Unmanned delivery method and device based on dry and branch end fusion

Abstract

The application provides an unmanned distribution method and device based on dry-branch-end fusion, and relates to the technical field of intelligent logistics.The method comprises the following steps: collecting distribution data of an unmanned distribution task; inputting the distribution data into a path planning model to obtain a global distribution path output by the path planning model; and controlling an unmanned device to execute the unmanned distribution task based on the global distribution path.The unmanned distribution method and device based on dry-branch-end fusion provided by the application convert the planned global optimal path into executable instructions to drive the unmanned device to work cooperatively, thereby realizing unmanned distribution in a complex and changeable city, reducing distribution cost, and improving distribution safety and distribution efficiency.

Description

Technical Field

[0001] This invention relates to the field of intelligent logistics technology, and in particular to an unmanned delivery method and device based on the integration of trunk, branch, and terminal logistics. Background Technology

[0002] Goods delivery usually relies on couriers. In long-distance delivery, relying on couriers is labor-intensive, inefficient, and poses high safety risks.

[0003] With the rapid development of artificial intelligence, existing technologies have gradually adopted driverless vehicles for unmanned delivery, mainly through a combination of driverless vehicles and human labor, to complete long-distance transportation on open roads. However, because driverless vehicles, motor vehicles, non-motor vehicles, and pedestrians share road rights, the complexity of road conditions is high, leading to higher traffic safety risks. At the same time, driverless vehicles exacerbate urban traffic congestion by occupying road resources. The delivery process is also constrained by traffic lights, congestion, speed limits, etc., resulting in uncontrollable long-distance delivery times. Furthermore, a large amount of human labor is still required for maintenance or last-mile delivery, resulting in high implementation costs and the inability to achieve a truly end-to-end unmanned closed loop. In the face of complex and ever-changing urban scenarios, delivery efficiency is difficult to guarantee. Summary of the Invention

[0004] This invention provides an unmanned delivery method and device based on trunk-branch-terminal integration, which solves the technical problems of high delivery cost, high safety risk and low efficiency in long-distance delivery in complex and ever-changing urban environments.

[0005] This invention provides an unmanned delivery method based on the fusion of zoning and local time, comprising the following steps: The system collects delivery data for unmanned delivery missions; the delivery data includes time-series data, spatial structure data, and static attribute data. The delivery data is input into a path planning model to obtain a global delivery path output by the model. The path planning model includes a temporal feature module, a topological feature module, a static feature module, a mapping module, and a global strategy module. The temporal feature module is used to extract temporal correlation features based on the temporal sequence data. The topological feature module is used to extract spatial topological features based on the spatial structure data. The static feature module is used to extract static features based on the static attribute data. The mapping module is used to map the temporal correlation features, the spatial topological features, and the static features into a spatiotemporal feature vector. The global strategy module is used to output the global delivery path based on the spatiotemporal feature vector. Based on the global delivery path, control the unmanned equipment to execute the unmanned delivery task.

[0006] According to the present invention, an unmanned delivery method based on stem-branch fusion is provided, wherein the time-series data includes at least one of the following: Current operational data for public transportation; Historical passenger flow data for the public transportation vehicles; Current meteorological data; Current road congestion data; Historical traffic accident data; Future weather data for a preset time period; The location data of the unmanned equipment.

[0007] According to the present invention, an unmanned delivery method based on the fusion of trunk and branch lines is provided, wherein the spatial structure data includes at least one of the following: Linked data between public transportation stations; The association data between the starting point of the unmanned delivery mission and the public transportation station; The association data between the destination of the unmanned delivery mission and the public transportation station.

[0008] According to the present invention, an unmanned delivery method based on the fusion of cyclical and end-of-cycle data is provided, wherein the static attribute data includes at least one of the following: The starting address data of the unmanned delivery mission; The destination address data of the unmanned delivery mission; The order information corresponding to the unmanned delivery task; The carrying parameters of the unmanned equipment; Basic attribute data of public transportation stations.

[0009] According to the present invention, an unmanned delivery method based on the fusion of trunk and branch lines is provided, wherein the global strategy module is specifically used for: Based on the improved spatiotemporal aligned attention operator and the spatiotemporal feature vector, scores corresponding to multiple potential delivery paths are calculated respectively; Based on the score, the global delivery route is determined from the potential delivery routes.

[0010] According to the present invention, an unmanned delivery method based on trunk-branch-terminal fusion is provided, wherein controlling unmanned equipment to execute the unmanned delivery task based on the global delivery path includes: Based on the near-end strategy optimization algorithm and the global delivery path, the unmanned delivery task is decomposed into multiple delivery sub-tasks; The unmanned equipment is controlled to execute the delivery sub-tasks sequentially.

