Unmanned forklift path planning method, system and equipment and readable storage medium

Abstract

The invention provides an unmanned forklift path planning method, system and device and a readable storage medium. The method comprises the steps that a storage reference base map and a road network topological structure are constructed; establishing mapping association between the nodes and the storage locations and storing the mapping association; real-time inventory data is maintained through a distributed cluster, and the storage location is updated synchronously; marking obstacle nodes based on the inventory information and the mapping relation; planning a path by adopting a path search algorithm and avoiding obstacles; and when the preset path section is completed during driving, the inventory updating obstacle is obtained again, the path is regenerated and compared, and an updating instruction is issued when a difference exists. According to the method and the device, the technical problem that the driving path is easy to fail due to the change of the obstacle node because the traditional unmanned forklift path planning cannot respond to the dynamic change of the warehouse inventory in real time is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of unmanned forklifts, and in particular to an unmanned forklift path planning method, system, device and readable storage medium. BACKGROUND

[0002] With the intelligent development of the logistics and warehousing industry, unmanned forklifts are widely used in warehouse cargo handling due to their high efficiency, precision and safety. In the context of explosive growth of e-commerce orders and increasing demand for warehouse space utilization, unmanned forklifts have achieved automation of processes such as loading, handling and stacking, significantly reducing labor costs and operational errors, and becoming the core driving force for efficient operation of modern warehouses. Path planning, as a core technology for autonomous operation, directly affects operational efficiency and safety, and is a key link in the intelligent upgrading of warehouse processes.

[0003] Traditional path planning methods are mostly based on static environment models, using preset paths or simple obstacle avoidance logic, which are difficult to cope with dynamic changes in inventory status and complex work scenarios in the warehouse environment. When goods frequently enter and exit the warehouse and multiple devices are working simultaneously, path congestion and collisions are likely to occur, causing delays in work and even causing damage to devices and loss of goods, significantly reducing warehouse efficiency and safety.

[0004] Therefore, it is of great significance to develop a planning method that responds to inventory changes in real time and dynamically optimizes paths. This not only improves the adaptive ability of unmanned forklifts in complex environments, ensuring work continuity and safety, but also promotes the development of warehouse logistics systems to higher efficiency and higher intelligence, meeting the needs of modern logistics for rapid response and precise operation. SUMMARY

[0005] The present application provides an unmanned forklift path planning method, system, device and readable storage medium, which solves the technical problem that traditional unmanned forklift path planning cannot respond to dynamic changes in warehouse inventory in real time, resulting in invalidation of driving paths due to changes in obstacle nodes.

[0006] According to one aspect of the present application, an unmanned forklift path planning method is provided, comprising: constructing a warehouse scene reference base map and generating a road network topology structure, the road network topology structure being constructed with nodes and edges as core elements, the nodes representing key coordinate positions in the warehouse area, and the edges representing passable path segments for the unmanned forklift; establishing a mapping and association relationship between the information of each node in the road network topology structure and the data of the warehouse location in the warehouse management system, and storing the mapping and association relationship in a data storage medium; The unmanned forklift control system and the warehouse management system maintain a real-time shared inventory data set through a distributed cluster. When an update event occurs to the warehouse location data of the warehouse management system, the updated inventory data is synchronized to the distributed cluster. The unmanned forklift control system obtains the in-stock location information from the real-time shared inventory data set through a cache client interface, determines an occupied node in the warehouse based on the mapping association, and marks the occupied node in the warehouse as an obstacle node; A path search algorithm is used to perform a path planning operation. During the node traversal process of the path search algorithm, an avoidance process is performed on the coordinate points marked as obstacle nodes to generate a target planning path. During the driving process of the unmanned forklift, after completing a preset number of path segments, the inventory information is obtained from the real-time shared inventory data set and the obstacle node mark is updated. The planned path is regenerated through the path search algorithm. If there is a difference between the regenerated planned path and the current execution path, a path update instruction is issued.

[0007] Furthermore, the construction of the warehouse scene reference base map and generation of the road network topology structure includes: using laser SLAM equipment in combination with visual acquisition equipment to obtain warehouse environment data, and using the Cartographer SLAM algorithm to generate a two-dimensional raster map as the scene reference base map; based on the scene reference base map, the road network topology structure is generated through a map editing tool, and the road network topology structure is marked with the travel direction attributes and width parameter thresholds of each path segment.

[0008] Furthermore, the establishment of a mapping association relationship between the node information in the road network topology structure and the storage location data of the warehouse management system includes: importing the node parameters of the road network topology structure and the storage location parameters of the warehouse management system, constructing a one-to-one mapping association relationship, configuring a unique identification code for each mapping association relationship, and setting a joint index structure for the storage location code and the node code in the data storage medium.

[0009] Furthermore, when an update event occurs in the warehouse location data of the warehouse management system, synchronizing the updated inventory data to the distributed cluster includes: deploying a data synchronization component in the warehouse management system, and when the warehouse location data changes, the data synchronization component synchronizes the updated inventory data to the distributed cluster in a preset data format through a transaction processing mechanism, and the distributed cluster adopts a hybrid persistence strategy.

