Path planning method, computer equipment and computer readable storage medium

Abstract

The embodiment of the invention discloses a path planning method, computer equipment and a computer readable storage medium. The method comprises the following steps that task information and state information of a storage robot are obtained; inputting the task information and the state information into a path planning model to obtain a path planning result; the path planning model comprises a global planning layer and a local optimization layer, the global planning layer is used for generating a global optimal path, and the local optimization layer is used for locally optimizing the global optimal path to output a path planning result; and controlling the storage robot to execute tasks according to the path planning result. Therefore, the task information and the state information of the storage robot can be input into the path planning model, the path planning is completed through the hybrid algorithm and the cooperative work of the global planning layer and the local optimization layer, the accuracy and the adaptability of path generation are improved, and the path planning efficiency is improved. The problem of low efficiency caused by traditional single-level planning is effectively avoided.

Description

Technical Field

[0001] This application belongs to the field of industrial automation, and in particular relates to a path planning method, computer equipment, and computer-readable storage medium. Background Technology

[0002] In the technological fields involving large-scale manufacturing, especially in the areas of automated production and intelligent control represented by the automotive and construction machinery industries, AGVs (Automated Guided Vehicles) have been widely adopted as warehouse robots to participate in manufacturing production lines, automating material handling, assembly assistance, and production processes. Traditional manufacturing parts and materials warehousing is characterized by its diverse types and specifications, frequent inbound and outbound operations, and random demand; therefore, efficient picking paths are crucial for ensuring production continuity. Current AGV scheduling systems rely on intelligent algorithms (such as deep reinforcement learning and improved spatiotemporal A* algorithms) combined with edge computing and 5G network slicing technology to achieve dynamic path planning and multi-objective optimization, but they face challenges such as insufficient real-time response and complex multi-device collaboration. How to improve the path planning capabilities of AGVs is a technical problem that urgently needs to be solved by those skilled in the art.

[0003] The preceding description is intended to provide general background information and does not necessarily constitute prior art. Summary of the Invention

[0004] The purpose of this application is to provide a path planning method, computer device, and computer-readable storage medium that can improve the path planning capabilities of warehouse robots.

[0005] To achieve the above objectives: In a first aspect, embodiments of this application provide a path planning method, comprising the following steps: obtaining task information and status information of a warehouse robot; inputting the task information and status information into a path planning model to obtain a path planning result; the path planning model includes a global planning layer and a local optimization layer, wherein the global planning layer is used to generate a globally optimal path, and the local optimization layer is used to perform local optimization on the globally optimal path to output the path planning result; and controlling the warehouse robot to perform tasks according to the path planning result.

[0006] In an optional embodiment of this application, the task information includes at least one of actual task information and predicted task information; obtaining the task information of the warehouse robot includes: obtaining the actual task information sent to the warehouse robot by the host computer; and / or obtaining the historical task information of the warehouse robot, inputting the historical task information into the task prediction model to obtain the predicted task information.

[0007] In an optional embodiment of this application, obtaining task information and status information of the warehousing robot includes: establishing a network connection with the warehousing robot and dividing the network into a first sub-network, a second sub-network, and a third sub-network; obtaining map data from the status information through the first sub-network; obtaining task information through the second sub-network and storing the task information; and obtaining sensor data from the status information through the third sub-network.

[0008] In an optional embodiment of this application, the task information and status information are input into the path planning model to obtain the path planning result, including: inputting the task information and status information into the global planning layer, the global planning layer performs a global path search on the task information and status information to generate the global optimal path; inputting the global optimal path into the local optimization layer, the local optimization layer dynamically adjusts the global optimal path to generate the path planning result.

[0009] In an optional embodiment of this application, the global planning layer performs a global path search on the task information and state information to generate a globally optimal path, including: obtaining the location information and energy consumption information in the state information, and constructing an objective function based on the task information, location information and energy consumption information; randomly generating an initial honey source set, which includes multiple honey sources, each honey source being a possible solution to the objective function, and each honey source corresponding to a candidate path; and using the artificial bee colony algorithm to iteratively optimize the initial honey source set until a preset convergence condition is met to obtain a globally optimal solution, which is then used as the globally optimal path.

[0010] In an optional embodiment of this application, the artificial bee colony algorithm is used to iteratively optimize the initial honey source set until a preset convergence condition is met. This includes: each iteration cycle comprising a hired bee phase, an observation bee phase, and a scout bee phase; during the hired bee phase, hired bees are generated based on the honey sources in the initial honey source set, with each hired bee corresponding to one honey source; the hired bees perform a neighborhood search in the solution space based on the corresponding honey source, updating the honey source position; and the fitness value of the updated honey source is calculated according to the objective function; during the observation bee phase, a preset probability selection mechanism is used to select the N honey sources with the highest fitness values ​​from all honey sources, where N is a preset integer; during the scout bee phase, honey sources whose fitness values ​​have not reached a preset threshold are randomly reset; the hired bee phase, observation bee phase, and scout bee phase are repeated until a preset maximum number of iterations is reached or the global optimal solution meets the convergence condition, where the global optimal solution is the honey source with the highest fitness value in each iteration cycle.

[0011] In an optional embodiment of this application, the local optimization layer dynamically adjusts the globally optimal path to generate path planning results, including: dividing the globally optimal path into multiple path segments, each path segment including at least one node; setting each node as a pheromone using an ant colony algorithm, the pheromone including concentration information, the concentration information being determined based on the globally optimal path; performing local iterative updates on each path segment, the local iterative update including: in the current path segment, selecting the next node based on the concentration information and a preset heuristic factor, and completing the connection between nodes; marking the connected nodes as updated path segments, and updating the concentration information of all nodes in the updated path segment; repeating the local iterative update until all path segments meet the optimization conditions; summarizing all updated path segments to form the path planning results.

[0012] In an optional embodiment of this application, controlling a warehouse robot to perform tasks based on path planning results includes: obtaining scene information of the warehouse robot; constructing a digital twin scene based on scene information and status information, the digital twin scene including a virtual map and a virtual robot; configuring control instructions corresponding to the path planning results to the virtual robot in the digital twin scene to drive the virtual robot to simulate operation in the virtual map; generating operation evaluation data based on the simulation operation results of the virtual robot; if the operation evaluation data indicates that the path planning results meet preset execution conditions, then the path planning results are sent to the actual warehouse robot for execution; if the operation evaluation data indicates that the path planning results do not meet the execution conditions, then the current path planning results are discarded and regenerated until the generated path planning results meet the execution conditions.

