Warehousing self-operation device for unmanned aerial vehicle distribution based on AI algorithm

Abstract

The invention relates to the technical field of warehouse logistics, in particular to an unmanned aerial vehicle distribution warehouse self-operation device based on an AI algorithm, which comprises an intelligent warehouse structure, the intelligent warehouse structure comprises a plurality of multi-layer honeycomb shelves, an environment sensing matrix and an unmanned aerial vehicle cluster, and the unmanned aerial vehicle cluster is provided with a multi-modal navigation module and an airborne AI calculation unit; and a cloud cooperative control system, wherein the intelligent storage structure, the unmanned aerial vehicle cluster and the cloud cooperative control system realize real-time data interaction through a 5G private network. Through deep fusion of an intelligent warehousing structure, an unmanned aerial vehicle cluster and a cloud cooperative control system, full-link unmanned intelligent scheduling of warehousing operation and tail end distribution is realized, the system adopts an improved RRT algorithm and a federal reinforcement learning framework, the robustness of unmanned aerial vehicle path planning in a complex environment is significantly improved, and the system has a wide application prospect. And dynamic optimization of virtual-real linkage is realized through a digital twinning technology, and the method is suitable for high-timeliness logistics scenes such as medicine cold chains and emergency materials.

Description

Technical Field

[0001] The present invention relates to a self-operating warehousing device, in particular to a self-operating warehousing device for unmanned aircraft delivery based on an AI algorithm, and belongs to the technical field of warehousing and logistics. Background Art

[0002] AI-powered drone delivery and warehousing systems significantly improve logistics efficiency, reduce costs, and promote the development of green logistics through dynamic scheduling, route optimization, and automated operations. This model has already gained commercial application in scenarios such as remote area logistics and urban instant delivery. In the future, as technology evolves and standards improve, drone warehousing and delivery will become key infrastructure for logistics industry upgrades. Existing technologies have the following defects: traditional warehouse robots only complete in-warehouse handling and lack coordination with drone delivery systems, resulting in efficiency bottlenecks in the outbound loading and takeoff links; drones rely on preset routes and lack real-time obstacle avoidance and path reconstruction capabilities in complex environments inside and outside the warehouse; multi-drone task allocation is based on static priority rules, making it difficult to dynamically respond to sudden orders or equipment failures.

[0003] Therefore, there is an urgent need to improve the warehouse self-operating device for unmanned aerial vehicle delivery based on AI algorithms to solve the above-mentioned problems. Summary of the Invention

[0004] The purpose of this invention is to provide a self-operating warehouse device for unmanned aircraft delivery based on AI algorithms. Through the deep integration of intelligent warehouse structure, drone cluster and cloud-based collaborative control system, it realizes unmanned intelligent scheduling of the entire link between warehousing operations and terminal delivery. The system adopts an improved RRT algorithm and a federated reinforcement learning framework to significantly improve the robustness of drone path planning in complex environments, and realizes dynamic optimization of virtual and real linkage through digital twin technology. It is suitable for high-timeliness logistics scenarios such as pharmaceutical cold chain and emergency supplies.

[0005] In order to achieve the above objectives, the main technical solutions adopted by the present invention include: A self-operating warehousing device for drone delivery based on AI algorithms, comprising: An intelligent storage structure comprising a plurality of multi-layer honeycomb shelves and an environmental sensing matrix disposed on the multi-layer honeycomb shelves, wherein the environmental sensing matrix comprises a laser radar, a thermal imaging camera, and a temperature and humidity sensor; A drone swarm, wherein the drone swarm is provided with a multimodal navigation module and an onboard AI computing unit; and A cloud-based collaborative control system, comprising a digital twin platform, a hybrid AI algorithm engine, a Transformer global task scheduling model, and an improved RRT; The intelligent storage structure, the drone cluster and the cloud-based collaborative control system realize real-time data interaction through the 5G private network.