[0010] Further, the path planning operation performed by the path search algorithm comprises: defining a node data structure based on a programming language, the node data structure comprising a coordinate parameter, a cost function value and a parent node pointer; storing a set of nodes to be explored in a priority queue, and performing node expansion in ascending order of the cost function value; and using a Manhattan distance algorithm to calculate a heuristic function of the cost function.

[0011] Further, the low-temperature warehouse narrow aisle turning optimization process further comprises: the unmanned forklift control system acquires load state information of the unmanned forklift, and when it is determined that the unmanned forklift is in an unloaded state, in a path planning stage, in-library and no-inventory location information of a ground layer of a rack on both sides of an aisle is retrieved, and a no-inventory area is selected as a turning node; the unmanned forklift is controlled to lower a fork height in a backward process, so that the fork is inserted into an empty space of the ground layer of the rack to complete a turning operation; an upper end surface of the ground layer of the rack is flush with the ground, and when there is no inventory stored in the ground layer of the rack, a turning gap is formed between the upper end surface of the ground layer of the rack and a lower end surface of a second layer of the rack.

[0012] According to a second aspect of the present application, there is provided an unmanned forklift path planning system, comprising: a map construction module, which constructs a warehouse scene reference base map and generates a road network topology structure, the road network topology structure being constructed with nodes and edges as core elements, the nodes representing key coordinate positions of a warehouse area, and the edges representing path segments available for an unmanned forklift to pass through; a data mapping module, which is connected to the map construction module through a communication interface, receives node parameters of the road network topology structure, and establishes a mapping and association relationship between the node parameters and inventory data of a warehouse management system, and stores the mapping and association relationship in a data storage medium; an inventory synchronization module, which is connected to the warehouse management system through a communication interface, maintains a real-time shared inventory data set, and when inventory data of the warehouse management system is updated, synchronizes the updated inventory data to the distributed cluster; an obstacle identification module, which is connected to the distributed cluster of the inventory synchronization module through a communication interface, acquires in-library inventory information, and is connected to the data mapping module through a fourth communication interface, determines in-library occupied nodes based on the mapping and association relationship, and marks the in-library occupied nodes as obstacle nodes; a path planning module, which is connected to the obstacle identification module through a communication interface, receives obstacle node marking information, and performs path planning operation by using a path search algorithm to generate a target planning path; The path update module is connected to the path planning module through the communication interface, receives the target planned path, and triggers the path planning module to regenerate the planned path every time a preset number of path segments are completed during the driving process of the unmanned forklift. If there is a difference between the regenerated planned path and the current execution path, a path update instruction is sent to the unmanned forklift through the communication interface.

[0013] Furthermore, it also includes a low-temperature storage diversion optimization module, including: The communication interface is used to connect with the unmanned forklift control system to obtain the load status information of the unmanned forklift; when it is determined that the unmanned forklift is in an unloaded state, the communication interface is used to retrieve the in-stock and out-of-stock position information of the first floor of the shelves on both sides of the aisle during the path planning stage, and the out-of-stock area is selected as the turning node; the communication interface is used to control the unmanned forklift to lower the height of the fork during the retreat process, so that the fork is inserted into the empty space on the first floor of the shelf to complete the turning operation; the upper end surface of the first floor of the shelf is kept flush with the ground, and when there is no goods stored on the first floor of the shelf, a turning gap is formed between the upper end surface of the first floor of the shelf and the lower end surface of the second floor of the shelf.

[0014] According to a third aspect of the present invention, there is provided an unmanned forklift path planning device, comprising: A processor and a memory, wherein the memory stores a computer program, and when the processor executes the computer program, the method steps provided in the first aspect are implemented; An unmanned forklift body, wherein the unmanned forklift path planning system provided in the second aspect is integrated into the unmanned forklift body.

[0015] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the method steps provided in the first aspect are implemented.

[0016] This invention provides a method, system, device, and readable storage medium for unmanned forklift path planning, resolving the technical issue of traditional unmanned forklift path planning's inability to respond to dynamic changes in warehouse inventory in real time, resulting in the vehicle's path being easily invalidated due to changes in obstruction nodes. By synchronizing inventory data in real time through a distributed cache cluster, establishing node-location mappings, and implementing a dynamic path update mechanism, this method ensures that path planning can promptly avoid dynamic obstacles and improve path adaptability. Furthermore, through a steering optimization process designed specifically for narrow aisles, the method effectively enhances the flexibility and safety of unmanned forklifts in narrow aisle scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The accompanying drawings, which form a part of this application, are included to provide a further understanding of the application, illustrate the preferred embodiments of the application and assist in

[0018] Figure 1 is a flow chart of a path planning method of a forklift truck according to an embodiment of the application;

[0019] Figure 2 is a block diagram of a path planning system of a forklift truck according to an embodiment of the application. DETAILED DESCRIPTION

[0020] It should be noted that the embodiments and features of the embodiments in the present application can be combined with each other without conflict. In order to make the person skilled in the art better understand the technical scheme of the present application, the technical scheme of the embodiments in the present application will be described clearly and completely below in conjunction with the drawings of the embodiments in the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by the person skilled in the art without creative labor should be within the protection scope of the present application.