[0013] Secondly, embodiments of this application provide a computer device, including: a processor and a memory storing a computer program, wherein when the processor runs the computer program, the steps of the above-described method are implemented.

[0014] Thirdly, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described method.

[0015] The embodiments of this application have the following beneficial effects: The method provided in this application includes the following steps: acquiring task information and status information of a warehousing robot; inputting the task information and status information into a path planning model to obtain path planning results; the path planning model includes a global planning layer and a local optimization layer, where the global planning layer generates the globally optimal path, and the local optimization layer optimizes the globally optimal path locally to output the path planning results; and controlling the warehousing robot to execute tasks based on the path planning results. Therefore, this application can perform path planning using the task information and status information of the warehousing robot. The global planning layer and the local optimization layer in the path planning model employ different algorithms. By using a hybrid algorithm and leveraging the collaborative work of the global planning layer and the local optimization layer, the accuracy and adaptability of path generation are improved, effectively avoiding the inefficiency problem caused by traditional single-level planning.

[0016] The above description is merely an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it according to the contents of the specification, and to make the above and other objects, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. It should be understood that the above general description and the following detailed description are merely exemplary and explanatory, and do not limit this application. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application 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 only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating a path planning method provided in one embodiment.

[0019] Figure 2 This is a schematic block diagram of the structure of a computer device provided in one embodiment. Detailed Implementation

[0020] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. In the following description relating to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements.

[0021] 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, components, features, and elements with the same names in different embodiments of this application may have the same meaning or different meanings, the specific meaning of which must be determined by its interpretation in that specific embodiment or further in conjunction with the context of that specific embodiment.

[0022] It should be understood that although the terms first, second, third, etc., may be used herein to describe various information, such information should not be limited to these terms. These terms are used only to distinguish information of the same type from one another. For example, without departing from the scope of this document, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if," as used herein, can be interpreted as "when," "when," or "in response to determination." Furthermore, as used herein, the singular forms "a," "an," and "the" are intended to also include the plural forms unless the context indicates otherwise. It should be further understood that the terms "comprising," "including," indicate the presence of the stated feature, step, operation, element, component, item, kind, and / or group, but do not exclude the presence, occurrence, or addition of one or more other features, steps, operations, elements, components, items, kinds, and / or groups. The terms "or" and "and / or" as used herein are to be interpreted as inclusive, or mean any one or any combination thereof. Therefore, "A, B, or C" or "A, B, and / or C" means "any one of the following: A; B; C; A and B; A and C; B and C; A, B, and C". Exceptions to this definition will only occur if the combination of elements, functions, steps, or operations is inherently mutually exclusive in some way.

[0023] It should be understood that although the steps in the flowcharts of this application's embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some of the steps in the figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0024] It should be noted that step designations such as S110 and S120 are used in this document for the purpose of more clearly and concisely describing the corresponding content, and do not constitute a substantial limitation on the order. In specific implementation, those skilled in the art may execute S120 first and then S110, etc., but these should all be within the protection scope of this application.

[0025] It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.

[0026] In the following description, the use of suffixes such as "module," "part," or "unit" to denote elements is solely for the purpose of illustrative purposes and has no specific meaning in itself. Therefore, "module," "part," or "unit" may be used interchangeably.

[0027] To facilitate understanding of this application, the following explanations are provided for the terms and technical objects that may be used in this application: AGV (Automated Guided Vehicle): A transport robot equipped with electromagnetic or optical automatic guidance devices. It is a mobile robot capable of autonomously driving and completing material handling tasks without driver intervention, relying on an automatic navigation system. It is commonly used in factories, warehousing and logistics facilities, ports, hospitals, and other similar settings to perform tasks such as handling, delivery, picking, and loading / unloading.

[0028] Artificial Bee Colony (ABC) is a swarm intelligence optimization algorithm that simulates the foraging behavior of bees. This algorithm uses food sources as candidate solutions and nectar source quality as the objective function value. It searches for the optimal solution through an iterative update mechanism. It features simple structure, few parameters, and strong global search capabilities, and is widely used in path planning, parameter optimization, image processing, and other fields.

[0029] Ant Colony Optimization (ACO) is a heuristic optimization algorithm that simulates the communication and cooperation of ants through pheromones during foraging. This algorithm correlates path selection probability with pheromone concentration and utilizes a positive feedback mechanism to gradually converge to a better solution, demonstrating good performance in solving combinatorial optimization problems such as the Traveling Salesman Problem, network routing, and task scheduling.

[0030] Digital twins refer to simulation processes that integrate multiple disciplines, physical quantities, scales, and probabilities, utilizing data such as physical models, sensor updates, and operational history to map physical equipment in virtual space, thereby reflecting the entire lifecycle of the corresponding physical equipment. They can synchronize the state, behavior, and situation of their physical counterparts in real time and guide the physical entity's decision-making and optimization through simulation and prediction.

[0031] Existing AGV scheduling systems mainly suffer from the following problems: 1. Static resource allocation: Most systems use fixed resource pools, which cannot cope with AGV idleness or overload caused by sudden changes in order volume. 2. Strong communication dependence: Laser-guided AGVs rely on a continuous network connection, and scheduling fails when the network is down. 3. Insufficient predictive ability: Existing digital twin applications are only used for real-time monitoring and do not achieve dynamic resource pre-allocation. 4. Low recovery efficiency: Fault recovery schemes require manual intervention and cannot autonomously reconstruct scheduling strategies. To address these issues, this application proposes a path planning method that integrates digital twins and swarm intelligence algorithms, enabling dynamic optimization and autonomous decision-making in the AGV scheduling system. For a clear description of the method provided in this embodiment, please refer to... Figure 1 This includes steps S110 to S130.

[0032] This application provides a path planning method, which can be executed by a computer device provided in this application embodiment. This device can be implemented in software and / or hardware. Specifically, the device can be a server with higher computing power, which communicates with a warehouse robot (hereinafter referred to as AGV) to obtain relevant information, processes it, and then feeds back the processed path planning result to the AGV for dynamic path adjustment. Alternatively, the executing entity can be a high-performance controller built into the AGV itself, with the AGV acting as an edge device to complete the path planning process autonomously. The specific implementation depends on actual needs and the distribution of computing resources, and is not limited thereto. For ease of understanding, this application describes the executing entity as a server.

