding time window, respectively. Indicates the safe power threshold. Indicates the minimum tolerable power. This represents the minimum tolerable signal-to-interference-plus-noise ratio. This represents the safe packet loss rate threshold. This indicates the maximum tolerable packet loss rate. Indicates the safe delay threshold. This indicates the maximum tolerable latency. Indicates the threshold for the duration of continuous safety interruption. Indicates the maximum tolerable duration of continuous interruptions. Indicates the threshold for the number of safe handovers. This indicates the maximum number of tolerable switching operations.

[0067] The comprehensive risk component of the s-th sampling point It can be represented as: ; ; in, , , , , , These are the weighting coefficients for the power risk component, signal-to-interference-plus-noise ratio risk component, packet loss rate risk component, delay risk component, continuous interruption duration risk component, and handover risk, respectively, and each weighting coefficient is greater than or equal to 0.

[0068] Within a sliding time window, the real-time risk components are aggregated using a time decay weight, resulting in: ; in, For the current moment The set consists of all the sampling points s within the previous sliding time window. , Indicates the duration of the sliding time window. This represents the sampling time of the s-th sampling point. This represents the time decay coefficient of real-time data.

[0069] Without probability calibration, This serves as a normalized real-time communication risk value; when it needs to be converted into a probability of communication interruption, historical interruption samples can be used for probability calibration. ; in, Indicates the first Each spatial location unit at the current moment The communication risk value (i.e., the probability of communication interruption). and These are calibration parameters fitted based on historical interruption markers. Therefore, various heterogeneous link quality indicators in real-time telemetry are converted into communication risk values ​​with a unified direction and a unified value range.

[0070] In one specific embodiment, if there are at least two data sources among the communication propagation model, historical link measurement data, signal-to-interference-plus-noise ratio prediction results and real-time link quality data in the current spatial location unit, then the communication risk values ​​calculated from each data source are weighted and fused to obtain the communication risk value of the current spatial location unit.

[0071] Specifically, when communication propagation models exist simultaneously Historical link measurement data Signal-to-interference-plus-noise ratio prediction results and real-time link quality data First, the methods described above are used to obtain the corresponding data sources. , , and Then, the fusion weights are determined based on the reliability of the data source, the number of valid samples, and the freshness of the data. Let the data source set be... , No. The first spatial location unit Unnormalized weights of data sources Represented as: ; in, Indicates the first The credibility of the data source Indicates the first The number of valid samples or the number of valid simulations for the data source. Indicates the first The parameter for adjusting the number of samples in a data source is as follows. Indicates the first Data delay of the data source from the current time t. Indicates the first Time decay coefficient of data source Therefore, normalized fusion weights for: ; Finally, the first Communication risk value of each spatial location unit Represented as: ; in, For the first The first spatial location unit The corresponding data of each class , , and . satisfy: ; When a data source becomes unavailable, its unnormalized weights are set to zero, and the risk calculation is performed by the remaining available data sources. Thus, the model calculation results are used to supplement areas without measured data, historical data is used to reflect long-term spatial distribution characteristics, prediction results are used to estimate the communication status of unsampled locations or future periods, and real-time data is used to correct for changes in the current flight environment.

[0072] In a specific embodiment, if there is no data source among the following—communication propagation model, historical link measurement data, signal-to-interference-plus-noise ratio prediction results, and real-time link quality data—within the current spatial location unit, the communication risk value of the current spatial location unit is obtained through spatial interpolation.

[0073] Specifically, for spatial location units lacking all of the aforementioned data sources, spatial interpolation can be performed based on the communication risk values ​​of adjacent spatial location units. Let... Indicates the first The set of adjacent spatial units of the _i_ spatial unit, then the _i_th spatial unit obtained by interpolation The communication risk value of a spatial location unit can be expressed as: ; in, Indicates the center of adjacent spatial location units and the first The three-dimensional distance between the centers of each spatial location unit This represents the distance decay index.

[0074] Step 3.3: Map the communication risk value of each spatial location unit to the corresponding spatial location to form a three-dimensional communication risk map that varies with horizontal position and height.

[0075] Specifically, based on the communication risk values ​​of each spatial location unit calculated above, these discrete communication risk values ​​are stored at their corresponding three-dimensional physical coordinate points through spatial coordinate lookup and data binding. This process transforms the abstract algorithm output into an intuitive spatial distribution pattern, thereby constructing a three-dimensional communication risk map that dynamically changes with horizontal position and height. For any spatial location... You can query the grid to which it belongs. It can also be obtained by 3D interpolation based on adjacent grids. .

[0076] The final output 3D communication risk map does not determine the mission path independently, but rather serves as a risk attribute in the 3D trajectory matrix of mission nodes for subsequent evaluation. For the same horizontal path, communication risk values ​​at different discrete altitude layers can be compared; for candidate segments with excessive risk, discrete altitude layer replacement or local 3D rerouting can be prioritized while keeping the pick-up and delivery access sequence unchanged.

[0077] For any three-dimensional candidate track consisting of several discrete waypoints The communication risk values ​​corresponding to each sampling point can be accumulated according to the sampling time to obtain the cumulative communication risk value of the matrix element: ; in, This represents the cumulative communication risk value of the j-th segment of the i-th candidate track. This represents the total number of sampling points for the j-th segment of the i-th candidate track. =1,2,..., , For the first Communication risk value for each sampling point Indicates the first The location coordinates of each sampling point Indicates the first Flight time at each sampling point.

[0078] Alternatively, average communication risk can be used. Maximum communication risk and cumulative duration of communication risk exceeding limits As a communication constraint indicator: ; ; ; in, This represents the communication risk threshold. When using segment length accumulation, the value in the above formula can be used as a reference. Replace with the corresponding flight segment length During feasibility verification, the average risk, maximum risk, or cumulative duration of risk exceeding the limit can be required to not exceed a preset threshold, and based on this, it can be determined whether to perform height layer replacement, local rerouting, or speed re-optimization.

[0079] Through the above processing, the communication risk of the three-dimensional mesh can be calculated using the communication propagation model, historical link measurement data, signal-to-interference-plus-noise ratio prediction results and real-time link quality data, respectively. When multiple data are available at the same time, the fused communication risk can be obtained by normalization weighting, so that the communication reliability is transformed from empirical judgment into constraints and cost terms that can be directly invoked by three-dimensional pathfinding, three-dimensional trajectory matrix generation of task nodes and feasibility verification.

[0080] Step 4: Perform a 3D path search in the hierarchical 3D directed graph to obtain candidate 3D trajectories between task nodes, and generate trajectory attribute matrices between task nodes corresponding to multiple discrete load levels based on the 3D communication risk graph.

[0081] Specifically, in the hierarchical 3D directed graph, the payload range of the UAV from empty to maximum permissible payload is first discretized into multiple payload levels, each payload level corresponding to a preset payload value. Then, for each payload level, any two task nodes in the task node set are mapped as source and sink points, respectively. A 3D path search is performed in the hierarchical 3D directed graph to obtain candidate 3D trajectories between task nodes under the current payload level. Subsequently, based on the 3D communication risk graph, the communication risk value corresponding to the waypoints, waypoint segments, or overall paths traversed by each candidate 3D trajectory is calculated. At the same time, at least one of the geometric attributes, energy consumption attributes, and time attributes can be determined by combining the 3D waypoint / edge sequence of the candidate 3D trajectory. Then, the multidimensional attributes of a candidate 3D trajectory between each ordered pair of task nodes are recorded to generate a trajectory attribute matrix between task nodes.