[0021] It should be noted that the terms "data mapping module", "inventory synchronization module" and the like in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so as to describe the embodiments of the present application herein. In addition, the terms "include", "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, product or device including a series of steps or units does not have to be limited to the steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0022] The technical scheme of the present application will be further described in detail below in conjunction with the drawings and specific embodiments. Embodiment one

[0023] Reference Figure 1 The path planning method of the forklift truck according to the embodiments of the present application includes: S102, a warehouse scene reference map is constructed and a road network topology structure is generated, the road network topology structure is constructed with nodes and edges as core elements, the nodes represent key coordinate positions of the warehouse area, and the edges represent passable path segments of the unmanned forklift. Further, a laser SLAM device is combined with a visual acquisition device to obtain warehouse environment data, a two-dimensional grid map is generated by using a Cartographer SLAM algorithm as a scene reference map; based on the scene reference map, a map editing tool is used to generate the road network topology structure, and the road network topology structure is labeled with the passing direction attribute and the width parameter threshold of each path segment.

[0024] In more detail, the laser SLAM device and the visual acquisition device establish corresponding connections through an embedded communication interface to form a multi-source data acquisition link; the laser SLAM device completes the collection and filtering and denoising processing of the three-dimensional point cloud data of the environment through a loaded laser ranging module and a preprocessing unit, and the visual acquisition device completes the collection and distortion correction processing of the texture feature data of the environment through a loaded image sensor and a preprocessing unit; the preprocessed data collected by the two types of devices are time and space aligned through a timestamp synchronization mechanism, and a standardized environment perception data set is generated after processing by a feature extraction submodule. The timestamp synchronization mechanism refers to adding accurate time labels to the collected data of the laser SLAM device and the visual acquisition device through a unified clock source, so that the environment data collected by different devices at the same time corresponds to the same timestamp, ensuring the consistency of multi-source data in the time dimension and providing a time and space reference for subsequent data fusion and map construction. The feature extraction refers to extracting geometric features such as planes and corner points from the preprocessed laser point cloud data, and extracting visual features such as edges, textures, and key points from visual image data, converting the original perception data into feature information with clear semantics, reducing data redundancy, and enhancing environmental recognition.

[0025] In more detail, the feature extraction module and the map construction operation unit are connected through a data bus, and the standardized environment perception data set is transmitted to the map construction operation unit. The standardized environment perception data set refers to a multi-source environment data set after format unification, space-time alignment and feature structuring processing, including laser point cloud coordinate data, visual feature descriptor and data collection timestamp information, which can be directly parsed and called by the SLAM algorithm engine. The map construction operation unit loads the Cartographer SLAM algorithm engine integrated in the path planning system, and through the cooperative operation of real-time pose estimation in the front end and loop optimization in the back end, the cumulative positioning error is eliminated. The optimized environment data is converted into a two-dimensional occupancy grid model through rasterization processing, and a scene reference base map containing static obstacle and passable area boundary information is generated, and the base map data is stored in the map database. SLAM is an English full name, which means simultaneous localization and mapping. It is a technology that constructs an environment map in real time and determines its own position at the same time through sensor data in an unknown environment. The Cartographer SLAM is an open source SLAM algorithm framework developed by Google, which supports multi-sensor fusion and can generate high-precision and globally consistent environment maps.

[0026] In more detail, the map database and the map editing tool are connected through a network communication protocol, and the scene reference base map data is transmitted to the map editing tool through an authorized access interface. The operator labels key coordinate points on the scene reference base map and defines them as topological nodes through the map editing tool, including job starting point, target point, turning node and area boundary node. Adjacent nodes are connected by path edge elements through topological editing to form a road network framework representing the passing relationship. Each path edge element is configured with passing direction attribute and width constraint parameter to complete the generation and structured storage of road network topology structure, and the storage data of the road network topology structure is connected to the unmanned forklift control system through an interface. The map editing tool integrates layer management, coordinate calibration, graphic drawing, attribute editing function modules, and can realize fine processing of base map data and structured construction of road network topology.

[0027] S104, a mapping association relationship between node information in the road network topology structure and the storage location data of the warehouse management system is established, and the mapping association relationship is stored in a data storage medium. Further, the node parameters of the road network topology structure and the storage location parameters of the warehouse management system are imported, a one-to-one mapping association relationship is constructed, a unique identification code is configured for each mapping association relationship, and a joint index structure is set for the storage location code and the node code in the data storage medium.