[0033] The server establishes a wired or wireless communication connection with the AGV. Wireless communication technologies may include, but are not limited to: Global System for Mobile Communication (GSM), Enhanced Data GSM Environment (EDGE), Wideband Code Division Multiple Access (W-CDMA), Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA), Bluetooth, Wireless Fidelity (WiFi) (such as IEEE 802.11a, IEEE 802.11b, IEEE 802.11g and / or IEEE 802.11n standards), Voice over Internet Protocol (VoIP), Worldwide Interoperability for Microwave Access (Wi-Max), other protocols for email, instant messaging, and short messages, and any other suitable communication protocols, even those not yet developed. The AGV is equipped with a multimodal perception system consisting of LiDAR, depth cameras, and infrared sensors to collect real-time information on obstacle locations, dynamic pedestrian trajectories, and shelf status in the warehouse environment. It may also include environmental sensing devices (such as LiDAR, UWB base stations, and vision sensors) to collect real-time physical environment data; this collected data is collectively referred to as status information. Furthermore, the AGV includes a communication module capable of establishing a communication connection with a server and uploading the collected status information to the server in real time. The AGV can also be configured with an edge computing unit to maintain basic obstacle avoidance and path planning capabilities even when communication with the server is interrupted. The edge computing unit can run lightweight models to quickly respond to changes in the local environment, ensuring operational safety and task continuity. The method provided in this application, executed by the server, includes the following steps.

[0034] Step S110: Obtain the task information and status information of the warehouse robot.

[0035] In one embodiment, the task information includes at least one of actual task information and predicted task information. Step S110: Obtaining task information of the warehouse robot includes: obtaining actual task information sent to the warehouse robot by the host computer; and / or, obtaining historical task information of the warehouse robot, inputting the historical task information into the task prediction model to obtain predicted task information.

[0036] In one implementation, the task information records the type of operation the AGV needs to perform, the target shelf number, the goods information, and the priority. This task information guides the AGV's path planning and action execution. Task information can include two types: actual task information and predicted task information. Actual task information is issued in real-time by the host computer system based on current order requirements, with clear operational instructions and timeliness requirements. Predicted task information, on the other hand, is derived through learning and analysis of historical task information, representing a sequence of tasks the AGV may execute within a specific future time period. Predicted task information can be output through a task prediction model. This model can employ an LSTM network architecture and incorporate an attention mechanism, effectively enhancing the model's ability to extract key path features and improving prediction accuracy. Using historical task information as input, the model learns from the temporal features and spatial distribution patterns of historical tasks, capturing the temporal regularity and regional hotspots of task occurrences, and outputting the probability distribution of tasks over a future period, ultimately producing the predicted task information.

[0037] Both actual task information and predicted task information can be used as input for task planning, guiding AGV path pre-planning and resource reservation. Actual task information serves as the primary basis for path planning, while predicted task information can be used to assist in optimizing the global scheduling strategy, improving the system's responsiveness to unexpected tasks and overall operational efficiency.

[0038] In one embodiment, step S110: obtaining task information and status information of the warehouse robot includes: establishing a network connection with the warehouse robot and dividing the network into a first sub-network, a second sub-network and a third sub-network; obtaining map data in the status information through the first sub-network; obtaining task information through the second sub-network and storing the task information; and obtaining sensor data in the status information through the third sub-network.

[0039] In one embodiment, as described above, when the executing entity is a server, the server establishes a communication connection with the AGV. In a preferred embodiment, network connection can be achieved via 5G communication. With 5G communication, task information and status information can be obtained from the AGV using 5G network slicing. 5G network slicing divides the network into a first sub-network, a second sub-network, and a third sub-network.

[0040] Specifically, the first sub-network can be an eMBB (Enhanced Mobile Broadband) slice, providing ultra-high bandwidth and high speed, which can be used to obtain map data in status information, specifically point cloud data collected by AGVs through devices such as LiDAR.

[0041] The second sub-network can be a URLLC (Ultra-Reliable and Low Latency Communications) slice, providing extremely low latency and high reliability, and can be used to acquire task information. After acquisition, the task information can be stored as evidence. Specifically, the task information can be uploaded to the blockchain for storage, thereby preventing malicious nodes from tampering with scheduling instructions. For the blockchain evidence storage method, considering that all AGVs are edge computing devices, a decentralized evidence storage mechanism can be adopted, with each AGV node jointly maintaining the blockchain ledger of task information to ensure that the data is immutable and traceable.

[0042] The third sub-network can be an mMTC (Massive Machine Type Communications) slice, supporting the access of massive numbers of devices to obtain sensor data in status information, including but not limited to AGV position, power, load, motor temperature, battery degradation rate, etc.

[0043] Furthermore, network slicing can not only be used to acquire task and status information, but also, after obtaining path planning results, leverage the extremely low latency and high reliability of the second sub-network within the network slice to issue commands to the AGV. Interaction with the AGV via network slicing ensures that different types of data transmission needs are accurately met, improving overall communication efficiency and system stability. The isolation and specialization of network slicing effectively avoids data congestion and interference, further guaranteeing the collaborative stability of the AGV cluster in high-density operating environments. Simultaneously, the dynamic resource allocation capability between different slices allows the system to adjust bandwidth and latency parameters in real time according to task priority, ensuring timely response and execution of critical instructions. Compared to existing scheduling schemes, latency is reduced by over 90%. The blockchain-based notarization mechanism for task information ensures the immutability and traceability of scheduling instructions through distributed node verification. Upon task completion, a smart contract is automatically triggered, recording the execution on the blockchain, thus constructing an end-to-end trusted scheduling system. This mechanism effectively prevents network attacks and data forgery risks, improving the security level and operational reliability of the AGV system in complex industrial scenarios.

[0044] Step S120: Input the task information and status information into the path planning model to obtain the path planning result.

[0045] In one embodiment, task information and status information are input into the path planning model to obtain path planning results, including: inputting task information and status information into the global planning layer, the global planning layer performs a global path search on the task information and status information to generate a globally optimal path; inputting the globally optimal path into the local optimization layer, the local optimization layer dynamically adjusts the globally optimal path to generate path planning results.

[0046] In one embodiment, the path planning model proposed in this application includes a two-layer architecture: a global planning layer and a local optimization layer. The global planning layer generates the globally optimal path, while the local optimization layer optimizes the globally optimal path locally to output the path planning result. The two layers employ different algorithms, using a hybrid algorithm and collaborative processing approach to ensure the efficiency and real-time performance of the path planning. In a preferred embodiment, the global planning layer uses an artificial bee colony algorithm for global path search, while the local optimization layer uses an ant colony algorithm to optimize the candidate paths output by the global planning layer locally. The implementation processes of each will be described in detail later.