[0082] In an optional embodiment, step 4 includes: Step 4.1: Discretize the payload range of the drone from empty to maximum allowable payload into multiple payload levels. Each payload level has a corresponding preset payload value, and the preset payload value increases with the payload level.

[0083] Specifically, the payload range of the drone from empty to maximum permissible payload is divided into several discrete payload levels according to preset intervals or rules. Each payload level corresponds to a specific preset payload value, and the corresponding preset payload value increases as the payload level number increases. For example, if the maximum permissible payload is 10 kg, it can be discretely divided into 5 levels, corresponding to preset payload values ​​of 2 kg, 4 kg, 6 kg, 8 kg and 10 kg, respectively. The first level corresponds to 2 kg, the second level corresponds to 4 kg, and so on.

[0084] Step 4.2: For each discrete load level, traverse each task node in the task node set as the starting task node, and traverse each task node other than the starting task node as the ending task node. Map the starting task node and the ending task node to the layered three-dimensional directed graph to determine the source point and sink point. Based on the source point and sink point, perform three-dimensional path search in the layered three-dimensional directed graph to obtain the candidate three-dimensional trajectory set that meets the preset constraints.

[0085] Specifically, firstly, for each discrete payload level, the preset payload value corresponding to that payload level is used as the payload value for the current path search, and remains unchanged throughout the current path search. Since the UAV only updates its payload after visiting a task node and performing pickup or delivery, a fixed payload is used for a single segment search between the same ordered task node pairs. Then, all task nodes in the task node set are traversed using a double loop. In the outer loop, each task node in the task node set is sequentially selected as the starting task node p. In the inner loop, each task node except the current starting task node p is sequentially selected as the ending task node q.

[0086] For each given starting task node p and ending task node q, firstly, based on the three-dimensional coordinates of the starting task node p and the ending task node q, search for graph nodes in the layered three-dimensional directed graph that perfectly match or are the nearest neighbors of their spatial positions (horizontal coordinates and height layers). Define the graph node matching the starting task node p as the source node of the path search, and define the graph node matching the ending task node q as the sink node of the path search. Then, with the source node as the starting point and the sink node as the ending point, perform a three-dimensional path search in the layered three-dimensional directed graph: initialize the priority queue, the minimum cost record table, and the parent node record table, add the source node to the priority queue, set the initial cost of the source node to 0, and set the initial cost of the remaining graph nodes to positive infinity. Assuming all candidate edge movement costs are non-negative, the system enters a cyclic expansion phase. The node with the smallest cumulative cost is selected from the current priority queue as the current expanded node, and its sink node is checked. If it is, the system proceeds to the path backtracking phase; otherwise, it searches all neighboring nodes of the current expanded node in the hierarchical 3D directed graph, including horizontal neighbors within the same height layer and vertical neighbors across different height layers. For each candidate edge connecting the current expanded node and a neighboring node, the corresponding cost function is called to calculate the movement cost based on the edge type (horizontal movement, ascending, or descending). Simultaneously, real-time checks are performed based on preset constraints, pruning and eliminating search branches that violate any preset constraints, preventing further expansion. For neighboring nodes that satisfy all preset constraints, the new cumulative cost from the source node to the neighboring node via the current expanded node is calculated. If this cumulative cost is less than the neighboring node's original minimum cost record, the minimum cost record and parent node record of the neighboring node are updated, and the neighboring node is added to the priority queue for subsequent expansion. After expanding all neighboring graph nodes of the current expanded graph node, the next graph node with the minimum cumulative cost is retrieved from the priority queue, and the above expansion, calculation, checking, and updating process is repeated. When the graph node retrieved from the priority queue is the sink, the process is reversed from the sink through the parent node record table, tracing back to the source node level by level, thus obtaining a complete three-dimensional waypoint sequence from the source node to the sink. This three-dimensional waypoint sequence is the candidate three-dimensional trajectory of the ordered node pair (p and q). If the sink has not been reached when the priority queue is empty, the trajectory attribute matrix element corresponding to the ordered task node pair under the current load level is marked as unreachable.

[0087] In the current discrete load level 3D path search process, different cost functions are used for horizontal movement edges, climb edges, and descent edges at the same level to reflect the energy consumption difference between load climb and horizontal flight and descent maneuvers. Furthermore, the cost coefficient of climb edges is higher than that of horizontal movement edges and descent edges. The cost functions are expressed as follows: ; ; ; ; in, Indicates the drone's payload capacity is At that time, along the edge The cost of horizontal movement along the edge during flight. Indicates the drone's payload capacity is At that time, along the edge The cost of climbing to the edge of flight, Indicates the drone's payload capacity. At that time, along the edge The cost of the descent edge of flight, Indicates passing through the edge Load capacity at that time This represents the horizontal distance that edge e moves. Representing an edge The corresponding communication risk value, Indicates the altitude climbed. Indicates descent altitude. , and The energy consumption conversion factor and cost factor vary with load. This represents the communication risk weight. The coefficient relationship indicates that the energy consumption cost of a load-bearing climb edge is higher than the cost of equivalent horizontal flight and descent maneuvers. When performing a three-dimensional path search for any discrete load level, the aforementioned load W is taken as the preset load value corresponding to the current discrete load level. The aforementioned energy consumption conversion coefficient and communication risk weight are all non-negative numbers, and the horizontal movement distance, climb altitude, descent altitude, and communication risk value of the edge are all non-negative quantities to ensure that the cumulative cost does not decrease as the path expands.

[0088] Optionally, the preset constraints include at least one of the following: spatial and obstacle constraints, flight performance constraints, communication and link constraints, and energy and endurance constraints. Spatial and obstacle constraints require that candidate edges not cross obstacle areas and no-fly zones, and that flight altitude be within permissible limits; flight performance constraints require that the rate of climb / descent, climb angle, horizontal speed, and turning angle not exceed the upper limit of the UAV's maneuverability; communication and link constraints require that the communication risk value or cumulative communication risk of candidate edges not exceed a preset threshold; and energy and endurance constraints require that the cumulative energy consumption from the source point to the current node not exceed the remaining battery energy of the UAV.

[0089] Step 4.3: For each discrete load level, calculate the cumulative communication risk value of the candidate three-dimensional trajectory using the three-dimensional communication risk map, and construct the trajectory attribute matrix between task nodes based on the cumulative communication risk value, track length and energy consumption of the candidate three-dimensional trajectory. Each preset load value corresponds to a trajectory attribute matrix, and the trajectory attribute matrix is ​​used to characterize the three-dimensional track attributes between task nodes.

[0090] Specifically, for the candidate 3D trajectory searched from the starting task node p to the ending task node q, firstly, in conjunction with the 3D communication risk map, the corresponding communication risk value is queried for each discrete sampling point or segment of the candidate 3D trajectory, and weighted and accumulated according to time or segment length to obtain the cumulative communication risk value of the trajectory; simultaneously, each edge in the candidate 3D trajectory is traversed, its horizontal distance is accumulated to obtain the total horizontal distance, the positive altitude change is accumulated to obtain the cumulative climb altitude, and the absolute value of the negative altitude change is accumulated to obtain the cumulative descent altitude. Based on the preset load value corresponding to the current discrete load level, the horizontal flight energy consumption, load climb energy consumption, and descent maneuver energy consumption are calculated respectively, thereby obtaining the expected flight time and expected total energy consumption; finally, the path search cost, total horizontal distance, cumulative climb altitude, cumulative descent altitude, load climb energy consumption, descent maneuver energy consumption, cumulative communication risk value, expected flight time, and the 3D waypoint sequence of the trajectory are encapsulated together as a matrix element of the ordered node pair in the trajectory attribute matrix.