[0028] In more detail, by establishing the mapping association relationship between the node information of the road network topology structure and the storage location data of the warehouse management system, the deep integration of physical space coordinates and business data is realized, and key data support is provided for the path planning system. By importing the node parameters of the road network topology and the storage location parameters of the warehouse management system, a one-to-one mapping association relationship is constructed and a unique identification code is configured, and a joint index structure is set for the storage location code and the node code in the data storage medium. The technical effect of the joint index structure of the storage location code and the node code mainly reflects in three aspects: first, the semantic binding of the physical path layer and the business data layer is realized, the abstract topology node coordinates are accurately corresponded with the specific warehouse storage location (such as shelf number, storage location), and the path planning target is transformed from pure coordinate movement to operation instruction execution that can directly respond to business needs; second, data association basis is provided for dynamic obstacle identification, the mapping rule of the storage location and the node is determined, the corresponding topology node is accurately positioned and marked as an obstacle node based on the real-time inventory data (such as storage location occupancy status) of the warehouse management system in the subsequent steps, and real-time response of the path planning to the dynamic state of the warehouse is realized; third, the efficiency and consistency of cross-system data interaction are guaranteed, the query matching efficiency of the storage location data and the node data is optimized through the design of the unique identification code and the joint index structure, data association confusion or redundancy is avoided, reliable data access support is provided for the unmanned forklift control system to quickly call the mapping relationship and update the obstacle state, and the cooperation and linkage of the path planning system and the warehouse management system are realized.

[0029] S106, the real-time shared inventory data set is maintained by the unmanned forklift control system and the warehouse management system through a distributed cluster, and when the storage location data of the warehouse management system has an update event, the updated inventory data is synchronized to the distributed cluster. Further, when the storage location data of the warehouse management system has an update event, the updated inventory data is synchronized to the distributed cluster, including: deploying a data synchronization component in the warehouse management system, when the storage location data changes, the data synchronization component synchronizes the updated inventory data to the distributed cluster in a preset data format through a transaction processing mechanism, and the distributed cluster adopts a hybrid persistence strategy.

[0030] In more detail, at the system deployment level, a high-availability data storage node is built through a Redis distributed cluster, which is a distributed storage solution provided by Redis to solve the limitations of single-node Redis in terms of storage capacity, concurrent access, and high availability. The Redis distributed cluster achieves sharded storage and load balancing of data by distributing data to multiple nodes, while supporting automatic failover to ensure continuous availability of services. The system uses Redis cluster to solve the following problems: the cluster disperses data to multiple nodes through sharding, supporting TB-level data storage and breaking through the upper limit of single-node storage; requests are dispersed to different nodes to avoid CPU and network bottlenecks of a single node, improving overall throughput and concurrent processing capacity; through master-slave replication and automatic failover, when the master node fails, the slave node can automatically upgrade to the master node, reducing service interruption time and ensuring high availability; the Redis cluster supports dynamic addition or removal of nodes, allowing the cluster capacity to be expanded without downtime, ensuring high availability.

[0031] In more detail, at the warehouse management system level, the data synchronization component is deployed internally in the warehouse management system. The data synchronization component captures real-time warehouse location data update events through Binlog analysis, and when the warehouse location state changes, such as warehousing, warehousing, or relocation changes, triggers a transaction processing mechanism to ensure the atomicity and consistency of data updates, and synchronizes the updated data to the Redis cluster in the pre-set key-value data format. The key value refers to the warehouse location code as the key and the inventory status as the value. The data synchronization component is a special tool or module used in distributed systems to achieve data consistency between different nodes, modules, or systems. The role of the transaction processing mechanism is to ensure the atomicity and consistency of warehouse location data updates and cache synchronization. When the warehouse management system's warehouse location data changes, the system automatically starts a transaction, binding "warehouse location data updates" and "data synchronization to cache" as an indivisible operation unit. Specifically, a data update operation is performed in the warehouse management system's database, such as modifying the inventory quantity, and ensuring that the operation is successfully committed; the data synchronization component is triggered to push the updated data to the Redis cluster, and the cache write success is confirmed through status feedback during the synchronization process; if either the database update or cache synchronization fails, the transaction mechanism will automatically trigger a rollback operation: the database retracts the executed update, and the cache discards invalid data write, ensuring consistent data state on both ends; if the database update or cache synchronization is successful, the transaction is committed and the data change takes effect.

[0032] In more detail, at the data persistence level, the Redis cluster adopts a hybrid persistence strategy combining RDB (Redis Database) snapshots and AOF (Append Only File) logs. The RDB is used to implement periodic full data persistence to ensure data integrity, and the AOF is used to record write operation instructions in real time to ensure that data changes are not lost. The RDB (Redis Database) snapshot is a mechanism for Redis to persist data by generating a full snapshot file (.rdb) of the memory data. Essentially, it compresses and stores all data in the Redis memory at a certain time in binary format to a static file on the disk. When the preset trigger condition (such as the configured time interval or the number of write operations) is met, the Redis master process creates a memory snapshot by forking a child process. The child process compresses and writes all key-value pair data in the current memory to a temporary RDB file, during which the master process continues to handle client requests. After writing is complete, the temporary file replaces the old RDB file, completing a full data persistence. Redis automatically loads the RDB file content into memory when it starts, restoring the data state. The AOF (Append Only File) log is a mechanism for Redis to persist data by recording all write operation commands (such as SET, HSET) to a log file. Essentially, it restores the data state by "replaying commands". When the Redis executes a write command, it first appends the command to the AOF buffer in memory, and then flushes the buffer content to the disk AOF file according to the configured synchronization strategy, such as once per second. As the AOF file size increases, Redis traverses the memory data by forking a child process to generate an optimized command set to replace the old log (AOF rewrite), reducing the file size. When Redis starts, it executes the commands in the AOF file line by line to restore the data to the latest state. The forking child process (Forking a Child Process) is a mechanism for creating a new process that allows a process, called the parent process, to copy itself by system call to generate a child process that is almost identical to it. The parent process and the child process execute in parallel after creation, but have independent memory space and execution context. By copying the parent process and letting the child process perform disk operations independently, Redis ensures high performance without blocking the service of the master process, and achieves reliable data persistence, balancing performance and data security. The forking child process is an important cornerstone of the high availability of the Redis distributed cluster.The mixed persistence strategy generates an RDB full snapshot through the Redis cluster, and then appends subsequent write commands in AOF format to the end of the RDB file to form a hybrid file. When starting, the RDB full data is loaded first, and then the incremental commands are replayed through the AOF log, while the fast recovery of RDB and the high security of AOF are taken into account.