[0047] In one embodiment, the global planning layer performs a global path search on the task information and state information to generate a globally optimal path, including: obtaining the location information and energy consumption information from the state information, and constructing an objective function based on the task information, location information, and energy consumption information; randomly generating an initial honey source set, which includes multiple honey sources, each honey source being a possible solution to the objective function, and each honey source corresponding to a candidate path; and iteratively optimizing the initial honey source set using the artificial bee colony algorithm until a preset convergence condition is met to obtain a globally optimal solution, which is then used as the globally optimal path.

[0048] In one implementation, the global planning layer utilizes the artificial bee colony algorithm for multi-objective dynamic optimization. Through a global optimization process, it finds an "optimal" initial or corrected path solution set, i.e., the globally optimal path. The global planning layer needs to use the AGV's state and task information to construct the objective function. Specifically, it needs to obtain the position and energy consumption information from the state information. The position information is the current coordinates of the AGV on the map; the energy consumption information can include the AGV's battery level, energy consumption per unit distance, etc.

[0049] Based on task information, location information, and energy consumption information, an objective function is constructed. The objective function is a weighted multi-objective function, and its goal is to minimize the combined values ​​of the following three items: total driving distance, total energy consumption, and task completion time.

[0050] The total travel distance is the total mileage of all AGVs completing their tasks, and the calculation method is as follows.

[0051] (1) In the above formula, For time; To determine the location coordinates of the AGV at time t based on the candidate paths. The predicted path is essentially the intermediate value calculated during the iteration process. The objective function targets the total travel distance. The optimization objective is to minimize the total driving distance.

[0052] The number of turns is determined for each candidate path. and number of accelerations and decelerations The number of turns It can be configured to count when the AGV's heading angle changes beyond a threshold (e.g., 15°) in adjacent time intervals, based on candidate paths. This applies to acceleration and deceleration counts. The counting method can be based on the candidate path, determining that a count occurs when the AGV's speed change exceeds a set threshold (e.g., 0.5 m / s) between adjacent time points. This is achieved by counting the number of turns. With the number of accelerations and decelerations Calculate total energy consumption The calculation method can be found in the following formula.

[0053] (2) In the above formula, and The preset energy consumption coefficient is set to a specific value based on the energy consumption situation in the actual scenario, for example... =0.3, =0.2.

[0054] Total task time is the total time taken for all AGVs from receiving the task to completing it. The calculation method can be found in the following formula.

[0055] (3) In the above formula, Total task time; Indicates the AGV number. Total number of AGVs; For the number The length of the candidate path for an individual can be determined directly from the candidate path; For the number The average speed of an individual can be determined directly using state information.

[0056] Therefore, after calculating and determining the combined value of the three terms, the objective function can be constructed, which can be expressed as the following formula.

[0057] (4) In the above formula, ~ These are the preset weighting coefficients. The objective function is used to generate new solutions and calculate fitness values, thereby evaluating the merits of each honey source location. After setting the objective function, iteration can begin. Initially, an initial set of honey source solutions is randomly generated, with each honey source representing a feasible path scheme. Each honey source is a possible solution of the objective function.

[0058] In one embodiment, an artificial bee colony algorithm is used to iteratively optimize the initial nectar source set until a preset convergence condition is met. This includes: each iteration cycle comprising a hired bee phase, an observation bee phase, and a scout bee phase; during the hired bee phase, hired bees are generated based on the nectar sources in the initial set, with each hired bee corresponding to one nectar source; the hired bees perform a neighborhood search in the solution space based on the corresponding nectar source, updating the nectar source position; and the fitness value of the updated nectar source is calculated according to the objective function; during the observation bee phase, a preset probability selection mechanism is used to select the N nectar sources with the highest fitness values ​​from all nectar sources, where N is a preset integer; during the scout bee phase, nectar sources whose fitness values ​​have not reached a preset threshold are randomly reset; the hired bee phase, observation bee phase, and scout bee phase are repeated until a preset maximum number of iterations is reached or the global optimal solution meets the convergence condition, where the global optimal solution is the nectar source with the highest fitness value in each iteration cycle.

[0059] In one implementation, the iterative process based on the artificial bee colony algorithm can be divided into three stages: the hired bee stage, the observation bee stage, and the scout bee stage. In the hired bee stage, hired bees are generated based on the nectar sources in the initial nectar source set, with each hired bee corresponding to one nectar source. These hired bees, along with the subsequent observation and scout bees, are virtual individuals simulated by the algorithm, used to perform local neighborhood searches of nectar source locations. Hired bees are used to adjust the order of path nodes or replace some nodes, thereby generating new solutions, and fitness values ​​are calculated based on the objective function to evaluate their performance. At the beginning of the iteration, hired bees are generated using the initial nectar source set to perform local neighborhood searches. In subsequent iterations, the nectar sources are updated, forming a new nectar source set. Similarly, the updated nectar source set is used as the basis for generating hired bees to continue neighborhood searches and fitness evaluations. For ease of description, the nectar source set updated in the current and subsequent iterations is collectively referred to as the initial nectar source set. That is to say, the initial nectar source set represents the set of nectar sources in the update process, not limited to the initially randomly selected set of nectar sources.

[0060] In the subsequent observation phase, a pre-defined probability selection mechanism, such as a roulette wheel selection mechanism based on fitness values, is used to select superior nectar sources from the nectar source set. Fitness values ​​characterize nectar source quality; nectar source quality and fitness values ​​are positively correlated—the higher the fitness value, the better the nectar source quality and the greater the probability of it being selected. In each iteration, N superior nectar sources can be selected for further development by the observation bees, where N is a pre-defined integer. Based on the selected superior nectar sources, the observation bees perform a more refined neighborhood search to further optimize the combination of path nodes.

[0061] During the scout bee phase, if the fitness value of a nectar source remains below a preset threshold after multiple iterations, it is determined that it has fallen into a local optimum. The nectar source is then abandoned, and a new nectar source is randomly generated by the scout bee to replace it, thereby enhancing population diversity. Specifically, this could involve changing the scout bee into a hired bee, which would then randomly search for new nectar sources.

[0062] Each iteration consists of repeating the hiring bee phase, the observation bee phase, and the scout bee phase. By cyclically executing these three phases, the nectar source set is continuously updated and the path solution is optimized. Each nectar source in the iteration process corresponds to a path solution, which consists of a series of ordered nodes representing a possible travel route, called a candidate path, and is used for calculating the fitness value and updating the path. After each iteration, the nectar source with the highest fitness value in the current round can be evaluated to determine whether it meets the preset convergence conditions, such as the fitness value of the nectar source being greater than the preset termination threshold, or the fitness value of the nectar source not changing significantly for multiple consecutive rounds. If the convergence conditions are met, the iteration terminates, and the nectar source with the highest fitness value is output as the global optimal solution. Alternatively, the iteration count can be recorded, and when the iteration count reaches the preset maximum number of iterations, regardless of whether there is a significant improvement, the iteration stops and the current optimal nectar source solution is output. Conversely, if the above conditions are not met, it means that the requirements have not been met, and the algorithm will continue to the next round of iteration until the global optimal solution is found.