[0091] In other words, when generating the elements of the trajectory attribute matrix, the edge attributes in the candidate 3D trajectory are accumulated to obtain attributes such as horizontal distance, cumulative climb altitude, cumulative descent altitude, communication risk value, and estimated flight time, which are represented as follows: ; ; ; in, Indicates from task node To the task node The total horizontal distance, Representing an edge Belongs to task node To the task node flight path , Representing an edge Horizontal length, Indicates from task node To the task node Cumulative elevation gain Indicates from task node To the task node The cumulative descent altitude, Representing an edge The height difference This indicates that only the climb portion is accumulated. This indicates that only the cumulative decrease is counted. Indicates from task node To the task node The estimated flight time Indicates drones along the edge Flight time, Indicates from task node To the task node The cumulative value of communication risks, Indicates drone along the edge The communication risk value for flight. This set of formulas is used to summarize the underlying edge attributes into the horizontal distance, cumulative ascent, cumulative descent, estimated time, and communication risk attributes in the matrix elements.

[0092] Therefore, for the f-th discrete load level, the matrix elements from task node p to task node q in its corresponding trajectory attribute matrix can be represented as: ; in, This indicates that for the f-th discrete load level, from the task node... To the task node matrix elements, Indicates from task node To the task node The path search cost is a scalar value accumulated based on the cost function of each candidate edge under the current load level, used for generating or sorting candidate 3D trajectories. Indicates from task node To the task node The total horizontal distance, Indicates from task node To the task node Energy consumption for load climbing Indicates from task node To the task node Reduced maneuvering energy consumption Indicates from task node To the task node The estimated total energy consumption, This represents a 3D waypoint sequence. The formula encapsulates the 3D pathfinding results between pairs of task nodes into a matrix attribute that can be invoked by access sequence optimization. Therefore, the trajectory attribute matrix of a task node not only stores the shortest geometric path between two nodes, but also the executable attributes of that path under different altitude layers, communication environments, and load conditions, facilitating direct invocation by access sequence optimization.

[0093] The above process is repeated for all discrete load levels to obtain multiple trajectory attribute matrices corresponding to different preset load values. These trajectory attribute matrices collectively characterize the executable attributes of candidate 3D trajectories between task nodes under different load conditions. During subsequent task access sequence optimization, the actual load is calculated segment by segment when executing the current segment. When the actual load equals a preset load value, the trajectory attribute matrix corresponding to that preset load value is called. When the actual load is between two adjacent preset load values, and the matrix elements of the two adjacent load levels contain candidate 3D trajectories with the same 3D waypoint sequence, linear interpolation is performed on the load-related numerical attributes of the candidate 3D trajectory according to the position of the actual load between the two adjacent preset load values ​​to improve the overall evaluation accuracy. Non-numerical attributes such as the 3D waypoint sequence are not interpolated. For hard constraints such as maximum payload, cumulative energy consumption, and endurance, the trajectory attribute matrix corresponding to the preset payload value that is not less than the actual payload and has the smallest difference from the actual payload is still used for conservative verification. When the candidate 3D trajectories corresponding to two adjacent payload levels are different, the aggregated attributes of different candidate 3D trajectories are not directly interpolated. Instead, the candidate 3D trajectory corresponding to the higher payload level is used for conservative evaluation, and the corresponding attributes are recalculated and the feasibility condition verification is performed in step 6 according to the actual payload of the current flight segment. When the actual payload exceeds the maximum allowable payload, the corresponding task access sequence is determined to not meet the payload constraints.

[0094] Step 5: Based on the trajectory attribute matrix between the task nodes corresponding to multiple discrete load levels, generate a set of candidate task access sequences under the condition of satisfying task execution constraints.

[0095] Specifically, firstly, a first task access sequence that satisfies the order of pick-up and delivery is generated based on the take-off and landing point locations and pick-up and delivery pairing relationships. Under the condition of satisfying the task execution constraints, a search algorithm is executed on the first task access sequence and a segment-by-segment query based on the trajectory attribute matrix corresponding to the actual load of each flight segment is performed to generate a set of candidate task access sequences.

[0096] Optionally, the task execution constraints include at least one of the following: pickup / delivery order constraints, maximum payload constraints, communication risk constraints, and energy consumption constraints. The pickup / delivery order constraints require that the pickup node in each group of pickup / delivery nodes accesses the corresponding delivery node before the delivery node; the maximum payload constraints require that the actual payload of the UAV at any point during the entire task execution process does not exceed the maximum allowable payload; the communication risk constraints require that the cumulative value of communication risk for each segment in the candidate task access sequence does not exceed a preset communication risk threshold; and the energy consumption constraints require that the total energy consumption of the candidate task access sequence does not exceed the UAV's battery capacity.

[0097] In an optional embodiment, step 5 includes: Step 5.1: Generate the first task access sequence based on the take-off and landing point location, pick-up and delivery pairing relationship, task node height attribute, and maximum load constraint.

[0098] Specifically, the start and end points of the sequence are determined based on the take-off and landing points. The pickup node in each group of pickup and delivery nodes is visited before the corresponding delivery node based on the pickup and delivery pairing relationship. The pickup and delivery points are classified into layers according to the task node height attribute to optimize the vertical maneuvering order. For example, priority is given to visiting all task nodes in the same height layer. For scenarios that require cross-layer access, the task nodes of each layer are visited in ascending or descending order of height. During the sequence generation process, the load change is checked segment by segment along the access order to ensure that the actual load at any time does not exceed the maximum allowable load. Finally, a first task access sequence that satisfies the pickup and delivery order relationship and load constraints is generated.

[0099] UAVs along the mission access sequence During flight, the first The actual load of a segment can be expressed as: ; ; ; in, Indicates by A task access sequence consisting of access nodes. Represents the first in the sequence One access node, Indicates execution up to the 1st After visiting each node, the actual payload of the drone during the flight segment. Indicates access node The resulting change in load capacity. This represents the weight of the nth item. This indicates the maximum allowable payload of the drone. The segmented definition represents the access to the pickup node. Increase cargo weight and access delivery nodes. The corresponding cargo weight is reduced when the load changes to zero when accessing the take-off and landing points; the load constraint is used to eliminate overloaded access sequences.

[0100] For the j-th segment, based on its actual load, the corresponding matrix elements are determined according to the load class selection and interpolation rules described in step 4: during comprehensive evaluation, linear interpolation is performed on the load-related numerical attributes if the interpolation conditions are met; during constraint checks, the matrix elements corresponding to the nearest load class not lower than the actual load are used for conservative evaluation; if the three-dimensional waypoint sequences corresponding to adjacent load classes are different, no interpolation is performed, and the trajectory and its attributes corresponding to the higher load class are used; if the corresponding matrix element is marked as unreachable, the mission access sequence is removed.

[0101] Step 5.2: Based on the actual load of each segment in the first mission access sequence, determine the corresponding trajectory attribute matrix according to the rules described in Step 4 to perform preliminary feasibility screening of the first mission access sequence; if the first mission access sequence meets the mission execution constraints, then use the first mission access sequence as the initial current mission access sequence and the current optimal mission access sequence, and add the first mission access sequence to the elite sequence set; if the first mission access sequence does not meet the mission execution constraints, then regenerate the first mission access sequence and perform preliminary feasibility screening again until a first mission access sequence that meets the mission execution constraints is obtained.