[0033] In the S106 step, through the low-delay characteristics of the Redis distributed cluster and the transaction synchronization mechanism, it is ensured that the unmanned forklift control system can obtain the latest inventory state in real time, providing millisecond-level response inventory data support for subsequent obstacle node marking, path dynamic adjustment and other processes, and realizing efficient data collaboration between the warehouse management system and the unmanned forklift control system.

[0034] In S108, the unmanned forklift control system obtains the in-warehouse storage location information from the real-time shared inventory data set through the cache client interface, determines the in-warehouse occupied node based on the mapping association relationship, and marks the in-warehouse occupied node as an obstacle node.

[0035] In S110, a heuristic path search algorithm is used to perform path planning calculation. In the node traversal process of the heuristic path search algorithm, avoidance processing is performed on the coordinate points marked as obstacle nodes to generate a target planning path. Further, the use of a heuristic path search algorithm to perform path planning calculation includes: defining a node data structure based on a programming language, the node data structure including three-dimensional coordinate parameters, cost function values, and parent node pointers; using a priority queue to store a set of nodes to be explored, and performing node expansion in ascending order of the cost function values; and using a Manhattan distance algorithm to calculate the heuristic function of the cost function. The heuristic path search algorithm balances search efficiency and path optimality by introducing a heuristic function. The core of the heuristic path search algorithm is to select the next node to be explored by evaluating the total cost function f(n), which is defined as: f(n) = g(n) + h(n), where g(n) represents the actual cost from the starting point to the current node, the actual cost being the length of the path traveled by the unmanned forklift, h(n) represents the estimated cost from the current node n to the target node, and the estimated cost is the heuristic function, which must satisfy the admissibility, i.e. h(n) < actual shortest path cost, to ensure that the heuristic path search algorithm finds the optimal solution.

[0036] In more detail, the node data structure is defined based on the Python or C++ programming language, and key coordinate information in the road network topology is abstracted into a computable digital object. The node data structure includes the three-dimensional coordinate parameter, the cost function value, and the parent node pointer; the three-dimensional coordinate parameter includes X / Y / Z axis positions for locating the physical position of the node; the cost function value includes the actual cost g value and the heuristic cost h value for evaluating the pros and cons of the path; and the parent node pointer is used to record the path backtracking relationship, supporting the reverse tracing from the target node to the starting point. At the same time, an obstacle identification field is set in the node data structure, which is associated with the obstacle node state marked in the S108 step in real time, providing a data basis for the avoidance logic.

[0037] In detail, the priority queue is an abstract data structure for ordered storage and dynamic scheduling based on node cost function value, and its core function is to efficiently manage the set of nodes to be explored in the path search process, ensuring that the heuristic path search algorithm always expands the node with the lowest cost in the optimal path direction, thereby improving the efficiency and accuracy of path planning. Specifically, the starting node is inserted into the priority queue in the initialization stage, and the priority queue is sorted in ascending order of the total cost function value f(n) of the node. The total cost function formula is f(n)=g(n)+h(n), where f(n) represents the total cost function value, g(n) represents the actual cost value, which is calculated based on the physical distance between nodes combined with path attributes weighted, including roadway width and number of turns, to ensure that the path meets the physical driving constraints of the unmanned forklift, and h(n) represents the heuristic cost value, which is calculated using the Manhattan distance algorithm. The sum of the horizontal and vertical coordinate differences between the node and the target node is used to estimate the remaining path length, guiding the Manhattan distance algorithm to quickly converge towards the target direction; the node with the smallest cost is extracted as the current exploration node at each iteration, and its adjacent nodes are generated, including four-way or eight-way adjacent nodes, and nodes marked as obstacles are filtered out, which is achieved by querying the obstacle identification field of the node data structure; the cost function value of the valid adjacent nodes is calculated, and if the node has not been explored or a better path is found, its cost parameter and parent node pointer are updated, and it is inserted into the priority queue for subsequent expansion. The Manhattan distance algorithm is an algorithm for calculating the distance between two points, named after its geometric measurement method similar to the layout of city blocks, especially suitable for grid-like environments. For two points A(x1, y1) and B(x2, y2) on a two-dimensional plane, the Manhattan distance = |x2-x1|+|y2-y1|, which represents the shortest path length from point A to point B when moving only in the horizontal and vertical directions. The unmanned forklift can usually only move straight or turn along the warehouse roadway, and cannot move diagonally, so the Manhattan distance accurately reflects the actual driving distance. The Manhattan distance calculation formula only requires simple addition, subtraction and absolute value operations, and has low complexity, making it suitable for real-time algorithms. When the Manhattan distance satisfies the heuristic function value h(n) < actual shortest path cost condition, the heuristic path search algorithm can ensure that the global optimal path is found when performing path planning operations.