[0063] Furthermore, before being output to the local optimization layer, the global optimal solution undergoes smoothing and feasibility verification to ensure its executability in a real-world operating environment. Smoothing can be achieved by smoothing path nodes using interpolation algorithms or curve fitting methods, eliminating unnecessary inflection points and reducing turning energy consumption and travel resistance. Feasibility verification includes checking whether the path avoids obstacles, whether there are path conflicts (multiple AGVs simultaneously traversing the same road segment), and whether it meets task constraints (such as whether the task's start and end points match the task information). Only global optimal solutions that pass the feasibility verification are passed to the local optimization layer for further processing; otherwise, they are regenerated until they pass the feasibility verification, thus ensuring the reliability and practicality of the global optimal solution.

[0064] In one embodiment, the local optimization layer dynamically adjusts the globally optimal path to generate path planning results, including: dividing the globally optimal path into multiple path segments, each path segment including at least one node; setting each node as a pheromone using an ant colony algorithm, the pheromone including concentration information determined according to the globally optimal path; performing local iterative updates on each path segment, the local iterative update including: in the current path segment, selecting the next node according to the concentration information and a preset heuristic factor, and completing the connection between nodes; marking the connected nodes as updated path segments, and updating the concentration information of all nodes in the updated path segment; repeating the local iterative update until all path segments meet the optimization conditions; and summarizing all updated path segments to form the path planning results.

[0065] In one implementation, the local optimization layer is used to perform local optimization and specific path generation on the globally optimal solution output by the global planning layer. Specifically, this can be achieved using the pheromone mechanism in the ant colony algorithm to find the specific shortest path in the graph network. Therefore, the input to the local optimization layer is the globally optimal solution output by the global planning layer, rather than directly obtaining the raw data of the AGV.

[0066] The process of local optimization is essentially a fine-tuning of the locally optimal path, enhancing its adaptability and robustness by dynamically updating pheromone concentration. Therefore, the globally optimal path can first be divided into multiple path segments, each containing at least one node. Nodes represent key locations on the path, such as turns, intersections, or task execution points. Connecting these nodes in sequence yields a path segment. The sum of these path segments represents the planned path corresponding to the globally optimal solution; in other words, a path segment is a local representation of the planned path. The division of the globally optimal path is not restricted; it can be an equal-length division or a dynamic division based on key nodes, flexibly set according to the complexity of the actual scenario without specific limitations.

[0067] The local optimization layer can proceed sequentially along path segments, optimizing each segment individually. The optimization process simulates the behavior of ants moving between nodes, using an ant colony algorithm to process each node within a path segment, assigning each node a pheromone, including its concentration. The concentration is determined based on the globally optimal path; specifically, the pheromone concentration is proportional to the nectar quality in the ant colony algorithm, which, as mentioned earlier, is determined by the objective function. Higher nectar quality leaves a higher pheromone concentration on its path, thus guiding the ant colony algorithm to favor these paths.

[0068] The local iterative update within a path segment involves selecting the next node based on concentration information and a preset heuristic factor. Higher pheromone concentrations increase the probability of an ant selecting that node. Similarly, ants are virtual individuals simulated in the algorithm to model candidate path selection during path search. Each ant moves between adjacent nodes according to probability rules, releasing pheromones along the paths it traverses. As iterations proceed, the pheromone concentration on better paths gradually increases, while the concentration on inferior paths decreases due to an evaporation mechanism. This achieves dynamic optimization and convergence of local paths. Connecting the updated information yields the optimized path segment, which is then marked as the updated path segment. During iteration, the concentration information corresponding to all pheromones decreases with the number of iterations; the pheromone evaporation mechanism ensures the algorithm does not stagnate in local optima, while positive feedback reinforces high-quality paths.

[0069] The process involves repeated local iterative updates to optimize node positions within a path segment until the segment meets the optimization criteria. The same optimization process is then applied to the next path segment. Optimization criteria can be determined based on changes in path segment curvature, consistency of node spacing, or algorithm convergence thresholds. A path segment is considered to meet the optimization criteria when the path length optimization between adjacent iterations is less than a preset threshold and smoothness constraints are satisfied. After completing local optimization for all path segments, the updated path segments are sequentially concatenated to form a complete optimized path, which is then marked as the path planning result.

[0070] In one implementation, the global planning layer uses the artificial bee colony algorithm, and the local optimization layer uses the bee colony algorithm. These two algorithms can also be replaced with other intelligent optimization algorithms, such as particle swarm optimization or genetic algorithms, adapted to the specific application scenario. However, in a preferred implementation, a hybrid algorithm mechanism combining the artificial bee colony algorithm and the ant colony algorithm is used. The artificial bee colony algorithm can complete a coarse global path planning within seconds (approximately 3 seconds in experiments), and typically outputs multiple globally optimal paths (e.g., 3). The ant colony algorithm then refines the path segments planned by the artificial bee colony algorithm, similarly completing local path adjustment and optimization within seconds (approximately 6 seconds in experiments).

[0071] The core innovation of this application lies in the hybridization of the artificial bee colony algorithm and the ant colony algorithm. Their influence is bidirectional, but primarily the results of the bee colony algorithm guide the optimization direction of the ant colony algorithm. The hybrid algorithm can produce the following effects: First, accelerated convergence: The ant colony algorithm no longer needs to start from zero and search randomly, but directly performs local optimization on the "good paths" selected by the bee colony algorithm. This greatly accelerates the convergence speed and improves efficiency. Second, avoidance of local optima: As a global optimizer, the artificial bee colony algorithm can escape the local optimum traps that the ant colony algorithm may fall into. It provides multiple high-quality initial solutions, allowing the ant colony algorithm to start optimization from different starting points, thus increasing the likelihood of finding the global optimum. Third, unconstrained intervention: The results of the artificial bee colony algorithm do not directly constrain the ant colony algorithm. It does not "force" ants to take a certain path, but guides and enhances the probability of ants choosing a specific path by changing the pheromone concentration.