[0006] Preferably, each of the multi-layer honeycomb shelves is provided with a plurality of evenly distributed electromagnetic pushing mechanisms, wherein the electromagnetic pushing mechanisms include shelf partitions, the bottom of the shelf partitions being fixedly provided on the multi-layer honeycomb shelves via partition support arms, a driving motor being fixedly provided at the middle of the partition support arms, a partition driving gear being provided at the output end of the driving motor, and a driving rack corresponding to the partition driving gear being fixedly provided on the bottom side surface of the shelf partitions, wherein when the partition driving gear is started, the partition driving gear is used to push the shelf partitions horizontally; The bottom side surface of the shelf partition is provided with symmetrically distributed guide rods, and both ends of the partition support arm are provided with guide arms, and the guide rods are slidably connected to the guide arms.

[0007] Preferably, an automatic weighing module is fixedly installed above the shelf partition, and the partition drive gear and the automatic weighing module are both connected to the cloud collaborative control system in a communication manner; A sensing module is fixedly provided on the side of the shelf partition, and the environment sensing matrix is fixedly provided on the sensing module.

[0008] Preferably, a parking board is fixedly provided on the upper side of the multi-layer honeycomb shelf, and the drone cluster is placed on the parking board; The shutdown board is provided with a wireless charging transmitter, and the drone cluster is provided with a wireless charging receiver electrically connected to the drone cluster battery, and the wireless charging transmitter corresponds to the wireless charging receiver.

[0009] Preferably, the multimodal navigation module integrates UWB indoor positioning and visual SLAM algorithms to achieve group obstacle avoidance knowledge sharing through federated learning, wherein the fusion process is: front end: Visual SLAM extracts ORB / SIFT features and calculates the initial pose; UWB indoor positioning provides a rough location, helping to solve the scale ambiguity problem of visual SLAM; rear end: Construct an optimization problem, where the algorithm formula is: ; Among them, T is the camera pose, P is the UWB anchor point position, is the visual reprojection error, is the UWB ranging error; Dynamic environmental adaptation: RANSAC is used to eliminate abnormal UWB ranging, and visual SLAM is used to detect dynamic objects to avoid UWB interference.

[0010] Preferably, the cloud-based collaborative control system adopts a hybrid decision-making architecture driven by the digital twin platform, including a global task scheduling model based on a hybrid AI algorithm engine, Transformer, and a dynamic path planning layer of an improved RRT.

[0011] Preferably, the motion state estimation algorithm of the dynamic path planning layer of the improved RRT is: a. Dynamic environment modeling; b. Improve sampling strategy; c. Real-time re-planning conditions.

[0012] Preferably, the dynamic environment modeling includes: Dynamic obstacle prediction: + ; in, is the predicted position of obstacle k at time t; , is the velocity and acceleration; ~ϰ(0,Σ k ) is the prediction noise; Dynamic collision risk field: ; is the obstacle influence radius, and σ is the smoothing factor.

[0013] Preferably, the improved sampling strategy includes: Dynamic hybrid sampling distribution: P ; Dynamic hybrid sampling distribution: ; Dynamic obstacle biased sampling: ; Weight , the faster the speed, the higher the sampling density; Adaptive parameter adjustment: , ; in, is the α value at the initial moment, λ is a positive constant, indicating the speed at which α decays over time, t is the time variable, and β represents the weight of global exploration.

[0014] Preferably, the real-time replanning conditions include: Spatiotemporal joint cost function: ; in is the target distance, is the dynamic risk score, is the curvature smoothing term, where the curvature smoothing term formula is: ; Real-time replanning conditions: ; is the risk threshold.

[0015] The present invention has at least the following beneficial effects: 1. Through the deep integration of intelligent warehousing structure, drone clusters and cloud-based collaborative control systems, unmanned intelligent scheduling of the entire chain of warehousing operations and terminal distribution is realized. The system adopts an improved RRT algorithm and a federated reinforcement learning framework to significantly improve the robustness of drone path planning in complex environments, and realizes dynamic optimization of virtual-real linkage through digital twin technology. It is suitable for high-efficiency logistics scenarios such as pharmaceutical cold chain and emergency supplies.