[0102] Step 5.3: Using the current first task access sequence as the initial optimal task access sequence, perform iterative optimization using an improved tabu search algorithm; wherein, in each iteration, the following steps are executed: Step 5.31: Perform a local perturbation operation on the current task access sequence to generate the second task access sequence.

[0103] Specifically, after the first task access sequence passes the initial feasibility screening, it is used as the initial current task access sequence and the current optimal task access sequence for iterative optimization. In each iteration, the current task access sequence updated after the previous iteration is used as the search starting point for the current iteration, and a local perturbation operation is performed on the current task access sequence to generate the second task access sequence for the current iteration.

[0104] Furthermore, the local perturbation operation includes at least one of the following: swapping the positions of two task nodes in the sequence, inserting a task node into another position in the sequence, reversing the order of a subsequence in the sequence, moving a group of fetch / send nodes as a whole to another position in the sequence, aggregating and adjusting multiple task nodes at the same height level to be accessed consecutively, or rearranging task nodes at different height levels in height order.

[0105] Alternatively, this step can employ neighborhood search, genetic algorithm, simulated annealing, ant colony algorithm, particle swarm optimization, large neighborhood search, variable neighborhood search, local search, or a combination of the above algorithms.

[0106] Step 5.32: Based on the actual load of each segment in the second task access sequence, call the corresponding trajectory attribute matrix to check whether the second task access sequence meets the task execution constraints. If it does not meet the constraints, discard the second task access sequence and keep the current task access sequence and the current optimal task access sequence unchanged. If it meets the constraints, check whether the second task access sequence is prohibited by the taboo list.

[0107] Step 5.33: Update the current optimal task access sequence based on the taboo check results and comprehensive evaluation value of the second task access sequence.

[0108] Specifically, if the second task access sequence is not prohibited, the trajectory attributes between adjacent task nodes in the second task access sequence obtained in step 5.32 are used to obtain the comprehensive evaluation value of the second task access sequence, and the second task access sequence is used as the new current task access sequence. If the comprehensive evaluation value of the second task access sequence is less than the comprehensive evaluation value of the current optimal task access sequence, the second task access sequence is updated to the new current optimal task access sequence. If there is no access sequence with the same name as the second task access sequence in the elite sequence set, the second task access sequence is added to the elite sequence set. Otherwise, the second task access sequence is discarded, and the current task access sequence and the current optimal task access sequence remain unchanged.

[0109] If the second task access sequence is prohibited, the trajectory attributes between adjacent task nodes in the second task access sequence obtained in step 5.32 are used to obtain the comprehensive evaluation value of the second task access sequence, and it is determined whether the comprehensive evaluation value of the second task access sequence is less than the comprehensive evaluation value of the current optimal task access sequence. If so, the second task access sequence is allowed to break the taboo restriction, and the second task access sequence is used as the new current task access sequence and the new current optimal task access sequence. If there is no access sequence with the same name as the second task access sequence in the elite sequence set, the second task access sequence is added to the elite sequence set. If not, the second task access sequence is discarded, and the current task access sequence and the current optimal task access sequence remain unchanged.

[0110] For a second task access sequence that is not prohibited by the taboo list or meets the taboo-breaking conditions, if it is different from the existing task access sequences in the elite sequence set, it is added to the elite sequence set; when the number of task access sequences in the elite sequence set exceeds the preset number, they are sorted in ascending order of comprehensive evaluation value, and the preset number of top-ranked task access sequences are retained, and the current best task access sequence is always retained.

[0111] In this embodiment, the comprehensive evaluation value is obtained by calling the corresponding trajectory attribute matrix according to the actual payload of each flight segment, acquiring the communication risk, estimated flight time, and cumulative energy consumption between adjacent task nodes segment by segment along the access sequence of the task access sequence, and normalizing each trajectory attribute when necessary, and then weighting and summing them according to preset weight coefficients. The weight coefficients can be configured according to the task objectives (such as prioritizing energy saving, prioritizing low risk, or prioritizing short time). The final comprehensive evaluation value is a scalar value used to measure the overall quality of the candidate sequence. The smaller the comprehensive evaluation value, the better the performance of the sequence under the given task objectives.

[0112] Here, the formula for calculating the comprehensive evaluation value is: ; Where J(π) represents the comprehensive evaluation value of the task access sequence π. Let π represent the communication risk value of the flight segment from task node p to task node q, T(π) represent the estimated total time of the mission access sequence π, E(π) represent the estimated total energy consumption of the mission access sequence π, and β, γ, and δ are weighting coefficients. This function is used to uniformly evaluate candidate access sequences among communication risk, mission time, and flight energy consumption.

[0113] For ease of implementation, task access sequence It can be represented by arranging task nodes, or it can be represented by... The variable represents the adjacent access relationship. If Represents task node Then directly access the task node. Then the candidate task access sequence must satisfy the in-degree and out-degree constraints of each task node: ; in, For 0-1 adjacent access variables, Represents task node Then directly access the task node. , This represents the set of task nodes. The first equation indicates that each task node selects only one successor node, and the second equation indicates that each task node is entered by only one predecessor node. The in-degree and out-degree constraints mentioned above are used to limit the adjacency connection relationship of the candidate access sequence. When using the adjacency access variable to represent the task access sequence, it is also necessary to combine the in-degree and out-degree constraints of the start and end points as well as the sub-loop elimination constraints to ensure that all task nodes constitute a complete task access sequence from the start point to the end point.

[0114] The order of fetching and delivery can be constrained by accessing position variables or accumulated load variables. Accessing position variables... For example, you can set: ; in, Indicates pickup node Position variable in the task access sequence Indicates the corresponding delivery node The position variable in the task access sequence. This formula requires that the access position of the pickup point in each pickup-delivery pair is earlier than that of the delivery point, in order to ensure the "pick-up first, delivery later" logistics execution order.

[0115] Simultaneously, the load capacity and remaining battery power are updated segment by segment along the task access sequence. If the maximum load capacity is exceeded or the remaining battery power is negative at any stage, the candidate task access sequence is determined to be infeasible. ; ; in, Indicates completion of the first Load after each access node Indicates execution of the first The remaining battery power or energy before the end of the flight segment Indicates the first Energy consumed per segment of flight This indicates the lower limit of safe remaining battery power. The above formula is used to simultaneously check whether the payload and battery energy meet the actual execution requirements of the drone during the access sequence progression. Wherein, the... The energy consumed by a flight segment is obtained from the corresponding trajectory attribute matrix according to the actual load of the segment and the aforementioned load level selection and interpolation rules.

[0116] Step 5.34: In each iteration, add the object to be subjected to the local perturbation operation to the taboo list and update the taboo duration of the taboo list.

[0117] Specifically, the tabu list is used to record recently executed local perturbation operations and their corresponding short-term memory structures. A certain tabu period is set for each recorded operation (the tabu period can be adaptively adjusted according to the number of task nodes, the size of the candidate neighborhood, the number of consecutive unimproved times, or the number of search iterations). This can prevent the algorithm from repeatedly returning to the solution it just left or repeatedly executing the same perturbation during the search process and getting stuck in a local loop. At the same time, by breaking the tabu restrictions, the algorithm can balance the ability of local fine-grained search and global exploration, thereby guiding the search to jump out of the local optimum and evolve towards a better solution.

[0118] After each iteration completes the local perturbation operation and generates the second task access sequence, the object directly affected by the local perturbation operation actually executed in that iteration is added to the taboo list as a taboo record, and the taboo period is set or refreshed for the record (that is, the object is prohibited from being the target of the same type of perturbation operation in the next few iterations). At the same time, the length and period of the taboo list are maintained according to the first-in-first-out or decay strategy, so that the early forbidden objects are automatically unbanned after the period expires. This prevents the algorithm from repeated searches and cyclical oscillations in the short term, while retaining the long-term exploration capability of high-quality regions.