[0011] The present invention also provides an unmanned delivery device based on the fusion of tectonic and end-of-line systems, comprising the following modules: The data acquisition module is used to collect delivery data for unmanned delivery tasks; the delivery data includes time-series data, spatial structure data, and static attribute data. A route planning module is used to input the delivery data into a route planning model to obtain a global delivery path output by the route planning model. The route planning model includes a temporal feature module, a topological feature module, a static feature module, a mapping module, and a global strategy module. The temporal feature module is used to extract temporal correlation features based on the temporal sequence data. The topological feature module is used to extract spatial topological features based on the spatial structure data. The static feature module is used to extract static features based on the static attribute data. The mapping module is used to map the temporal correlation features, the spatial topological features, and the static features into a spatiotemporal feature vector. The global strategy module is used to output the global delivery path based on the spatiotemporal feature vector. The unmanned delivery module is used to control unmanned equipment to perform the unmanned delivery task based on the global delivery path.

[0012] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the unmanned delivery method based on trunk-branch fusion as described above.

[0013] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the unmanned delivery method based on trunk-branch fusion as described above.

[0014] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the unmanned delivery method based on trunk-branch fusion as described above.

[0015] This invention provides an unmanned delivery method and apparatus based on the fusion of time, space, and static attributes. It collects delivery data from unmanned delivery tasks, including time-series data, spatial structure data, and static attribute data. This comprehensive collection of multi-dimensional delivery data provides complete and accurate input information for subsequent path planning. The delivery data is then input into a path planning model to obtain the global delivery path output by the model. The path planning model includes a time-series feature module, a topology feature module, a static feature module, a mapping module, and a global strategy module. The time-series feature module extracts time-series correlation features from the time-series data; the topology feature module extracts spatial topology features from the spatial structure data; the static feature module extracts static features from the static attribute data; and the mapping module maps the time-series correlation features to the static attribute data. The system maps temporal and spatial topological features and static features into spatiotemporal feature vectors. A global strategy module outputs a global delivery path based on these spatiotemporal feature vectors. This path planning model, integrating temporal feature modules, topological feature modules, static feature modules, and mapping and global strategy modules, performs deep feature extraction and fusion on multi-source heterogeneous data. Ultimately, it generates the optimal global delivery path under complex spatiotemporal constraints, improving the reliability and efficiency of long-distance unmanned delivery. Based on the global delivery path, unmanned equipment is controlled to execute unmanned delivery tasks, transforming the planned optimal path into executable instructions that drive collaborative operation of unmanned equipment. This translates the algorithmic optimization results into precise and automated delivery actions in the physical world, enabling unmanned delivery in complex and ever-changing urban environments, reducing delivery costs, and improving safety and efficiency. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating the unmanned delivery method based on the fusion of trunk and branch lines provided by the present invention.

[0018] Figure 2 This is a schematic diagram of the unmanned delivery device based on the fusion of the main trunk and branch trunk provided by the present invention.

[0019] Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0021] Current technologies employ unmanned vehicles for unmanned delivery, primarily combining unmanned vehicles with human labor to complete long-distance transportation on open roads. However, this approach suffers from several problems: unmanned vehicles, motor vehicles, non-motorized vehicles, and pedestrians share road rights, facing challenges from complex traffic flow interference and emergency response; human delivery also carries the risk of traffic accidents, placing continuous pressure on public safety throughout the entire delivery process; a large number of ground transportation equipment and delivery personnel occupy urban road resources, especially during peak hours and in core areas, where delivery flow overlaps with commuting traffic, further increasing the road network load, and delivery time is significantly affected by road condition fluctuations; ground transportation is constrained by factors such as traffic lights, traffic control, road congestion, and speed limits for unmanned vehicles, making long-distance transportation time uncontrollable; complex scenarios significantly increase time losses, making it difficult to optimize end-to-end delivery cycles; some road sections still require human labor for delivery, and long-distance unmanned vehicle highway delivery requires maintenance personnel, resulting in consistently high labor costs in large-scale operations, preventing the achievement of full-chain unmanned operation and limiting the cost optimization and scale expansion of the delivery network.

[0022] To address these issues, this invention proposes an unmanned delivery method and device based on the fusion of trunk and branch lines, which solves the technical problems of high delivery costs, high safety risks, and low efficiency in long-distance goods delivery in complex and ever-changing cities.

[0023] The following is combined with Figures 1 to 3 The present invention describes an unmanned delivery method and apparatus based on the fusion of the main stem and branch lines.