[0016] 2. When items need to be transferred, the partition drive gear is driven by the drive motor to rotate, and at the same time, the drive rack at the bottom of the shelf partition is driven, and the shelf partition is driven to the outside of the intelligent storage structure through the drive rack. Of course, it can be done on both sides, which improves the flexibility and scope of use, facilitates drone grasping, has a simple structure, a high degree of automation, and avoids contact, thereby improving the efficiency of transfer. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings: Figure 1 It is the electrical principle diagram of the present invention; Figure 2 A three-dimensional diagram of the intelligent storage structure of the present invention; Figure 3 The electromagnetic pushing mechanism structure of the present invention Figure 1 ; Figure 4 The electromagnetic pushing mechanism structure of the present invention Figure 2 ; Figure 5 It is a structural diagram of the shutdown board of the present invention.

[0018] In the figure, 1. Intelligent storage structure; 101. Multi-layer honeycomb shelves; 102. Environmental perception matrix; 103. Shutdown board; 1031. Wireless charging transmitter; 1021. LiDAR; 1022. Thermal imaging camera; 1023. Temperature and humidity sensor; 2. Drone cluster; 201. Multimodal navigation module; 202. Airborne AI computing unit; 203. Wireless charging receiver; 3. Cloud-based collaborative control system; 301. Digital twin platform; 302. Hybrid AI algorithm engine; 303. Transformer global task scheduling model; 304. Improved RRT; 4. Electromagnetic pushing mechanism; 401. Shelf partition; 402. Partition drive gear; 403. Partition support arm; 404. Drive rack; 405. Guide arm; 406. Guide rod; 407. Drive motor; 408. Automatic weighing module; 409. Induction module. DETAILED DESCRIPTION

[0019] The following will describe the implementation methods of the present application in detail with reference to the accompanying drawings and examples, so that the implementation process of how the present application applies technical means to solve technical problems and achieve technical effects can be fully understood and implemented accordingly.

[0020] like Figure 1-Figure 5As shown, the warehousing self-operating device for unmanned aircraft delivery based on AI algorithm provided in this embodiment includes: an intelligent warehousing structure 1, the intelligent warehousing structure 1 includes multiple multi-layer honeycomb shelves 101 and an environmental perception matrix 102 arranged on the multi-layer honeycomb shelves 101, the environmental perception matrix 102 includes a laser radar 1021, a thermal imaging camera 1022 and a temperature and humidity sensor 1023, a sensing module 409 is fixedly provided on the side of the shelf partition 401, and the environmental perception matrix 102 is fixedly provided on the sensing module 409. The environmental perception matrix 102 is a differential geometry mapping of environment-decision, which provides a differentiable decision basis for the autonomous system through structured representation of the spatiotemporal threat gradient. Its design directly affects the security, real-time performance and collaborative efficiency of the system, and a drone cluster 2. A multimodal navigation module is provided on the drone cluster 2. Block 201 and airborne AI computing unit 202; and cloud-based collaborative control system 3, cloud-based collaborative control system 3 includes digital twin platform 301, hybrid AI algorithm engine 302, Transformer's global task scheduling model 303 and improved RRT304, intelligent warehousing structure 1, drone cluster 2 and cloud-based collaborative control system 3 realize real-time data interaction through 5G private network, and realize full-link unmanned intelligent scheduling of warehousing operations and terminal distribution through the deep integration of intelligent warehousing structure 1, drone cluster 2 and cloud-based collaborative control system 3. The system adopts improved RRT304 algorithm and federated reinforcement learning framework to significantly improve the robustness of drone path planning in complex environments, and realizes dynamic optimization of virtual-real linkage through digital twin technology, which is suitable for high-timeliness logistics scenarios such as pharmaceutical cold chain and emergency supplies.