[0072] Although the primary impact is unidirectional, the local optimization results of the ant colony algorithm can also serve as feedback to adjust the iterative strategy of the bee colony algorithm. For example, if the ant colony algorithm consistently fails to find high-quality paths in a certain area, it may indicate insufficient exploration by the bee colony algorithm in that area, allowing it to adjust the search strategy of its scout bees. In summary, the bee colony algorithm is responsible for "site selection," finding the most promising AGV path combinations, while the ant colony algorithm is responsible for "road construction," performing refined path planning within the potential areas identified by the bee colony algorithm, thus achieving an efficient and robust distributed AGV intelligent scheduling method.

[0073] It can now be proven that the convergence speed of this hybrid algorithm is improved. Let the solution space of the AGV scheduling problem be... The optimal solution set is The algorithm's iterative process is modeled as a Markov chain. ,in Indicates the first The candidate path set at each iteration. In the bee colony stage, m initial candidate paths (bee role) are generated through global exploration, covering key regions of the solution space. In the ant colony stage, the candidate paths are locally optimized based on a pheromone feedback mechanism (ant role), updating the pheromone matrix. Pheromone matrix Represented as time At that time, path segment Information on pheromone concentration.

[0074] To analyze the exploration capability of the global planning layer, Lemma 1 needs to be introduced. Lemma 1 specifically relates to the coverage probability of the artificial bee colony algorithm, which states that in the bee colony stage, if the initial number of candidate paths satisfies condition 1, then at least one path will enter the target area. The probability of the inner optimal solution. Let be the neighborhood coverage probability threshold, which is also the target coverage accuracy. Condition 1 can be expressed as the following formula.

[0075] (5) In the above formula, This represents the initial number of candidate paths in the bee colony stage, which is also the number of parallel searches. This represents the probability that a single bee colony search covers the neighborhood of the optimal solution. The probability of the optimal solution can be expressed as follows.

[0076] (6) The proof is as follows: the probability that a single bee colony search does not cover the optimal neighborhood is... ,but The probability that none of the independent searches covered the area is: So, to satisfy... Then, condition 1 can be obtained directly.

[0077] For the convergence rate analysis of the local optimization layer, Lemma 2 can be introduced. Lemma 2 is the pheromone-driven convergence rate, which can be specifically defined by setting the pheromone update rule, expressed as follows.

[0078] (7) In the above formula, Let be the preset pheromone evaporation coefficient. Then, the expected time for the local optimization layer to converge to a local optimum satisfies the following equation.

[0079] (8) In the above formula, The expected time value for a local optimum; and These represent the maximum and minimum values ​​of pheromone concentration information, respectively.

[0080] The proof can be made by first constructing the potential function, which is expressed as the following formula.

[0081] (9) In the above formula, For state The steady-state probability is calculated. The decay rate of the potential function, analyzed through pheromone difference, is expressed as follows.

[0082] (10) In the above formula, Let be the parameter representing the difference in path quality. The expected convergence time can be derived using the geometric series summation formula, thus completing the proof of Lemma 2.

[0083] The proof of the accelerated convergence of the bee colony-ant colony hybrid algorithm can be based on a joint analysis of Lemmas 1 and 2. First, Lemma 1 is defined, characterizing the convergence speed advantage of the hybrid algorithm. Specifically, the expected convergence time of the bee colony-ant colony hybrid algorithm... The following expression is satisfied.

[0084] (11) Expected convergence time of traditional ant colony algorithm The following expression is satisfied.

[0085] (12) A comparison of equations (11) and (12) shows that the convergence speed of the hybrid algorithm is improved by a factor of [number missing]. The following expression is satisfied.

[0086] (13) It can be seen from equation (13) that, due to Therefore, the convergence speed is increased by a factor of 1. It will be significantly improved.

[0087] For the proof process, the initial cover will be accelerated in the bee colony algorithm stage of the global planning layer, based on Lemma 1, through... The parallel search quickly locates the neighborhood of the optimal solution, optimizing the initial state to... Approaching optimality. In the local optimization layer, ant colony optimization is used to accelerate local convergence, reaching a point where the target value has been narrowed down. In the solution space of the neighborhood, the pheromone update mechanism guides path optimization with a higher probability, and this process can be represented by the following formula.

[0088] (14) The total expected time of the bee colony-ant colony hybrid algorithm is superimposed, and the total time of the hybrid algorithm is obtained from the bee colony pre-search. Local optimization of ant colonies Combined with composition, it achieves superlinear acceleration.

[0089] Based on this, the numerical simulation comparison can be verified through experiments, as shown in Table 1.

[0090] Table 1

[0091] Experimental data show that the hybrid algorithm significantly outperforms the traditional ant colony algorithm in both convergence speed and stability. Under the same test environment, the average number of convergence iterations of the bee colony-ant colony hybrid algorithm is reduced by approximately 56%, and the time required to reach the target solution is shortened by nearly 60%. The correctness of Lemmas 1 and 2 is verified by combining theoretical derivation and simulation results, further supporting the validity of Theorem 1. The influence of key parameters is as follows.

[0092] (15) As can be seen from equation (15) and Table 1, the experimental results are consistent with the theoretical predictions. The decay trend is evident. Therefore, the bee colony-ant colony hybrid algorithm, which combines the fast response of bee colonies with the local optimization of ant colonies, improves path planning speed by 3 times compared to the traditional ant colony algorithm, while reducing congestion rate by 25%.

[0093] Step S130: Control the warehouse robot to perform tasks based on the path planning results.

[0094] In one embodiment, controlling a warehouse robot to perform tasks based on path planning results includes: acquiring scene information of the warehouse robot's location; constructing a digital twin scene based on the scene information and status information, the digital twin scene including a virtual map and a virtual robot; configuring control instructions corresponding to the path planning results to the virtual robot in the digital twin scene to drive the virtual robot to simulate operation in the virtual map; generating operation evaluation data based on the simulation operation results of the virtual robot; if the operation evaluation data indicates that the path planning results meet preset execution conditions, then sending the path planning results to the actual warehouse robot for execution; if the operation evaluation data indicates that the path planning results do not meet the execution conditions, then discarding the current path planning results and regenerating them until the generated path planning results meet the execution conditions.

[0095] In one implementation, the path planning results can also be verified before being sent to the AGV control. The verification process can take place in a digital twin virtual environment, therefore, a digital twin scene needs to be constructed first. This requires acquiring the scene information where the AGV is located. Scene information can be divided into fixed and variable information. Fixed information refers to the unchanging parts, such as shelf layout and road boundaries, which can be directly obtained from map data of the AGV's scene. Variable information includes dynamic obstacles and the positions of other robots, which need to be determined by the point data in the AGV's collected status information. Finally, the digital twin scene is constructed using the scene information and status information. The digital twin scene includes a virtual map and a virtual robot. The virtual map is constructed based on the point cloud data in the scene information and status information, while the virtual robot is simulated using the AGV's status information.