[0024] Figure 1 This is a flowchart illustrating the unmanned delivery method based on the fusion of trunk and branch lines provided by the present invention, as shown below. Figure 1 As shown, the method includes the following steps: Step 101: Collect delivery data for unmanned delivery tasks; the delivery data includes time-series data, spatial structure data, and static attribute data; Specifically, delivery data for unmanned delivery tasks is collected from multiple sources. For example, user-end devices can collect order information (such as the sender / pickup person's detailed address, departure time, etc.); the operation and maintenance end can collect equipment battery level, transport capacity, whether there is a fault, current location, etc.; the interface end can collect vehicle type, vehicle transport capacity, arrival and departure time, whether the vehicle / site is in use, vehicle / site location, etc.

[0025] According to preset rules, delivery data can be divided into time-series data, spatial structure data, and static attribute data. Time-series data reflects information that changes over time, spatial structure data reflects information about physical location and connectivity, and static attribute data reflects information about fixed or semi-fixed attributes. These data collectively constitute a digital description of the dynamic environment in which the unmanned delivery task takes place.

[0026] Furthermore, the time-series data includes at least one of the following: Current operational data for public transportation; Historical passenger flow data for the public transportation vehicles; Current meteorological data; Current road congestion data; Historical traffic accident data; Future weather data for a preset time period; The location data of the unmanned equipment.

[0027] Specifically, the time-series data includes one or more of the following: current operational data of public transportation vehicles, historical passenger flow data of public transportation vehicles, current meteorological data (such as weather forecasts for the areas they pass through), current road congestion data, historical traffic accident data, future meteorological data for a preset time period, and location data of unmanned equipment. Public transportation vehicles can be city buses, city subways, or high-speed trains, etc.

[0028] This invention provides richer dynamic environmental information for long-distance path planning in unmanned delivery tasks by collecting time-series data, thereby improving the reliability of unmanned delivery and ensuring delivery quality.

[0029] Furthermore, the spatial structure data includes at least one of the following: Linked data between public transportation stations; The association data between the starting point of the unmanned delivery mission and the public transportation station; The association data between the destination of the unmanned delivery mission and the public transportation station.

[0030] Specifically, the spatial structure data includes correlation data between public transportation stations, correlation data between the starting point and public transportation stations in unmanned delivery tasks, and correlation data between the destination and public transportation stations in unmanned delivery tasks.

[0031] For example, in long-distance unmanned delivery tasks that require controlling unmanned equipment to take public transportation, it is not only necessary to have the correlation data between bus stops, but also to clarify the walking / cycling path topology between the starting point (such as a community express locker) and the nearest bus stop, as well as the correlation between the destination (such as an office building) and the target bus stop, so as to achieve end-to-end unmanned delivery.

[0032] This invention, through the collection of spatial structure data and the detailed modeling of spatial connections, ensures the physical feasibility of the planned path, achieves precise delivery, and improves delivery efficiency and user experience.

[0033] Furthermore, the static attribute data includes at least one of the following: The starting address data of the unmanned delivery mission; The destination address data of the unmanned delivery mission; The order information corresponding to the unmanned delivery task; The carrying parameters of the unmanned equipment; Basic attribute data of public transportation stations.

[0034] Specifically, static attribute data includes one or more of the following: the starting address data of the unmanned delivery mission, the ending address data of the unmanned delivery mission, the order information corresponding to the unmanned delivery mission, the carrying parameters of the unmanned equipment, and the basic attribute data of public transportation stations. Among these, order information includes one or more of the following: delivery time, category of delivered goods, weight of delivered goods, and volume of delivered goods; carrying parameters include one or more of the following: rated load capacity, range, equipment size, and operating speed; and basic attribute data includes inherent attributes such as the type of station (e.g., bus station, subway station, or high-speed rail station), carrying capacity, whether there are charging facilities, and whether it is an indoor station.

[0035] For example, for a medical reagent order requiring cold chain delivery, the order information includes "temperature control at 2-8℃ throughout the process." The intelligent operation cabin (i.e., the unmanned equipment undertaking the unmanned delivery task) specifies in its transport parameters that its cold chain module has a 4-hour battery life. Basic attribute data from public transportation stations shows that subway station A has charging facilities, while bus station B does not. Therefore, the route with charging facilities and a travel time of less than 4 hours is prioritized, ensuring that the delivery is completed using the intelligent operation cabin with the cold chain module.

[0036] This invention collects static attribute data and incorporates rich static constraints into decision-making, thereby meeting customizable personalized delivery needs (such as cold chain and high-value items), and reasonably matching equipment capabilities with task requirements, optimizing resource allocation, and reducing overall delivery costs while ensuring delivery quality.

[0037] Based on the above embodiments, the present invention integrates multi-source heterogeneous data from transportation systems, users, and equipment to collect time-series data, spatial structure data, and static attribute data of unmanned delivery tasks, providing a data foundation for generating a long-distance delivery route that fully utilizes a mature public transportation network and adapts to a dynamic environment.