[0119] Step 5.35: Determine whether the preset iteration termination condition is met; if it is met, sort the task access sequences in the elite sequence set according to the comprehensive evaluation value in ascending order, and output the sorted candidate task access sequence set; if it is not met, execute the next iteration.

[0120] Specifically, after each iteration, it is determined whether the current iteration meets the preset iteration termination condition. If it does, the iteration process is terminated, and the task access sequences in the elite sequence set are sorted in ascending order of comprehensive evaluation value. The sorted elite sequence set is then output as the candidate task access sequence set. If it does not meet the condition, the process returns to the next iteration and continues to perform local perturbation, constraint checking, tabu judgment, current task access sequence update, current optimal task access sequence update, and elite sequence set update on the current task access sequence until the termination condition is met.

[0121] Optionally, the iteration termination condition includes at least one of the following: reaching the maximum number of iterations, the current optimal task access sequence not being updated in a number of consecutive iterations, the search time exceeding a preset time limit, or the comprehensive evaluation value converging to a preset threshold.

[0122] Step 6: Generate candidate 3D logistics trajectories based on the candidate task access sequence set, and perform feasibility condition verification and correction operations on each segment of the candidate 3D logistics trajectory to obtain a 3D logistics trajectory that meets the feasibility conditions.

[0123] Specifically, firstly, according to the sorting of the candidate task access sequence set determined in step 5, the current candidate task access sequence is selected; the candidate 3D trajectories of each segment corresponding to the current candidate task access sequence are stitched together to form a candidate 3D logistics trajectory. Subsequently, following the segment order of the candidate 3D logistics trajectory, starting from the first segment, the feasibility conditions are checked for each segment in turn; if the current segment passes all feasibility condition checks, the next segment is checked; if the current segment fails any check, it is marked as a problem segment and corrected. After correction, the feasibility condition checks are re-executed starting from the corrected current segment until all segments pass all checks, and finally, the task access sequence that meets the feasibility conditions and its corresponding 3D logistics trajectory are output.

[0124] This embodiment can also calculate the actual payload segment by segment, calculating the horizontal flight energy consumption, payload climb energy consumption, and descent maneuver energy consumption based on horizontal distance, cumulative climb altitude, and cumulative descent altitude, respectively. The total energy consumption of the k-th segment can be broken down as follows: ; ; ; in, Indicates the first Total energy consumption per flight segment Indicates energy consumption for horizontal flight. Indicates energy consumption for load climbing. This indicates a decrease in motor energy consumption. Indicates the first The actual load of each flight segment This indicates the cumulative altitude climbed during this flight segment. This indicates the cumulative altitude descent for this flight segment. and These represent the load-related climbing and descent energy consumption coefficients, respectively. This represents the energy consumption coefficient for horizontal flight. The coefficient inequality is used to show that the energy consumption coefficient for heavy-load climb is greater than that for horizontal flight and descent maneuvers.

[0125] When the objective is to minimize flight time, the energy consumption for climb and descent maneuvers can be deducted from the battery energy budget first, and then a speed optimization model can be established for the horizontal segment with the goal of minimizing total time and constrained by the remaining energy budget. When the objective is to minimize energy consumption, the optimal speed for energy consumption can be solved segment by segment based on the load-related power model.

[0126] In some embodiments, horizontal flight energy consumption can be further expressed as a function of horizontal distance, flight speed, and payload, and payload climb energy consumption can be expressed as a function of actual payload and climb altitude: ; ; ; in, Indicates speed as Load capacity is Horizontal flight power at that time Indicates the first Horizontal distance of each flight segment This means that the horizontal energy consumption is obtained by multiplying the horizontal flight time by the horizontal power. Indicates the weight of the drone. Represents gravitational acceleration. and These represent the energy conversion factors for the ascent and descent processes, respectively. This represents the energy conversion factor for horizontal flight. This set of formulas is used to estimate vertical maneuver energy consumption using actual payload, UAV weight, and altitude changes, and emphasizes that energy consumption for heavy-load climb increases in tandem with actual payload and cumulative climb altitude.

[0127] In an optional embodiment, step 6 includes: Step 6.1: Segment the candidate 3D trajectories of each flight segment corresponding to the current candidate task access sequence to form a candidate 3D logistics trajectory.

[0128] Specifically, according to the sorting of the candidate task access sequence set determined in step 5, the current candidate task access sequence is selected; for each pair of adjacent task nodes in the current candidate task access sequence, the corresponding trajectory attribute matrix is ​​determined according to the actual load of the corresponding flight segment and the load level selection rule in step 5, and the candidate three-dimensional trajectory recorded in the corresponding matrix element is read; the candidate three-dimensional trajectories of each flight segment are spliced ​​in the access order, and the duplicate task node waypoints at the connection of adjacent flight segments are deleted to form the candidate three-dimensional logistics trajectory corresponding to the current candidate task access sequence.

[0129] Step 6.2: Follow the segment order of the candidate three-dimensional logistics trajectory corresponding to the current candidate task access sequence. Starting from the first segment, perform feasibility condition verification on the current segment. Feasibility condition verification includes battery energy consumption verification, speed constraint verification, climb and descent rate verification, altitude range verification, no-fly zone verification, and communication reliability verification.

[0130] Specifically, following the segment order of the candidate 3D logistics trajectory corresponding to the current candidate task access sequence, starting from the first segment, the following six feasibility condition checks are performed sequentially on the current segment: (1) Battery energy consumption verification: Compare the cumulative power consumption of the current flight segment and all previous flight segments with the available battery energy. If the cumulative power consumption exceeds the available battery energy or the remaining power is lower than the safety limit, the current flight segment is determined not to meet the battery energy consumption constraint. If the cumulative power consumption does not exceed the available battery energy and the remaining power is not lower than the safety limit, the current flight segment is determined to meet the battery energy consumption constraint.

[0131] (2) Speed ​​constraint verification: The horizontal and vertical flight speeds of the current flight segment are compared with the preset speed allowable ranges. If the horizontal and / or vertical flight speeds exceed the allowable ranges, the current flight segment is determined to not meet the speed constraints.

[0132] (3) Climb and descent rate verification: Calculate the actual climb rate or descent rate based on the altitude difference and passage time of adjacent three-dimensional waypoints in the current flight segment, and compare it with the maximum climb rate or maximum descent rate of the UAV. If the actual value exceeds the maximum allowable value, it is determined that the current flight segment does not meet the climb and descent rate constraint.

[0133] (4) Altitude range verification: Check whether the altitude of all three-dimensional waypoints in the current flight segment is within the allowable flight altitude range. If any waypoint has an altitude that exceeds the allowable range, the flight segment is determined to not meet the altitude range constraint.

[0134] (5) No-fly zone verification: The three-dimensional spatial position of the current flight segment is spatially intersected with the boundary of the no-fly zone. If the current flight segment crosses or enters the no-fly zone, it is determined that the current flight segment does not meet the no-fly zone constraints.

[0135] (6) Communication Reliability Verification: The communication risk attribute value of the current flight segment is compared with the preset communication risk upper limit. If the communication risk attribute value of the current flight segment is greater than the communication risk upper limit, the current flight segment is determined not to meet the communication risk constraints. The communication risk attribute value includes the average communication risk value, the maximum communication risk value, and the duration of continuous risk exceeding the limit.