[0038] In more detail, the heuristic path search algorithm triggers a path backtracking process when it extracts the target node from the priority queue. The path backtracking process traces back from the target node to the starting node through the parent node pointer chain of the node data structure, forming an original path sequence. The path backtracking process eliminates redundant nodes and reduces path inflection points by calling the Douglas-Peucker algorithm. The Douglas-Peucker algorithm is a classic algorithm for simplifying curves or polylines by retaining key feature points to reduce data volume while maintaining the approximation of the original shape as much as possible. The Douglas-Peucker algorithm calculation steps include: step one, taking the starting point and endpoint of the path as initial key points; step two, traversing all intermediate points in the path, calculating the perpendicular distance of each point to the line connecting the starting point and the endpoint, and finding the maximum distance point; step three, if the maximum distance is greater than or equal to a preset threshold, the point is retained as a key point, and the path is divided into two segments: starting point → maximum distance point, maximum distance point → endpoint. Repeat steps two and three for each segment distance, and if the maximum distance is less than the preset threshold, discard all intermediate points and only retain the starting point and endpoint; step four, repeat steps one to three for each sub-path after splitting until all sub-paths cannot be further split. The path backtracking process uses a collision detection function to check whether the path contains any unaverted obstacle nodes, and compares it with the obstacle marker library in real time to ensure path safety. The collision detection function is an algorithm for determining whether two or more objects collide or have potential collision risks. In path planning, the simplified path is verified for segment collision detection to detect whether it passes through known obstacles such as shelves and walls. If a collision is detected, the original path is returned or the preset threshold is adjusted to simplify the path again to ensure the safety of the AGV obstacle avoidance. The final target planning path is output in coordinate array format, including control parameters such as the length, turning angle, and travel speed of each path segment. The target planning path is output to the motion control module of the AGV.

[0039] S112, during the travel of the AGV, after completing a preset number of path segments, the in-warehouse information is reacquired from the real-time shared inventory data set and the obstacle node markers are updated, the planning path is regenerated by the heuristic path search algorithm, and if the regenerated planning path differs from the current execution path, a path update instruction is issued.

[0040] In more detail, the unmanned forklift confirms the arrival at the target node by an encoder to complete a path segment, and a counter is incremented by 1 each time the unmanned forklift completes a travel of the path segment. The unmanned forklift control system monitors the counter value in real time to determine whether the counter value reaches a preset threshold. If the preset threshold is reached, a path re-planning process is triggered, and a heuristic path search algorithm is called to generate a planning path with the current position as a starting point and the original target point as an ending point. A similarity index of the generated planning path and the current execution path is calculated, and the similarity index includes: index one, path coincidence degree, which refers to the proportion of the number of path overlapping nodes to the total number of nodes; index two, total distance difference, which refers to the percentage of the length difference between the generated planning path and the current execution path; and index three, turning number change, which refers to the number of newly added or reduced turning nodes. If any of the similarity indexes exceeds a preset threshold, it is determined that there is a significant difference. If there is a significant difference, a path update instruction is generated, including the node coordinate sequence of the new path and the speed control parameter. If there is no significant difference, the current path execution is maintained to avoid efficiency loss caused by frequent adjustment.

[0041] Further, the unmanned forklift path planning method of the embodiment of the present application further includes a low-temperature warehouse narrow lane turning optimization process. The unmanned forklift control system obtains the load state information of the unmanned forklift, and when it is determined that the unmanned forklift is in an unloaded state, the in-stock and no-stock position information of the ground layer of the lane on both sides of the rack is retrieved in the path planning stage, and the no-stock area is selected as the turning node. The unmanned forklift is controlled to lower the fork height during the backward process, so that the fork is inserted into the empty space of the ground layer of the rack to complete the turning operation. The upper end surface of the ground layer of the rack is flush with the ground, and when there is no goods stored in the ground layer of the rack, a turning gap is formed between the upper end surface of the ground layer of the rack and the lower end surface of the ground layer of the rack. Embodiment two

[0042] Reference Figure 2 The unmanned forklift path planning system of the embodiment of the present application includes:

[0043] A map construction module constructs a warehouse scene reference base map and generates a road network topology structure. The road network topology structure is constructed with nodes and edges as core elements. The nodes represent key coordinate positions of the warehouse area, and the edges represent path segments that can be traveled by the unmanned forklift.