[0038] Step 102: Input the delivery data into the path planning model to obtain the global delivery path output by the path planning model; the path planning model includes a temporal feature module, a topological feature module, a static feature module, a mapping module, and a global strategy module; the temporal feature module is used to extract temporal correlation features based on the temporal sequence data; the topological feature module is used to extract spatial topological features based on the spatial structure data; the static feature module is used to extract static features based on the static attribute data; the mapping module is used to map the temporal correlation features, the spatial topological features, and the static features into a spatiotemporal feature vector; the global strategy module is used to output the global delivery path based on the spatiotemporal feature vector; Specifically, the path planning model mines deep features in the data through three parallel feature extraction branches (i.e., sequential feature module, topological feature module, and static feature module); then, the mapping module merges these features into a unified spatiotemporal feature vector; finally, the global strategy module searches for and outputs a globally optimal delivery route that comprehensively considers time, space, and static constraints in the complex spatiotemporal network composed of "trunk lines + branch lines + terminal lines" based on this spatiotemporal feature vector, thus achieving efficient and low-cost unmanned delivery based on the fusion of trunk, branch, and terminal lines.

[0039] The temporal feature module uses a Transformer encoder to process temporal sequence data and extract temporal correlation features; the topological feature module uses a graph neural network (GNN) to process spatial structure data and extract spatial topological features based on road network topology relationships and station correlation data in the spatial structure data; the static feature module uses a multilayer perceptron (MLP) to process static attribute data and extract static features; the mapping module first maps the temporal correlation features, spatial topological features, and static features to a high-dimensional feature space of the same dimension, then adds sinusoidal position encoding to fuse timestamp information, performs vector concatenation, and finally outputs a standardized spatiotemporal feature vector; the global strategy module is used to output a global delivery path based on the spatiotemporal feature vector.

[0040] For example, in one embodiment, relying on public transportation networks such as subways, high-speed rail, and buses, an intelligent operation cabin is seamlessly connected to transport equipment such as outdoor transport platforms, multi-scenario adaptable Automated Guided Vehicles (AGVs), and auxiliary cargo drones. This achieves efficient collaborative transportation of "trunk lines + branch lines + last-mile delivery," and can be viewed as an urban / intercity logistics network consisting of warehouses, logistics stations, enterprise parks, and residential areas as leaf nodes, and public transportation stations (such as subway stations, bus stations, and high-speed rail stations) as branch nodes. The trunk line refers to the transport path between branch nodes, using multi-scenario adaptable AGVs to ride public transportation facilities and transport goods through the trunk public transportation network; the branch line refers to the transport path between branch nodes and leaf nodes, using outdoor transport platforms to achieve short-distance connections for the last 1km; and the last-mile delivery refers to the transport path within a leaf node, using multi-scenario adaptable AGVs or auxiliary cargo drones to achieve precise delivery of goods to the customer's door.

[0041] The route planning model performs a global analysis based on delivery data and outputs a global delivery route. This global delivery route takes into account the off-peak time windows of various public transportation tools or transport equipment, making use of idle transport capacity for cargo transportation, achieving both passenger and freight transport, reducing the burden on ground traffic, and improving urban logistics efficiency.

[0042] For example, in one embodiment, after a sender places an order on the user's end, the delivery system collects: 1) time-series data, such as real-time arrival and departure times of subway trains and historical passenger flow data; 2) spatial structure data, such as the road network topology and station relationships between sender points, subway stations, and receiver points; and 3) static attribute data, such as sender / recipient addresses, order weight and volume, and cold chain temperature control requirements of the intelligent operating cabin. Based on the collected data, a spatiotemporal graph model is created (e.g., modeling subway stations and AGVs as nodes with time windows). The time-series feature module analyzes subway train frequency patterns, the topology feature module understands the subway network structure, and the static feature module encodes order attributes. After these features are mapped into spatiotemporal feature vectors by the mapping module, the global strategy module calculates the optimal global delivery path from the sender point to the receiver point based on the spatiotemporal feature vectors. For example, a multi-scenario adapted AGV (a robot that can autonomously navigate indoors and in a vehicle, such as a navigation robot based on Simultaneous Localization and Mapping (SLAM)) picks up the goods at point A, transports them to subway station X via an outdoor transport platform (an unmanned vehicle suitable for outdoor roads, such as an unmanned delivery vehicle), then takes the intelligent work cabin on subway line M, is connected to the outdoor transport platform after exiting the station, and finally completes the last-mile delivery by a multi-scenario adapted AGV or an auxiliary cargo drone.