[0021] Further, such as Figure 3 and Figure 4 As shown, each multi-layer honeycomb shelf 101 is provided with a plurality of evenly distributed electromagnetic pushing mechanisms 4, the electromagnetic pushing mechanism 4 includes a shelf partition 401, the bottom of the shelf partition 401 is fixedly provided on the multi-layer honeycomb shelf 101 through a partition support arm 403, a driving motor 407 is fixedly provided in the middle of the partition support arm 403, and a partition driving gear 402 is provided at the output end of the driving motor 407, and a driving rack 404 corresponding to the partition driving gear 402 is fixedly provided on the bottom side of the shelf partition 401, wherein the partition driving gear When 402 is started, the partition driving gear 402 is used to push the shelf partition 401 horizontally. When the items need to be transferred, the partition driving gear 402 is driven to rotate by the driving motor 407, and at the same time drives the driving rack 404 at the bottom of the shelf partition 401, and drives the shelf partition 401 to the outside of the intelligent storage structure 1 through the driving rack 404. Of course, both sides can be used to improve the flexibility and scope of use, and facilitate the drone to grab. It has a simple structure and a high degree of automation, while avoiding collisions and improving the efficiency of transfer; In order to maintain the stability of the shelf partition 401 in movement, symmetrically distributed guide rods 406 are provided on the bottom side of the shelf partition 401. Guide arms 405 are provided at both ends of the partition support arm 403. The guide rods 406 are slidably connected to the guide arms 405. The guide arms 405 are fixedly provided on the shelf partition 401. The guide rods 406 at the bottom of the shelf partition 401 play a certain guiding role for the shelf partition 401, thereby improving the stability of the movement of the shelf partition 401. In addition, if Figure 4 As shown, an automatic weighing module 408 is fixedly installed above the shelf partition 401. The partition drive gear 402 and the automatic weighing module 408 are both connected to the cloud collaborative control system 3 to establish a communication connection. The automatic weighing module 408 can weigh the weight of the items, improving the flexibility of use. The cloud collaborative control system 3 generates the optimal drone task matching based on the order and plans the warehouse pickup route and external delivery route; Furthermore, Figure 5 As shown, a shutdown board 103 is fixedly provided on the upper side of the multi-layer honeycomb shelf 101, and the drone cluster 2 is placed on the shutdown board 103. The drone cluster 2 is parked on the shutdown board 103 of the multi-layer honeycomb shelf 101 and is in a start-up state at any time, thereby improving timeliness. Due to the long-term use of the drone cluster 2, the drone cluster 2 needs to be charged. A wireless charging transmitter 1031 is provided on the shutdown board 103, and a wireless charging receiver 203 electrically connected to the battery of the drone cluster 2 is provided on the drone cluster 2. The wireless charging transmitter 1031 corresponds to the wireless charging receiver 203. A wireless charging transmitter 1031 is fixedly provided on the shutdown board 103, and a wireless charging receiver 203 is fixedly provided on the drone cluster 2. Therefore, when the drone cluster 2 stays on the shutdown board 103, the wireless charging transmitter 1031 corresponds to the wireless charging receiver 203. Therefore, the drone cluster 2 can be charged at any time without manual operation by the staff. The degree of automation is high, which improves the efficiency of transportation.