[0136] Step 6.3: If the current flight segment passes all feasibility condition checks, determine whether the current flight segment is the last flight segment. If so, use the current task access sequence and its corresponding candidate 3D logistics trajectory as the task access sequence and 3D logistics trajectory that meet the feasibility conditions. If not, set the next flight segment as the current flight segment and continue to perform feasibility condition checks.

[0137] If the current flight segment fails any of the feasibility condition checks, the current flight segment will be marked as a problem flight segment, and the corresponding correction strategy will be selected to correct the current flight segment according to the type of check that failed.

[0138] Specifically, when the current flight segment fails the battery energy consumption verification, the system first attempts to re-solve for a speed with better energy consumption while keeping the current access sequence and flight segment geometry unchanged. By adjusting the speed, the system reduces the horizontal flight power, thereby reducing the horizontal flight energy consumption of the flight segment. If the flight segment still does not meet the battery energy consumption constraint after speed optimization, the system further replaces the altitude layer or three-dimensional path of the high-energy-consumption flight segment, prioritizing the reduction of the cumulative climb altitude under heavy load conditions to reduce the energy consumption of the load climb. If the above local flight segment correction measures still cannot restore the feasibility of battery energy consumption, the candidate sequence is fed back to the mission access sequence optimization module, which adjusts the access order of the pick-up and delivery nodes or directly removes the candidate sequence.

[0139] When the current flight segment fails the speed constraint verification, the horizontal or vertical speed of the current flight segment is first reconfigured to the preset allowable range, and the adjusted transit time, energy consumption, and communication risk attributes of the current flight segment are recalculated simultaneously. If other constraints still cannot be met within the allowable speed range, the altitude layer of the current flight segment is further adjusted, the vertical switching process is extended, or the three-dimensional path of the problematic flight segment is regenerated to ensure that the overall feasibility of the flight segment is restored under the premise of meeting the speed constraints.

[0140] If the current flight segment fails the climb-descent rate verification, the vertical speed is reduced first, the vertical switching time is extended, or an intermediate altitude layer is inserted between adjacent altitude layers. If the corresponding cross-layer edge is still not executable under the allowed flight parameters, the cross-layer edge is deleted, and the problematic flight segment is re-performed with layered three-dimensional routing.

[0141] If the current flight segment fails the altitude range verification, and a non-task node exceeds the altitude limit, it will be adjusted to an adjacent feasible altitude layer; if a task node has multiple candidate arrival altitudes, a new combination of start and end altitudes that meets the task requirements will be selected; if the altitude adjustment conflicts with obstacles, communication reliability, or energy consumption constraints, the corresponding three-dimensional trajectory matrix elements of the task node will be regenerated.

[0142] If the current flight segment fails the no-fly zone check, the corresponding motion edge of the current flight segment is marked as impassable. Instead of simply adjusting the speed to retain the flight segment, the local three-dimensional rerouting is performed to bypass the no-fly zone while keeping the start and end relationships of the task nodes unchanged. If there is no feasible bypass flight segment, the corresponding matrix element is marked as unreachable, and the candidate access sequence is re-evaluated or replaced.

[0143] If the current flight segment fails the communication reliability check, it will be corrected by replacing the altitude layer, re-optimizing the speed, or using local three-dimensional rerouting.

[0144] When a current flight segment fails multiple checks simultaneously, spatial hard constraints such as no-fly zones and altitude ranges are addressed first, followed by flight performance constraints such as speed and climb / descent rates. Subsequently, communication risks and battery power consumption are jointly corrected within the feasible space. Each time an altitude layer is replaced, speed is reconfigured, an intermediate altitude layer is inserted, or a local 3D rerouting is completed, the 3D waypoint sequence of the corresponding flight segment in the current candidate 3D logistics trajectory is updated synchronously. The time, power consumption, remaining battery power, and communication risks of the affected flight segment and its subsequent segments are recalculated, and the "check-positioning-segment correction-segment check" iteration is repeated.

[0145] For example, when a flight segment simultaneously has both communication risk exceeding limits and energy consumption exceeding limits, the correction target for the problematic flight segment can be expressed as: ; in, Represents task node To the task node Candidate alternative routes between This indicates the overall cost of the alternative flight segment compared to the original flight segment. This indicates the weight of the penalty for exceeding communication risk limits. Indicates communication risks on alternative flight segments. This indicates the upper limit of permitted communication risks. This indicates the weight of the penalty for exceeding energy consumption limits. This indicates the estimated energy consumption of the alternative flight segment. This indicates the current remaining available energy. This indicates that only the portion exceeding the constraints will be penalized. This objective function is used to select the flight segment with the lower additional cost and constraint violation from candidate alternative flight segments when both communication risk and energy consumption exceed limits simultaneously.

[0146] For flight segments with high communication risks and available altitude levels, the altitude level replacement target can be represented as: ; in, Indicates the selected alternative height layer. This indicates that the candidate height layer belongs to the height layer set. Indicates flight segment At altitude The following communication risks Indicates switching to the height layer The cost of additional vertical maneuvering required, This represents the vertical maneuver cost weight. This formula is used to select an alternative altitude layer with a smaller overall communication risk and additional vertical maneuver cost without changing the mission access sequence.

[0147] Step 6.4: Start the feasibility condition verification from the current segment after the correction, and continue until all segments pass the full feasibility condition verification to obtain a three-dimensional logistics trajectory that meets the feasibility conditions.

[0148] Specifically, after correcting the current flight segment, the feasibility condition verification is re-executed starting from the corrected current flight segment. If the corrected flight segment and its subsequent flight segments pass all verifications, the process continues until all flight segments pass verifications. If any flight segments still fail verifications, the problematic flight segments are marked again and the corresponding correction operation is performed. This process is iterated until all flight segments pass all feasibility condition verifications, and the feasibility-satisfied task access sequence and its corresponding 3D logistics trajectory are output. In addition, if the current candidate task access sequence still has flight segments that fail feasibility condition verifications after reaching a preset number of iterations, it is determined that the sequence cannot be restored to feasibility under the current correction strategy, and it is marked as an infeasible sequence and removed. Subsequently, the next candidate task access sequence in the candidate task access sequence set is selected, and the segment-by-segment feasibility condition verification and correction operation is restarted until a task access sequence that satisfies all feasibility condition verifications is found, or no feasible task access sequence and its corresponding 3D logistics trajectory are found after traversing all candidate sequences.

[0149] Finally, the mission access sequence that meets the feasibility conditions and its corresponding 3D logistics trajectory are output as step 6. Optionally, the 3D logistics trajectory can be further converted into a flight plan that the UAV can directly execute, and output a set of command parameters consistent with the flight control interface. The output results include a 3D waypoint sequence, corresponding altitude layer sequence, speed configuration for each flight segment, climb / descent maneuvers, payload change records, communication risk annotations, and necessary alarm threshold configurations. This output result can be sent to the flight control system, mission management system, or ground station system for UAV actual execution and mission process monitoring.

[0150] Thus, this invention completes a full closed-loop process from task data acquisition, 3D space construction, communication risk modeling, trajectory matrix generation, access sequence optimization, energy consumption communication verification to trajectory output.

[0151] Compared to logistics route planning based on a two-dimensional distance matrix, this invention can directly incorporate the height difference of delivery points and the energy consumption of load climbing into the evaluation of the access sequence. For candidate routes with similar horizontal distances but different delivery heights, the system can identify routes with more heavy-load climbing and tends to select three-dimensional logistics trajectories with more reasonable energy consumption and lower execution risk.