[0044] Specifically, the map construction module configures an environment perception unit, a map construction operation unit, and a topology construction unit. The environment perception unit collects three-dimensional point cloud data of a warehouse environment through a laser SLAM device, which includes a laser ranging module and synchronously collects texture feature data through a visual acquisition device, which includes an image sensor. A preprocessing unit of the laser SLAM device performs filtering and denoising processing on the three-dimensional point cloud data, and a preprocessing unit of the visual acquisition device performs distortion correction processing on image data. The laser SLAM device and the visual acquisition device form a data link through an embedded communication interface, add time labels to the preprocessed data through a timestamp synchronization mechanism, and realize spatio-temporal alignment of raw data.

[0045] Specifically, the environment perception unit transmits the aligned raw data to a feature extraction submodule of the map construction operation unit through a communication interface; the feature extraction submodule extracts geometric features such as planes and corner points from the three-dimensional point cloud data, extracts visual features such as edges and textures from the image data, generates a standardized environment perception data set, and the standardized environment perception data set includes coordinates, feature descriptors, timestamps, etc. The map construction operation unit loads a Cartographer SLAM algorithm engine, eliminates cumulative errors through real-time pose estimation and optimization, converts environment data into a two-dimensional occupancy grid model, generates a scene reference base map, and stores it in a map database. The map database transmits the reference base map data to a map editing tool of the topology construction unit through a network communication protocol via an authorized interface. An operator labels key coordinate points through the map editing tool, the key coordinate points include starting points, target points, turning nodes, etc., and are defined as topology nodes. Adjacent nodes are connected by path edge elements through a topology editing function to form a road network framework, and each path edge is configured with a passing direction attribute and a width parameter threshold. The topology construction unit receives the base map data through a communication interface, completes the structured storage of the road network topology structure, and establishes communication with an unmanned forklift control system through an interface.

[0046] The data mapping module is connected with the map construction module through a communication interface, receives node parameters of the road network topology structure, and establishes a mapping and association relationship between the node parameters and storage location data of a warehouse management system, and stores the mapping and association relationship in a data storage medium.

[0047] Specifically, the data mapping module includes a parameter import unit, an association construction unit, and an index configuration unit. The parameter import unit imports node parameters of the road network topology structure from the map construction module through a communication interface, the node parameters including coordinates, codes, etc. The parameter import unit imports bin parameters including bin numbers and bin information from the warehouse management system through a communication interface. The parameter import unit transmits the node parameters and the bin parameters to the association construction unit through a communication interface, constructs a one-to-one mapping relationship between the node and the bin, and configures a unique identification code for each mapping relationship. The association construction unit transmits the mapping relationship data to the index configuration unit through a communication interface. The index configuration unit sets a joint index structure for the bin code and the node code in the data storage medium, optimizes the query efficiency, and realizes the association and binding of physical space coordinates and business data.

[0048] The inventory synchronization module is connected with the warehouse management system through a communication interface, maintains a real-time shared inventory data set, and synchronizes the updated inventory data to the distributed cluster when the bin data of the warehouse management system has an update event.

[0049] Specifically, the bin data of the warehouse management system has an update event, such as warehousing, de-warehousing, and shifting. The data synchronization component deployed in the warehouse management system captures changes in real time through Binlog analysis. The data synchronization component starts a transaction processing mechanism, binds “bin data update” and “cache synchronization” as an atomic operation, executes the update in the database and confirms the submission first, and then triggers the synchronization process. The data synchronization component synchronizes the updated inventory data to the distributed cluster (Redis cluster) in a pre-set key-value pair format through a communication interface. The key value refers to the bin code as the key and the inventory status as the value. The Redis cluster uses sharding storage to achieve load balancing, and guarantees high availability through master-slave replication and automatic failover. The Redis cluster enables a hybrid persistence strategy, generates full data backup periodically through RDB snapshot, and records write operation instructions in real time through AOF log, ensuring data integrity and change losslessness.

[0050] The obstacle identification module is connected with the distributed cluster of the inventory synchronization module through a communication interface, obtains in-stock bin information, and is connected with the data mapping module through a fourth communication interface. Based on the mapping association relationship, the in-stock occupied node is determined, and the in-stock occupied node is marked as an obstacle node.

[0051] Specifically, the obstacle identification module obtains real-time in-stock location information from the Redis cluster through a communication interface / cache client interface, obtains the mapping relationship between nodes and locations from the data mapping module through a communication interface, and locates the corresponding topological node based on the in-stock information. The topological node occupied by the in-stock is marked as an obstacle node, and the obstacle identification field in the node data structure is updated synchronously.

[0052] The path planning module is connected with the obstacle identification module through a communication interface, receives obstacle node marking information, and performs path planning calculation using a path search algorithm to generate a target planning path.