[0043] Based on the above embodiments, and addressing the multi-objective optimization requirements of this invention, a total loss function for the path planning model in pre-supervised training is also designed for updating network weights via deep learning backpropagation, achieving a balance between timeliness, energy consumption, and reliability. The expression is as follows: In the formula, This represents the total loss function, and the training objective is to minimize this value; , , These are all weighting coefficients, which can be dynamically adjusted according to business needs (such as prioritizing cold chain delivery time and prioritizing cost for general parts); Indicates the total delivery time across the entire supply chain; This represents the total energy consumption of all transportation equipment across the entire supply chain. This represents the connection risk entropy, which is predicted by the path planning model based on historical data and real-time status, and is used to quantify the failure probability of the connection process.

[0044] This invention uses a path planning model to perform global analysis of delivery data, dynamically adapting to different trunk networks such as subways, high-speed rail, and buses, and planning collaborative paths that integrate multiple transportation devices. This effectively improves transportation efficiency, reduces labor costs, enhances delivery safety, and reduces unnecessary travel by optimizing paths, thus contributing to energy conservation and emission reduction and further reducing delivery costs.

[0045] Step 103: Based on the global delivery path, control the unmanned equipment to execute the unmanned delivery task.

[0046] Specifically, the abstract global delivery path is decomposed into a series of atomic tasks or hardware instructions that can be understood and executed by unmanned equipment (such as AGVs, drones, and outdoor transport platforms), and these devices are coordinated and controlled to complete operations such as docking, moving, loading, and unloading in sequence and collaboratively, ultimately achieving end-to-end unmanned delivery.

[0047] This invention provides an unmanned delivery method based on the fusion of time, space, and static attributes. It collects delivery data from unmanned delivery tasks, including time-series data, spatial structure data, and static attribute data. This comprehensive collection of multi-dimensional delivery data provides complete and accurate input information for subsequent path planning. The delivery data is then input into a path planning model to obtain the global delivery path output by the model. The path planning model includes a time-series feature module, a topology feature module, a static feature module, a mapping module, and a global strategy module. The time-series feature module extracts time-series correlation features from the time-series data; the topology feature module extracts spatial topology features from the spatial structure data; the static feature module extracts static features from the static attribute data; and the mapping module maps the time-series correlation features... The system maps spatial topological features and static features into spatiotemporal feature vectors. A global strategy module outputs a global delivery path based on these spatiotemporal feature vectors. This path planning model, integrating temporal feature modules, topological feature modules, static feature modules, and mapping and global strategy modules, performs deep feature extraction and fusion on multi-source heterogeneous data. Ultimately, it generates the optimal global delivery path under complex spatiotemporal constraints, improving the reliability and efficiency of long-distance unmanned delivery. Based on the global delivery path, unmanned equipment is controlled to execute unmanned delivery tasks, transforming the planned optimal path into executable instructions. This drives collaborative operation of unmanned equipment, translating the algorithmic optimization results into precise and automated delivery actions in the physical world. This enables unmanned delivery in complex and ever-changing urban environments, reducing delivery costs and improving safety and efficiency.

[0048] Furthermore, the global policy module is specifically used for: Based on the improved spatiotemporal aligned attention operator and the spatiotemporal feature vector, scores corresponding to multiple potential delivery paths are calculated respectively; Based on the score, the global delivery route is determined from the potential delivery routes.

[0049] Specifically, the global strategy module is an improved Transformer encoder, which replaces the original Transformer attention operator with a spatiotemporally aligned attention operator. This operator is used to capture spatial topological associations and temporal window matching relationships based on spatiotemporal feature vectors. The expression for the spatiotemporally aligned attention operator is as follows: In the formula, the parameters are defined as follows: Indicates the potential path score. This represents the query vector for the current node i; The key vector representing the target node j; , The weight matrix is ​​trainable. The feature dimension is used for scaling to avoid gradient explosion; The function representing the hard constraint penalty for spatiotemporal feasibility is when... <0 (which could indicate that the unmanned device missed the departure time of public transportation) or When ≤0 (which can indicate that public transportation has no remaining capacity), If the value approaches 0, the attention score of that node pair is forced to zero, eliminating infeasible delivery routes at the source; when the time window matches and capacity is sufficient, =1, retain the complete feature relevance score; 、 These are time sensitivity coefficient and space cost coefficient, respectively, which can be adjusted according to the needs of the scenario; This represents the difference between the arrival time of unmanned equipment and the departure time of public transportation, used to penalize delivery routes with mismatched times; This represents the physical travel distance between nodes, used to balance the transportation costs along the route.

[0050] The global strategy module calculates the scores corresponding to multiple potential delivery paths based on the input spatiotemporal feature vectors, and then selects the potential delivery path with the highest score as the global delivery path.