[0152] This invention decouples the underlying 3D pathfinding results from the upper-layer fetching and access sequence optimization through a 3D trajectory matrix of task nodes. The underlying matrix is ​​responsible for recording the 3D waypoints and trajectory attributes between each pair of nodes, while the upper-layer optimization is responsible for combining the access order and verifying fetching and access constraints, thereby reducing redundant pathfinding calculations while maintaining the realism of the 3D space.

[0153] This invention, through load-related climb / descent energy consumption modeling, can distinguish between light-load segments, heavy-load segments, climb segments, and descent segments during the mission access sequence evaluation phase. This avoids the system generating too many climb actions under heavy load conditions, improves the accuracy of battery energy budget, and reduces the risks of energy consumption exceeding limits, infeasible speed configuration, and mission return midway.

[0154] This invention embeds a communication risk map into a three-dimensional trajectory matrix and a feasibility verification process, making communication reliability a constraint check item before the execution of the logistics trajectory. Unlike schemes that simply optimize the altitude profile around communication quality, this invention prioritizes maintaining the main lines of pick-up and delivery access sequences and load energy consumption evaluation, and uses altitude adjustments or rerouting to reduce communication risks on local segments.

[0155] This invention can be applied to scenarios such as rooftop delivery, multi-building delivery in industrial parks, campus logistics, material transfer in factories, and emergency material delivery. These scenarios typically feature relatively concentrated pickup points, highly dispersed delivery points, communication coverage affected by buildings, and significant energy consumption during heavy-load climbing. This invention can output an executable three-dimensional logistics trajectory that includes waypoint sequences, altitude layer sequences, speed configurations, load changes, and communication risk annotations.

[0156] Example 2

[0157] Please see Figure 2 , Figure 2 This is a schematic diagram of a hierarchical three-dimensional UAV logistics trajectory planning system provided by the present invention. Based on Embodiment 1, the present invention further provides a hierarchical three-dimensional UAV logistics trajectory planning system, which includes: The data acquisition module is used to acquire logistics task information in the task area. The logistics task information includes task node information and environmental constraint information. The task node information includes a set of task nodes consisting of take-off and landing points, several pickup nodes, and several delivery nodes. The layered directed graph construction module is used to discretize the task region in the height direction and construct a layered 3D directed graph; The communication risk modeling module is used to construct a three-dimensional communication risk map based on the communication risk values ​​of different spatial location units in the task area. The trajectory matrix generation module is used to perform a three-dimensional path search in a layered three-dimensional directed graph to obtain candidate three-dimensional trajectories between task nodes, so as to generate trajectory attribute matrices between task nodes corresponding to multiple discrete load levels based on the three-dimensional communication risk graph. The access sequence optimization module is used to generate a set of candidate task access sequences based on the trajectory attribute matrix between task nodes corresponding to multiple discrete load levels, under the condition of satisfying task execution constraints. The verification and correction module is used to generate candidate 3D logistics trajectories based on the candidate task access sequence set, and to perform feasibility condition verification and correction operations on each segment of the candidate 3D logistics trajectory to obtain a 3D logistics trajectory that meets the feasibility conditions.

[0158] The implementation method and principle of the hierarchical three-dimensional UAV logistics trajectory planning system provided in this embodiment are similar to those of the hierarchical three-dimensional UAV logistics trajectory planning method provided in Embodiment 1, and will not be repeated here.

[0159] In the description of this invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0160] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, any modifications made without departing from the inventive concept should be considered within the scope of protection of the present invention.

Claims

1. A hierarchical three-dimensional UAV logistics trajectory planning method, characterized in that, include: Obtain logistics task information for the task area. The logistics task information includes task node information and environmental constraint information. The task node information includes a set of task nodes consisting of a take-off and landing point, several pickup nodes, and several delivery nodes. The task region is discretized in the height direction to construct a hierarchical three-dimensional directed graph; A three-dimensional communication risk map is constructed based on the communication risk values ​​of different spatial location units in the task area; A three-dimensional path search is performed in the layered three-dimensional directed graph to obtain candidate three-dimensional trajectories between task nodes, so as to generate a trajectory attribute matrix between task nodes corresponding to multiple discrete load levels based on the three-dimensional communication risk graph. Based on the trajectory attribute matrix between task nodes corresponding to multiple discrete load levels, a set of candidate task access sequences is generated under the condition of satisfying task execution constraints. Candidate 3D logistics trajectories are generated based on the candidate task access sequence set, and feasibility condition verification and correction operations are performed on each segment of the candidate 3D logistics trajectory to obtain a 3D logistics trajectory that meets the feasibility conditions.

2. The hierarchical three-dimensional UAV logistics trajectory planning method according to claim 1, characterized in that, Each task node in the task node set has a three-dimensional spatial location attribute. Each pickup node and its corresponding delivery node in the task node set constitute a pickup-delivery node pair, and the access order of each pair of pickup-delivery nodes is the pickup node and the delivery node in sequence. The environmental constraint information includes at least one of the following: no-fly zone distribution, obstacle distribution, flight altitude range, horizontal speed range, vertical speed range, maximum rate of climb, maximum rate of descent, communication station location, and communication reliability threshold.

3. The hierarchical three-dimensional UAV logistics trajectory planning method according to claim 1, characterized in that, The task region is discretized along the height direction to construct a hierarchical three-dimensional directed graph, including: The task region is discretized into multiple discrete height layers in the height direction; A horizontal motion network is established within each of the discrete height layers, wherein the horizontal moving edges in the horizontal motion network are used to connect different horizontal positions within the same discrete height layer; A vertical switching edge is established between different discrete height layers, wherein the vertical switching edge is used to connect the same horizontal position or adjacent horizontal positions between different discrete height layers; Feasibility screening is performed on the horizontal motion network and the vertical switching edge, and horizontal movement edges and vertical switching edges that do not meet the preset conditions are eliminated to construct a hierarchical three-dimensional directed graph. The hierarchical three-dimensional directed graph includes several discrete height layers, passable horizontal movement edges in each discrete height layer, and passable vertical switching edges connecting different discrete height layers.

4. The hierarchical three-dimensional UAV logistics trajectory planning method according to claim 1, characterized in that, Based on the communication risk values ​​of different spatial location units within the task area, a three-dimensional communication risk map is constructed, including: The task area is divided into several spatial location units; Based on at least one data source among the communication propagation model, historical link measurement data, signal-to-interference-plus-noise ratio prediction results, and real-time link quality data, a communication risk value is determined for each spatial location unit, wherein the communication risk value is used to characterize the possibility that the communication quality within the spatial location unit does not meet the preset reliability requirements; The communication risk value of each spatial location unit is mapped to the corresponding spatial location to form a three-dimensional communication risk map that varies with horizontal position and height.

5. The hierarchical three-dimensional UAV logistics trajectory planning method according to claim 4, characterized in that, Based on at least one data source selected from communication propagation models, historical link measurement data, signal-to-interference-plus-noise ratio (SIR) prediction results, and real-time link quality data, the communication risk value of each spatial location unit is determined, including: If only one of the following data sources exists within the current spatial location unit: the communication propagation model, the historical link measurement data, the signal-to-interference-plus-noise ratio prediction result, and the real-time link quality data, then the communication risk value of the current spatial location unit is determined based on the data source. If at least two of the following data sources exist within the current spatial location unit: the communication propagation model, the historical link measurement data, the signal-to-interference-plus-noise ratio prediction result, and the real-time link quality data, then the communication risk values ​​calculated from each of the data sources are weighted and fused to obtain the communication risk value of the current spatial location unit. If no data source exists for the current spatial location unit, including the communication propagation model, the historical link measurement data, the signal-to-interference-plus-noise ratio prediction result, and the real-time link quality data, then the communication risk value of the current spatial location unit is obtained through spatial interpolation.