[0022] Furthermore, the multimodal navigation module 201 integrates UWB indoor positioning and visual SLAM algorithms to achieve group obstacle avoidance knowledge sharing through federated learning. The fusion process is as follows: Front-end inter-frame tracking: Visual SLAM extracts ORB / SIFT features and calculates the initial pose; UWB indoor positioning provides a rough position, helping to solve the scale ambiguity problem of visual SLAM and monocular SLAM; Backend global optimization: Construct an optimization problem, where the algorithm formula is: ; Among them, T is the camera pose, P is the UWB anchor point position, is the visual reprojection error, is the UWB ranging error; Dynamic environmental adaptation: Use RANSAC to eliminate abnormal UWB ranging, and visual SLAM to detect dynamic objects to avoid UWB interference; Moreover, the cloud-based collaborative control system 3 adopts a hybrid decision-making architecture driven by a digital twin platform 301, which includes a global task scheduling model 303 based on a hybrid AI algorithm engine 302, a Transformer, and a dynamic path planning layer with an improved RRT; The motion state estimation algorithm of the dynamic path planning layer of the improved RRT is: a. Dynamic environment modeling; Dynamic environment modeling includes: Dynamic obstacle prediction: + ; in, is the predicted position of obstacle k at time t; , is the velocity and acceleration; ~ϰ(0,Σ k ) is the prediction noise; Dynamic collision risk field: ; is the obstacle influence radius, σ is the smoothing factor, which constrains the path turning angle, reduces unnecessary bends, and makes the movement smoother and more energy-efficient. For example, the path is optimized by B-spline curve to generate a smooth trajectory. Asymptotic optimality: Through the RRT algorithm, the optimal solution is gradually approached by reselecting parent nodes and rewiring mechanisms to ensure path quality; b. Improve sampling strategy; Improved sampling strategies include: Dynamic hybrid sampling distribution: P ; Dynamic hybrid sampling distribution: ; Dynamic obstacle biased sampling: ; Weight , the faster the speed, the higher the sampling density; Adaptive parameter adjustment: , ; in, is the α value at the initial moment, λ is a positive constant, indicating the rate at which α decays over time, t is the time variable, and β represents the weight of global exploration; c. Real-time replanning conditions; Real-time replanning conditions include: Spatiotemporal joint cost function: ; in is the target distance, is the dynamic risk score, is the curvature smoothing term, where the curvature smoothing term formula is: ; Real-time replanning conditions: ; For the risk threshold, by introducing the target bias probability, the algorithm is improved to make the sampling points more concentrated in the target area and reduce invalid exploration. For example, the "target point direction weight" is used to guide the expansion of new nodes to avoid blind search. The motion state estimation algorithm of the dynamic path planning layer of RRT is improved to remove redundant points in the path, shorten the path length, and improve the planning speed.

[0023] For example, certain words are used in the specification and claims to refer to specific components. Those skilled in the art should understand that hardware manufacturers may use different terms to refer to the same component. This specification and claims do not use differences in names as a way to distinguish components, but use differences in the functions of the components as the criteria for distinction. For example, "including" mentioned throughout the specification and claims is an open term, so it should be interpreted as "including but not limited to". "Approximately" means that within an acceptable error range, those skilled in the art can solve technical problems within a certain error range and basically achieve technical effects.

[0024] It should be noted that the terms "include," "comprises," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a product or system comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such product or system. In the absence of further limitations, an element defined by the phrase "comprising a..." does not exclude the presence of other identical elements in the product or system comprising the element.

[0025] The foregoing description shows and describes several preferred embodiments of the present invention. However, as previously stated, it should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. Rather, the present invention can be used in various other combinations, modifications, and environments and can be modified within the scope of the inventive concept described herein by the teachings above or by techniques or knowledge in the relevant art. Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention are intended to be within the scope of the appended claims.

Claims

1. A self-operating storage device for unmanned aircraft delivery based on AI algorithm, characterized in that: include: An intelligent storage structure (1), the intelligent storage structure (1) comprising a plurality of multi-layer honeycomb shelves (101) and an environment sensing matrix (102) arranged on the multi-layer honeycomb shelves (101), the environment sensing matrix (102) comprising a laser radar (1021), a thermal imaging camera (1022), and a temperature and humidity sensor (1023); A drone cluster (2), wherein the drone cluster (2) is provided with a multimodal navigation module (201) and an onboard AI computing unit (202); A cloud-based collaborative control system (3), wherein the cloud-based collaborative control system (3) includes a digital twin platform (301), a hybrid AI algorithm engine (302), a Transformer global task scheduling model (303), and an improved RRT (304); The intelligent storage structure (1), the drone cluster (2) and the cloud-based collaborative control system (3) realize real-time data interaction through a 5G private network.

2. The self-operating storage device for unmanned aircraft delivery based on AI algorithm according to claim 1 is characterized by: Each of the multi-layer honeycomb shelves (101) is provided with a plurality of evenly distributed electromagnetic pushing mechanisms (4), the electromagnetic pushing mechanisms (4) comprising a shelf partition (401), the bottom of the shelf partition (401) being fixedly provided on the multi-layer honeycomb shelf (101) via a partition support arm (403), a driving motor (407) being fixedly provided in the middle of the partition support arm (403), a partition driving gear (402) being provided at the output end of the driving motor (407), a driving rack (404) corresponding to the partition driving gear (402) being fixedly provided on the bottom side surface of the shelf partition (401), wherein when the partition driving gear (402) is started, the partition driving gear (402) is used to push the shelf partition (401) horizontally; The bottom side of the shelf partition (401) is provided with symmetrically distributed guide rods (406), and both ends of the partition support arm (403) are provided with guide arms (405), and the guide rods (406) are slidably connected to the guide arms (405).