[0051] For example, in one embodiment, if it is necessary to calculate the score of the potential delivery path from the current node i (such as an outdoor delivery platform that has just arrived at a bus stop) to the next target node j (such as a bus that is about to leave the station), the spatiotemporal alignment attention operator performs the following calculation: if the platform arrival time is later than the bus departure time ( If the value is less than 0, the spatiotemporal feasibility function Φ outputs close to 0, forcing the attention score of the path to be zero, thus avoiding the selection of invalid paths that "miss the train" in the early stages of calculation.

[0052] This invention improves the efficiency and reliability of path planning by designing a spatiotemporally aligned attention operator, ensuring that the final output global delivery path is not only mathematically optimal, but also absolutely feasible under physical spatiotemporal constraints. This enables long-distance delivery in complex and ever-changing cities, improving delivery efficiency and on-time delivery rate.

[0053] Furthermore, controlling the unmanned equipment to execute the unmanned delivery task based on the global delivery path includes: Based on the near-end strategy optimization algorithm and the global delivery path, the unmanned delivery task is decomposed into multiple delivery sub-tasks; The unmanned equipment is controlled to execute the delivery sub-tasks sequentially.

[0054] Specifically, the global delivery route is decomposed into a series of executable delivery sub-tasks using the Proximal Policy Optimization (PPO) algorithm, thereby enabling micro-level real-time control and adjustment based on macro-level route planning.

[0055] In this embodiment of the invention, during the decomposition of unmanned delivery tasks, a distinction is made between global and local reward functions, the optimization objective of reinforcement learning is clearly defined, and the loss function of deep learning is completely decoupled.

[0056] The expression for the global policy layer reward function is as follows: In the formula, This represents the total global reward. Basic rewards for completing unmanned delivery missions. Incentives will be given for saving time and effort. For energy efficiency rewards, As a reward for connection reliability, All are weighting coefficients.

[0057] The expression for the reward function of the local execution layer is as follows: In the formula, Indicates the total reward in a localized area. Basic rewards for completing delivery sub-tasks. Penalty for invalid waiting time Penalty for unnecessary energy consumption Punishment for security risks, This is the penalty coefficient.

[0058] For example, in one embodiment, the global strategy module outputs the global delivery path corresponding to the unmanned delivery task "The multi-scenario adaptable AGV needs to connect with the outdoor transport platform at the door of subway car No. 3 at 10:00". The PPO algorithm decomposes this unmanned delivery task into: delivery sub-task 1 - control the AGV to move from the car seat to the door at 9:55; delivery sub-task 2 - wait at the door and identify the docking signal; delivery sub-task 3 - perform the precise docking action at 9:59. If a temporary obstacle (such as other passengers' luggage) is encountered when performing delivery sub-task 1, the PPO algorithm can adjust the local path of the AGV in real time to ensure that subsequent delivery sub-tasks are not delayed.

[0059] The embodiments of the present invention enhance the robustness of unmanned delivery in the face of real-time and local disturbances (such as dynamic obstacles and short-term delays) through a hierarchical control architecture of "global planning + local execution". It realizes the decomposability and flexible adjustment of the delivery process, and ensures that even in complex dynamic environments, unmanned delivery tasks can be completed steadily and step by step, thereby guaranteeing the reliability and success rate of the entire delivery chain.

[0060] The unmanned delivery device based on trunk-branch-terminal fusion provided by the present invention will be described below. The unmanned delivery device based on trunk-branch-terminal fusion described below can be referred to in correspondence with the unmanned delivery method based on trunk-branch-terminal fusion described above.

[0061] Figure 2 This is a schematic diagram of the unmanned delivery device based on the fusion of trunk and branch lines provided by the present invention, as shown below. Figure 2 As shown. This embodiment of the invention provides an unmanned delivery device based on trunk-branch-terminal fusion, comprising a data acquisition module 201, a path planning module 202, and an unmanned delivery module 203, wherein: The data acquisition module 201 is used to collect delivery data for unmanned delivery tasks; the delivery data includes time-series data, spatial structure data, and static attribute data; the path planning module 202 is used to input the delivery data into a path planning model to obtain a global delivery path output by the path planning model; the path planning model includes a time-series feature module, a topology feature module, a static feature module, a mapping module, and a global strategy module; the time-series feature module is used to extract time-series correlation features based on the time-series data; the topology feature module is used to extract spatial topology features based on the spatial structure data; the static feature module is used to extract static features based on the static attribute data; the mapping module is used to map the time-series correlation features, the spatial topology features, and the static features into a spatiotemporal feature vector; the global strategy module is used to output the global delivery path based on the spatiotemporal feature vector; the unmanned delivery module 203 is used to control unmanned equipment to execute the unmanned delivery task based on the global delivery path.