6. The hierarchical three-dimensional UAV logistics trajectory planning method according to claim 1, characterized in that, A 3D path search is performed in the hierarchical 3D directed graph to obtain candidate 3D trajectories between task nodes. Based on the 3D communication risk graph, a trajectory attribute matrix is ​​generated between task nodes corresponding to multiple discrete load levels, including: The payload range of the drone from empty to maximum permissible payload is discretized into multiple payload levels, each payload level has a corresponding preset payload value, and the preset payload value increases with the payload level. For each discrete load level, each task node in the task node set is traversed as the starting task node, and each task node other than the starting task node is traversed as the ending task node. The starting task node and the ending task node are mapped to the layered three-dimensional directed graph to determine the source point and the sink point. Based on the source point and the sink point, a three-dimensional path search is performed in the layered three-dimensional directed graph to obtain candidate three-dimensional trajectories that meet preset constraints. For each discrete load level, the cumulative communication risk value of the candidate three-dimensional trajectory is calculated using the three-dimensional communication risk map. Based on the cumulative communication risk value, track length, and energy consumption of the candidate three-dimensional trajectory, a trajectory attribute matrix between task nodes is constructed. Each preset load value corresponds to one trajectory attribute matrix, and the trajectory attribute matrix is ​​used to characterize the three-dimensional track attributes between task nodes.

7. The hierarchical three-dimensional UAV logistics trajectory planning method according to claim 1, characterized in that, Based on the trajectory attribute matrix between task nodes corresponding to multiple discrete load levels, a set of candidate task access sequences is generated under the condition of satisfying task execution constraints, including: The first task access sequence is generated based on the take-off and landing point location, pick-up and delivery pairing relationship, task node height attribute, and maximum load constraint; Based on the actual load of each segment in the first mission access sequence, the corresponding trajectory attribute matrix is ​​determined to perform preliminary feasibility screening of the first mission access sequence. If the first mission access sequence meets the mission execution constraints, the first mission access sequence is used as the initial optimal mission access sequence and added to the elite sequence set. If the first mission access sequence does not meet the mission execution constraints, the first mission access sequence is regenerated and preliminary feasibility screening is performed again until a first mission access sequence that meets the mission execution constraints is obtained. Using the current first task access sequence as the initial optimal task access sequence, an improved tabu search algorithm is used for iterative optimization; wherein, in each iteration, the following steps are performed: Perform a local perturbation operation on the current task access sequence to generate a second task access sequence; Based on the actual load of each segment in the second task access sequence, the corresponding trajectory attribute matrix is ​​called to check whether the second task access sequence meets the task execution constraints. If it does not meet the constraints, the second task access sequence is discarded, and the current task access sequence and the current optimal task access sequence remain unchanged. If it meets the constraints, the second task access sequence is checked to see if it is prohibited by the taboo list. Update the current optimal task access sequence based on the tabu check results and comprehensive evaluation value of the second task access sequence; During each iteration, the object to be subjected to local perturbation operations is added to the taboo list, and the taboo duration of the taboo list is updated. Determine whether the preset iteration termination condition is met. If it is met, sort the task access sequences in the elite sequence set in ascending order of comprehensive evaluation value and output the sorted candidate task access sequence set. If it is not met, execute the next iteration.

8. The hierarchical three-dimensional UAV logistics trajectory planning method according to claim 7, characterized in that, Based on the tabu check results and comprehensive evaluation value of the second task access sequence, update the current optimal task access sequence, including: If the second task access sequence is not prohibited, the comprehensive evaluation value of the second task access sequence is obtained by using the trajectory attributes between each adjacent task node in the second task access sequence, and the second task access sequence is used as the new current task access sequence. If the comprehensive evaluation value of the second task access sequence is less than the comprehensive evaluation value of the current best task access sequence, the second task access sequence is updated to the new current best task access sequence, and the second task access sequence is added to the elite sequence set. If the second task access sequence is prohibited, the comprehensive evaluation value of the second task access sequence is obtained by using the trajectory attributes between each adjacent task node in the second task access sequence. It is then determined whether the comprehensive evaluation value of the second task access sequence is less than the comprehensive evaluation value of the current optimal task access sequence. If so, the second task access sequence is allowed to break the taboo restriction, and the second task access sequence is used as the new current task access sequence and the new current optimal task access sequence. The second task access sequence is then added to the elite sequence set. If not, the second task access sequence is discarded, and the current task access sequence and the current optimal task access sequence remain unchanged.

9. The hierarchical three-dimensional UAV logistics trajectory planning method according to claim 1, characterized in that, Candidate 3D logistics trajectories are generated based on the candidate task access sequence set, and feasibility condition verification and correction operations are performed on each segment of the candidate 3D logistics trajectories to obtain 3D logistics trajectories that meet the feasibility conditions, including: The candidate 3D trajectories of each flight segment corresponding to the current candidate task access sequence are stitched together to form a candidate 3D logistics trajectory. Following the segment order of the candidate three-dimensional logistics trajectory corresponding to the current candidate task access sequence, starting from the first segment, the feasibility condition verification of the current segment is performed. The feasibility condition verification includes battery energy consumption verification, speed constraint verification, climb and descent rate verification, altitude range verification, no-fly zone verification, and communication reliability verification. If the current flight segment passes all feasibility condition checks, then determine whether the current flight segment is the last flight segment. If so, then the current task access sequence and its corresponding candidate 3D logistics trajectory are taken as the task access sequence and 3D logistics trajectory that meet the feasibility conditions. If not, then the next flight segment is set as the current flight segment, and the feasibility condition check continues. If the current flight segment fails any of the feasibility condition checks, the current flight segment will be marked as a problem flight segment, and the corresponding correction strategy will be selected according to the type of failure to correct the current flight segment, resulting in the corrected current flight segment. Starting from the corrected current route segment, the feasibility condition verification will be re-executed until all routes pass the full feasibility condition verification, resulting in a three-dimensional logistics trajectory that meets the feasibility conditions.

10. A hierarchical three-dimensional unmanned aerial vehicle (UAV) logistics trajectory planning system, characterized in that, include: The data acquisition module is used to acquire logistics task information in the task area. The logistics task information includes task node information and environmental constraint information. The task node information includes a set of task nodes consisting of a take-off and landing point, several pickup nodes, and several delivery nodes. A hierarchical directed graph construction module is used to discretize the task region in the height direction and construct a hierarchical three-dimensional directed graph; The communication risk modeling module is used to construct a three-dimensional communication risk map based on the communication risk values ​​of different spatial location units in the task area. The trajectory matrix generation module is used to perform a three-dimensional path search in the layered three-dimensional directed graph to obtain candidate three-dimensional trajectories between task nodes, so as to generate a trajectory attribute matrix between task nodes corresponding to multiple discrete load levels based on the three-dimensional communication risk graph. The access sequence optimization module is used to generate a set of candidate task access sequences based on the trajectory attribute matrix between task nodes corresponding to multiple discrete load levels, under the condition of satisfying task execution constraints. The verification and correction module is used to generate candidate three-dimensional logistics trajectories based on the candidate task access sequence in the candidate task access sequence set, and to perform feasibility condition verification and correction operations on the candidate three-dimensional logistics trajectories segment by segment to obtain three-dimensional logistics trajectories that meet the feasibility conditions.