[0096] The simulation runs the virtual robot's path planning results within a virtual map. This involves configuring the path planning results as control commands and sending them to the virtual robot to drive its movement within the virtual map. The simulation monitors the robot's operational status and collects operational evaluation data. This evaluation data may include, but is not limited to, path length, obstacle avoidance success rate, task completion time, and number of conflicts. The evaluation data is then analyzed to determine if preset execution conditions are met. These conditions include indicators such as path safety, task timeliness, and resource utilization, which are directly related to the data included in the evaluation data. Specific conditions are set according to the needs of the actual application scenario; therefore, no specific conditions are defined here.

[0097] If the preset execution conditions are met, the path planning result is sent to the AGV to execute the task; if not, the current path planning result is discarded and regenerated until the generated path planning result meets the execution conditions.

[0098] Furthermore, dynamic resource pre-allocation can be implemented in digital twin scenarios. As mentioned earlier, task information may be predicted through predictive models, thus allowing for the forecasting of task demands over a future period. Based on these predictions, virtual scheduling and resource reservation can be performed within the digital twin scenario, including pre-allocating AGVs, charging stations, and path resources for upcoming tasks to avoid resource contention and scheduling conflicts. Dynamic resource pre-allocation can expand the virtual resource pool in advance when peak task periods are predicted, thereby better simulating resource scarcity in real-world situations and improving the overall system scheduling efficiency and response speed. Through dynamic resource pool allocation in digital twins, resource pre-allocation and virtual-real linkage are achieved. Simulation tests show that this solution improves the ability to handle sudden order surges by 50%, a significant improvement compared to existing monitoring-based digital twin technologies.

[0099] In one embodiment, the preceding examples described a path planning method with a server as the execution entity. In this embodiment, an AGV (Automated Guided Vehicle) is used as the execution entity for a simplified description. When using an AGV as the execution entity, the AGV needs to act as an edge computing device, requiring sufficient computing power and storage resources to run a lightweight path planning model. The AGV can be equipped with a local AI chip to run the lightweight path planning model. This lightweight model can extract core capabilities from the original large model using techniques such as model distillation and teacher-student model compression, reducing computational overhead while maintaining path planning accuracy.

[0100] This implementation can serve as a supplement or alternative to server-based execution, and can be implemented in situations with high communication latency or unstable networks. For example, when the AGV detects an interruption in its communication connection with the server or a latency exceeding a threshold, it automatically switches to the local lightweight model for real-time path planning, ensuring the continuity and safety of task execution. The execution process maintains consistency with the server, and task and status information must still be updated to the local model in real time to ensure the accuracy of path planning. For specific execution procedures, please refer to the preceding description; they will not be repeated here. After completing path planning, the local lightweight model immediately outputs control commands to drive the AGV's movement and uses sensors to provide real-time feedback on environmental changes, dynamically adjusting its travel strategy.

[0101] Furthermore, this compensation mechanism applies not only to AGVs but also to the server. When the server detects that an AGV is offline, it can immediately activate the virtual resource pool compensation mechanism. This mechanism allocates nearby AGVs, retrieves the unfinished tasks of the faulty equipment via blockchain, and quickly generates a new path based on digital twin simulation. This mechanism ensures seamless task continuity and avoids operational interruptions.

[0102] Furthermore, it's understandable that the interruption of communication between the AGV and the server is not merely a disruption of path planning capabilities, but also affects the continuity of task execution and the coordination of overall scheduling. To address this, after equipment recovery, local execution logs are synchronized to the server for iterative optimization of the global model, forming a closed-loop learning mechanism. This design balances real-time performance and intelligence, significantly improving autonomous decision-making capabilities and system robustness in complex warehousing environments. Through the deep integration of local decision-making and cloud collaboration, the system achieves efficient response and resilient fault tolerance in dynamic environments.

[0103] This application can also be applied to the offline self-healing capability of AGVs. When an AGV detects an abnormal shutdown or task interruption, it can still complete basic scheduling based on local AI, such as replanning the route, avoiding obstacles, or returning to the charging station, ensuring equipment safety and task recoverability. Compared to existing technologies, the response time is reduced by more than 80%, and a higher task recovery success rate is achieved in simulation tests.

[0104] Therefore, the method provided in this application includes the following steps: obtaining task information and status information of a warehousing robot; inputting the task information and status information into a path planning model to obtain path planning results; the path planning model includes a global planning layer and a local optimization layer, the global planning layer is used to generate the globally optimal path, and the local optimization layer is used to perform local optimization on the globally optimal path to output path planning results; controlling the warehousing robot to execute tasks according to the path planning results. Therefore, this application can perform path planning using the task information and status information of a warehousing robot. The global planning layer and the local optimization layer in the path planning model employ different algorithms. By using a hybrid algorithm and utilizing the collaborative work of the global planning layer and the local optimization layer, the accuracy and adaptability of path generation are improved, effectively avoiding the inefficiency problem caused by traditional single-level planning.

[0105] Figure 2 An internal structural diagram of a computer device in one embodiment is shown. This computer device can specifically be a terminal or a server. Figure 2 As shown, the device includes: a processor 310 and a memory 311 storing a computer program; wherein, Figure 2 The processor 310 shown in the diagram does not indicate that there is only one processor 310, but only indicates the positional relationship of the processor 310 relative to other devices. In practical applications, there can be one or more processors 310; similarly, Figure 2 The memory 311 illustrated herein has the same meaning, that is, it is only used to indicate the positional relationship of memory 311 relative to other devices. In practical applications, there can be one or more memories 311. When the processor 310 runs the computer program, the method applied to the above-mentioned device is implemented.

[0106] The device may also include at least one network interface 312. The various components of the device are coupled together via a bus system 313. It is understood that the bus system 313 is used to implement communication between these components. In addition to a data bus, the bus system 313 also includes a power bus, a control bus, and a status signal bus. However, for clarity, in... Figure 2 The general designated all buses as Bus System 313.