3. The AI algorithm-based unmanned aircraft delivery warehousing self-operating device according to claim 2, characterized in that: An automatic weighing module (408) is fixedly provided above the shelf partition (401), and both the partition drive gear (402) and the automatic weighing module (408) establish communication connections with the cloud collaborative control system (3); A sensing module (409) is fixedly provided on the side of the shelf partition (401), and the environment sensing matrix (102) is fixedly provided on the sensing module (409).

4. The self-operating storage device for unmanned aircraft delivery based on AI algorithm according to claim 2 is characterized by: A parking board (103) is fixedly provided on the upper side of the multi-layer honeycomb shelf (101), and the drone cluster (2) is placed on the parking board (103); A wireless charging transmitter (1031) is provided on the shutdown plate (103), and a wireless charging receiver (203) electrically connected to the battery of the drone cluster (2) is provided on the drone cluster (2), and the wireless charging transmitter (1031) corresponds to the wireless charging receiver (203).

5. The AI algorithm-based unmanned aircraft delivery warehousing self-operating device according to claim 1, characterized in that: The multimodal navigation module (201) integrates UWB indoor positioning and visual SLAM algorithms, and realizes group obstacle avoidance knowledge sharing through federated learning, wherein the fusion process is: Front-end (inter-frame tracking): Visual SLAM extracts ORB / SIFT features and calculates the initial pose; UWB indoor positioning provides a rough location, helping to solve the scale ambiguity problem of visual SLAM (monocular SLAM); Backend (global optimization): Construct an optimization problem, where the algorithm formula is: ； Among them, T is the camera pose, P is the UWB anchor point position, is the visual reprojection error, is the UWB ranging error; Dynamic environmental adaptation: RANSAC is used to eliminate abnormal UWB ranging, and visual SLAM is used to detect dynamic objects to avoid UWB interference.

6. The AI algorithm-based unmanned aircraft delivery warehousing self-operating device according to claim 1, characterized in that: The cloud-based collaborative control system (3) adopts a hybrid decision-making architecture driven by the digital twin platform (301), including a hybrid AI algorithm engine (302), a Transformer-based global task scheduling model (303), and an improved RRT dynamic path planning layer.

7. The self-operating storage device for unmanned aircraft delivery based on AI algorithm according to claim 6, characterized in that: The motion state estimation algorithm of the dynamic path planning layer of the improved RRT is: a. Dynamic environment modeling; b. Improve sampling strategy; c. Real-time re-planning conditions.

8. The self-operating storage device for unmanned aircraft delivery based on AI algorithm according to claim 7, characterized in that: The dynamic environment modeling includes: Dynamic obstacle prediction: + ； in, is the predicted position of obstacle k at time t; , is the velocity and acceleration; ~ϰ(0,Σ k ) is the prediction noise; Dynamic collision risk field: ； is the obstacle influence radius, and σ is the smoothing factor.

9. The self-operating storage device for unmanned aircraft delivery based on AI algorithm according to claim 7, characterized in that: The improved sampling strategy includes: Dynamic hybrid sampling distribution: P ； Dynamic hybrid sampling distribution: ； Dynamic obstacle biased sampling: ； Weight , the faster the speed, the higher the sampling density; Adaptive parameter adjustment: ， ； in, is the α value at the initial moment, λ is a positive constant, indicating the speed at which α decays over time, t is the time variable, and β represents the weight of global exploration.

10. The self-operating storage device for unmanned aircraft delivery based on AI algorithm according to claim 7, characterized in that: The real-time replanning conditions include: Spatiotemporal joint cost function: ； in is the target distance, is the dynamic risk score, is the curvature smoothing term, where the curvature smoothing term formula is: ； Real-time replanning conditions: ； is the risk threshold.

Images