[0062] The unmanned delivery device based on the fusion of time, space, and static attributes provided by this invention collects delivery data for unmanned delivery tasks. This delivery data includes time-series data, spatial structure data, and static attribute data, thus comprehensively collecting multi-dimensional delivery data in terms of time, space, and static attributes. This provides complete and accurate input information for subsequent path planning. The delivery data is then input into a path planning model to obtain the global delivery path output by the model. The path planning model includes a time-series feature module, a topology feature module, a static feature module, a mapping module, and a global strategy module. The time-series feature module extracts time-series correlation features based on the time-series data; the topology feature module extracts spatial topology features based on the spatial structure data; the static feature module extracts static features based on the static attribute data; and the mapping module maps the time-series correlation features... The system maps spatial topological features and static features into spatiotemporal feature vectors. A global strategy module outputs a global delivery path based on these spatiotemporal feature vectors. This path planning model, integrating temporal feature modules, topological feature modules, static feature modules, and mapping and global strategy modules, performs deep feature extraction and fusion on multi-source heterogeneous data. Ultimately, it generates the optimal global delivery path under complex spatiotemporal constraints, improving the reliability and efficiency of long-distance unmanned delivery. Based on the global delivery path, unmanned equipment is controlled to execute unmanned delivery tasks, transforming the planned optimal path into executable instructions. This drives collaborative operation of unmanned equipment, translating the algorithmic optimization results into precise and automated delivery actions in the physical world. This enables unmanned delivery in complex and ever-changing urban environments, reducing delivery costs and improving safety and efficiency.

[0063] Figure 3 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 3 As shown, the electronic device may include: a processor 310, a communications interface 320, a memory 330, and a communication bus 340, wherein the processor 310, the communications interface 320, and the memory 330 communicate with each other via the communication bus 340. The processor 310 can call logical instructions in the memory 330 to execute an unmanned delivery method based on trunk-branch-terminal fusion, the method including: The system collects delivery data for unmanned delivery missions; the delivery data includes time-series data, spatial structure data, and static attribute data. The delivery data is input into a path planning model to obtain a global delivery path output by the model. The path planning model includes a temporal feature module, a topological feature module, a static feature module, a mapping module, and a global strategy module. The temporal feature module is used to extract temporal correlation features based on the temporal sequence data. The topological feature module is used to extract spatial topological features based on the spatial structure data. The static feature module is used to extract static features based on the static attribute data. The mapping module is used to map the temporal correlation features, the spatial topological features, and the static features into a spatiotemporal feature vector. The global strategy module is used to output the global delivery path based on the spatiotemporal feature vector. Based on the global delivery path, control the unmanned equipment to execute the unmanned delivery task.

[0064] Furthermore, the logical instructions in the aforementioned memory 330 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0065] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program that can be stored on a non-transitory computer-readable storage medium, wherein when the computer program is executed by a processor, the computer is able to execute the unmanned delivery method based on the fusion of trunk and branch lines provided by the above methods, the method comprising: The system collects delivery data for unmanned delivery missions; the delivery data includes time-series data, spatial structure data, and static attribute data. The delivery data is input into a path planning model to obtain a global delivery path output by the model. The path planning model includes a temporal feature module, a topological feature module, a static feature module, a mapping module, and a global strategy module. The temporal feature module is used to extract temporal correlation features based on the temporal sequence data. The topological feature module is used to extract spatial topological features based on the spatial structure data. The static feature module is used to extract static features based on the static attribute data. The mapping module is used to map the temporal correlation features, the spatial topological features, and the static features into a spatiotemporal feature vector. The global strategy module is used to output the global delivery path based on the spatiotemporal feature vector. Based on the global delivery path, control the unmanned equipment to execute the unmanned delivery task.

[0066] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the unmanned delivery method based on trunk-branch fusion provided by the above methods, the method comprising: The system collects delivery data for unmanned delivery missions; the delivery data includes time-series data, spatial structure data, and static attribute data. The delivery data is input into a path planning model to obtain a global delivery path output by the model. The path planning model includes a temporal feature module, a topological feature module, a static feature module, a mapping module, and a global strategy module. The temporal feature module is used to extract temporal correlation features based on the temporal sequence data. The topological feature module is used to extract spatial topological features based on the spatial structure data. The static feature module is used to extract static features based on the static attribute data. The mapping module is used to map the temporal correlation features, the spatial topological features, and the static features into a spatiotemporal feature vector. The global strategy module is used to output the global delivery path based on the spatiotemporal feature vector. Based on the global delivery path, control the unmanned equipment to execute the unmanned delivery task.