[0107] The memory 311 can be volatile memory or non-volatile memory, or both. Non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), ferromagnetic random access memory (FRAM), flash memory, magnetic surface memory, optical disc, or compact disc read-only memory (CD-ROM); magnetic surface memory can be disk storage or magnetic tape storage. Volatile memory can be random access memory (RAM), used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Synchronous Static Random Access Memory (SSRAM), Dynamic Random Access Memory (DRAM), Synchronous Dynamic Random Access Memory (SDRAM), Double Data Rate Synchronous Dynamic Random Access Memory (DDRSDRAM), Enhanced Synchronous Dynamic Random Access Memory (ESDRAM), SyncLink Dynamic Random Access Memory (SLDRAM), and Direct Rambus Random Access Memory (DRRAM).The memory 311 described in the embodiments of the present invention is intended to include, but is not limited to, these and any other suitable types of memory.

[0108] The memory 311 in this embodiment of the invention is used to store various types of data to support the operation of the device. Examples of this data include: any computer programs used to operate on the device, such as operating systems and applications; contact data; phonebook data; messages; pictures; videos, etc. The operating system includes various system programs, such as the framework layer, core library layer, driver layer, etc., used to implement various basic services and handle hardware-based tasks. Applications can include various applications, such as media players, browsers, etc., used to implement various application services. Here, the program implementing the method of this embodiment of the invention can be included in the application.

[0109] Based on the same inventive concept as the foregoing embodiments, this embodiment also provides a computer-readable storage medium storing a computer program. The computer-readable storage medium can be a magnetic random access memory (FRAM), a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a flash memory, a magnetic surface memory, an optical disc, or a compact disc read-only memory (CD-ROM), etc.; it can also be various devices including one or any combination of the above-mentioned memories, such as mobile phones, computers, tablet devices, personal digital assistants, etc. When the computer program stored in the computer-readable storage medium is run by a processor, it implements the above method. For the specific steps implemented when the computer program is executed by the processor, please refer to [link to relevant documentation]. Figure 1 The description of the illustrated embodiments will not be repeated here.

[0110] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0111] In this document, the terms “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, which includes not only the elements listed but also other elements not expressly listed.

[0112] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A path planning method, characterized in that, Includes the following steps: Obtain task and status information of the warehouse robot; The task information and status information are input into the path planning model to obtain the path planning result. The path planning model includes a global planning layer and a local optimization layer. The global planning layer is used to generate the globally optimal path, and the local optimization layer is used to perform local optimization on the globally optimal path to output the path planning result. The warehouse robot is controlled to perform tasks based on the path planning results.

2. The path planning method as described in claim 1, characterized in that, The task information includes at least one of actual task information and predicted task information; The acquisition of task information for the warehouse robot includes: Obtain the actual task information sent from the host computer to the warehouse robot; And / or, The historical task information of the warehouse robot is obtained, and the historical task information is input into the task prediction model to obtain the predicted task information.

3. The path planning method as described in claim 1, characterized in that, The acquisition of task information and status information of the warehouse robot includes: Establish a network connection with the warehouse robot, and divide the network into a first sub-network, a second sub-network, and a third sub-network; The map data in the status information is obtained through the first sub-network; The task information is obtained through the second sub-network, and the task information is stored as evidence. Sensor data from the status information is obtained through the third sub-network.

4. The path planning method as described in claim 1, characterized in that, The step of inputting the task information and status information into the path planning model to obtain the path planning result includes: The task information and the status information are input into the global planning layer, and the global planning layer performs a global path search on the task information and status information to generate the global optimal path; The globally optimal path is input into the local optimization layer, which dynamically adjusts the globally optimal path to generate the path planning result.

5. The path planning method as described in claim 4, characterized in that... The global planning layer performs a global path search on the task information and status information to generate the globally optimal path, including: Obtain the location information and energy consumption information from the status information, and construct an objective function based on the task information, the location information, and the energy consumption information; An initial honey source set is randomly generated, which includes multiple honey sources. Each honey source is a possible solution to the objective function, and each honey source corresponds to a candidate path. The initial nectar source set is iteratively optimized using the artificial bee colony algorithm until the preset convergence condition is met, and the globally optimal solution is obtained, which is taken as the globally optimal path.

6. The path planning method as described in claim 5, characterized in that, The step of iteratively optimizing the initial nectar source set using the artificial bee colony algorithm until a preset convergence condition is met includes: Each iteration round includes a hired bee phase, an observation bee phase, and a scout bee phase; During the hired bee stage, hired bees are generated based on the nectar sources in the initial nectar source set, with each hired bee corresponding to one nectar source; the hired bee performs a neighborhood search in the solution space based on the corresponding nectar source to update the nectar source position; and calculates the updated fitness value of the nectar source according to the objective function. During the bee observation phase, a preset probability selection mechanism is used to select the N nectar sources with the highest fitness values ​​from all the nectar sources, where N is a preset integer. During the scout bee phase, nectar sources whose fitness values ​​have not reached a preset threshold are randomly reset; The hired bee phase, the observation bee phase, and the scout bee phase are repeated until the preset maximum number of iterations is reached or the global optimal solution meets the convergence condition. The global optimal solution is the nectar source with the highest fitness value in each iteration.

7. The path planning method as described in claim 4, characterized in that, The local optimization layer dynamically adjusts the globally optimal path to generate the path planning result, including: The globally optimal path is divided into multiple path segments, and each path segment includes at least one node. Each node is assigned a pheromone using an ant colony algorithm. The pheromone includes concentration information, which is determined based on the globally optimal path. Each path segment is locally iteratively updated, and the local iterative update includes: in the current path segment, selecting the next node according to the concentration information and a preset heuristic factor, and completing the connection between the nodes; marking the connected nodes as the updated path segment, and updating the concentration information of all nodes in the updated path segment; Repeat the local iterative update until all path segments meet the optimization conditions; summarize all the updated path segments to form the path planning result.

8. The path planning method as described in claim 1, characterized in that, The step of controlling the warehouse robot to perform tasks based on the path planning results includes: Obtain the scene information where the warehouse robot is located; construct a digital twin scene based on the scene information and the status information, the digital twin scene including a virtual map and a virtual robot; In the digital twin scenario, control commands corresponding to the path planning results are configured to the virtual robot to drive the virtual robot to simulate operation in the virtual map; Based on the simulation results of the virtual robot, operational evaluation data is generated; If the operational evaluation data indicates that the path planning result meets the preset execution conditions, then the path planning result is sent to the actual warehouse robot for execution. If the operational evaluation data indicates that the path planning result does not meet the execution conditions, the current path planning result is discarded and regenerated until the generated path planning result meets the execution conditions.

9. A computer device, characterized in that, Including processor and memory; The processor is configured to execute a computer program stored in the memory to implement the method as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method as described in any one of claims 1 to 8.

Images