[0067] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0068] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0069] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0070] In this application's embodiments, "determine B based on A" means that factor A must be considered when determining B. It is not limited to "B can be determined based solely on A," but should also include: "determine B based on A and C," "determine B based on A, C, and E," "determine C based on A, and further determine B based on C," etc. Additionally, it can include using A as a condition for determining B, for example, "when A meets the first condition, determine B using the first method"; another example, "when A meets the second condition, determine B," etc.; another example, "when A meets the third condition, determine B based on the first parameter," etc. Of course, it can also be a condition where A is a factor in determining B, for example, "when A meets the first condition, determine C using the first method, and further determine B based on C," etc.

[0071] It should also be noted that the terms "target," "first," and "second" in this invention are used to distinguish similar objects, not to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first" and "second" are generally of the same class, without limiting the number of objects; for example, the first object can be one or more.

[0072] In this invention, the term "multiple" refers to two or more kinds, and other quantifiers are similar.

[0073] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for unmanned delivery based on the fusion of the sexagenary cycle and the end of the lunar calendar, characterized in that, include: The system collects delivery data for unmanned delivery missions; the delivery data includes time-series data, spatial structure data, and static attribute data. The delivery data is input into a path planning model to obtain the global delivery path output by the path planning model. The path planning model includes a temporal feature module, a topological feature module, a static feature module, a mapping module, and a global strategy module. The temporal feature module is used to extract temporal correlation features based on the temporal sequence data. The topological feature module is used to extract spatial topological features based on the spatial structure data. The static feature module is used to extract static features based on the static attribute data. The mapping module is used to map the temporal correlation features, the spatial topology features, and the static features into spatiotemporal feature vectors; the global strategy module is used to output the global delivery path based on the spatiotemporal feature vectors. Based on the global delivery path, control the unmanned equipment to execute the unmanned delivery task.

2. The unmanned delivery method based on the fusion of solar and cyclical branches according to claim 1, characterized in that, The time-series data includes at least one of the following: Current operational data for public transportation; Historical passenger flow data for the public transportation vehicles; Current meteorological data; Current road congestion data; Historical traffic accident data; Future weather data for a preset time period; The location data of the unmanned equipment.

3. The unmanned delivery method based on the fusion of solar and cyclical branches according to claim 1, characterized in that, The spatial structure data includes at least one of the following: Linked data between public transportation stations; The association data between the starting point of the unmanned delivery mission and the public transportation station; The association data between the destination of the unmanned delivery mission and the public transportation station.

4. The unmanned delivery method based on the fusion of solar and cyclical branches according to claim 1, characterized in that, The static attribute data includes at least one of the following: The starting address data of the unmanned delivery mission; The destination address data of the unmanned delivery mission; The order information corresponding to the unmanned delivery task; The carrying parameters of the unmanned equipment; Basic attribute data of public transportation stations.

5. The unmanned delivery method based on the fusion of solar and cyclical branches according to claim 1, characterized in that, The global policy module is specifically used for: Based on the improved spatiotemporal aligned attention operator and the spatiotemporal feature vector, scores corresponding to multiple potential delivery paths are calculated respectively; Based on the score, the global delivery route is determined from the potential delivery routes.

6. The unmanned delivery method based on the fusion of solar and cyclical branches according to claim 1, characterized in that, The step of controlling unmanned equipment to execute the unmanned delivery task based on the global delivery path includes: Based on the near-end strategy optimization algorithm and the global delivery path, the unmanned delivery task is decomposed into multiple delivery sub-tasks; The unmanned equipment is controlled to execute the delivery sub-tasks sequentially.

7. An unmanned delivery device based on the fusion of cyclical and seasonal elements, characterized in that, include: The data acquisition module is used to collect delivery data for unmanned delivery tasks; the delivery data includes time-series data, spatial structure data, and static attribute data. The route planning module is used to input the delivery data into the route planning model to obtain the global delivery route output by the route planning model. The route planning model includes a temporal feature module, a topology feature module, a static feature module, a mapping module, and a global strategy module. The temporal feature module is used to extract temporal correlation features based on the temporal sequence data. The topology feature module is used to extract spatial topology features based on the spatial structure data. The static feature module is used to extract static features based on the static attribute data. The mapping module is used to map the temporal correlation features, the spatial topology features, and the static features into spatiotemporal feature vectors; the global strategy module is used to output the global delivery path based on the spatiotemporal feature vectors. The unmanned delivery module is used to control unmanned equipment to perform the unmanned delivery task based on the global delivery path.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the unmanned delivery method based on the fusion of trunk and branch lines as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the unmanned delivery method based on the fusion of trunk and branch lines as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the unmanned delivery method based on the fusion of trunk and branch lines as described in any one of claims 1 to 6.

Images