[0053] Specifically, the path planning module includes a data structure definition unit and a priority queue management unit. The core of the heuristic path search algorithm is to select the next node to be explored by evaluating the total cost function f(n), which is defined as: f(n) = g(n) + h(n), where g(n) represents the actual cost from the starting point to the current node, the actual cost refers to the path length traveled by the unmanned forklift, and h(n) represents the estimated cost from the current node n to the target node, the estimated cost refers to the heuristic function, which must satisfy the adoptability, i.e. h(n) < actual shortest path cost, to ensure that the heuristic path search algorithm finds the optimal solution. In more detail, the data structure definition unit defines a node data structure based on Python / C++, which includes three-dimensional coordinate parameters (X / Y / Z axes), cost function values (g value, h value), parent node pointer, and obstacle identification field. The obstacle identification field is associated with the obstacle node state marked by the obstacle identification module in real time, providing data basis for obstacle avoidance logic. The priority queue management unit obtains the node data structure through a communication interface, initializes the priority queue and inserts the starting node, and the priority queue is sorted in ascending order according to the total cost function value f(n) = g(n) + h(n). In each iteration, the node with the smallest f(n) is extracted from the priority queue, its adjacent nodes are expanded, obstacle nodes are filtered, and the g(n) value and h(n) value of the new node are calculated. If the new node has not been explored or a better path has been found, the cost parameters and parent node pointer of the new node are updated and inserted into the priority queue. When the target node is extracted, the original path is generated by backtracking through the parent node pointer, and the target planning path is output to the motion control module of the unmanned forklift after simplification and verification.

[0054] A path updating module is connected with the path planning module through a communication interface, receives a target planning path, and triggers the path planning module to regenerate a planning path after each preset number of path segments is completed during driving of the unmanned forklift. If the regenerated planning path is different from the currently executed path, a path updating instruction is issued to the unmanned forklift through the communication interface.

[0055] Specifically, during driving of the unmanned forklift, a path segment is completed when the encoder confirms arrival at a target node. The counter is incremented by 1 each time the unmanned forklift completes a path segment. The path updating module monitors the counter value in real time, and triggers a path re-planning process when the preset threshold is reached, for example, when the counter reaches 5. The path updating module triggers the path planning module through the communication interface to reacquire in-warehouse information from the Redis cluster, update obstacle node markers, call the path search algorithm to regenerate a planning path with the current position as the starting point and the original target as the end point, calculate a similarity index of the new and old paths, including path coincidence, distance difference, and turning number change, and determine that there is a significant difference if any of the similarity indexes exceeds the preset threshold. If there is a significant difference, the path updating module generates a path updating instruction including node coordinate sequences and speed control parameters of the new path, and transmits the path updating instruction to the motion control module of the unmanned forklift through the communication interface. If there is no significant difference, the current path is maintained to avoid efficiency loss caused by frequent adjustments.

[0056] Further, the unmanned forklift path planning system further includes a low-temperature warehouse turning optimization module. The low-temperature warehouse turning optimization module is connected with the unmanned forklift control system through a communication interface, and obtains load state information of the unmanned forklift. When it is determined that the unmanned forklift is in an unloaded state, in-warehouse and no-warehouse position information of the ground layer of the racks on both sides of the aisle is retrieved through the communication interface during path planning, and a no-warehouse area is selected as a turning node. The height of the forks is lowered through the communication interface to control the unmanned forklift during backward movement, so that the forks are inserted into the empty space of the ground layer of the racks to complete the turning operation. The upper end surface of the ground layer of the racks is flush with the ground, and when there is no goods stored in the ground layer of the racks, a turning gap is formed between the upper end surface of the ground layer of the racks and the lower end surface of the second layer of the racks.

[0057] Compared with the prior art, the embodiment one of the application achieves the following technical effects: the unmanned forklift path planning method of the embodiment of the application realizes high precision and high adaptability of warehouse scene path planning through whole-process technical optimization. A high-precision scene reference base map is constructed through multi-source data fusion, laser SLAM and visual equipment are spatio-temporally aligned and feature extraction is performed, the Cartographer SLAM algorithm is combined to eliminate cumulative errors, a generated grid map accurately depicts the environment boundary, and a road network topology structure constructed by the topology editing tool realizes structured expression of the physical space. Mapping association and joint indexing design of nodes and storage locations realize deep fusion of physical paths and business data, provide accurate association basis for dynamic obstacle identification, and improve cross-system data interaction efficiency. The Redis cluster and transaction mechanism guarantee real-time synchronization of inventory data, the hybrid persistence strategy ensures data security, and supports accurate marking of obstacle nodes. The heuristic path search algorithm combines Manhattan distance and priority queue scheduling to quickly generate a globally optimal path, which is directly adapted to the unmanned forklift motion control system after simplification and collision verification. The dynamic re-planning mechanism realizes on-demand adjustment through path segment monitoring and similarity judgment, ensuring the optimality of the path in a dynamic scene. The low-temperature narrow lane turning optimization process utilizes the space under the shelves to solve the problem of special scene operation. The overall method significantly improves the map precision, planning efficiency and dynamic response capability, guarantees safe and efficient operation of the unmanned forklift, and adapts to the needs of various warehouse scenes.

[0058] Compared with the prior art, the second embodiment of the present application achieves the following technical effects: the unmanned forklift path planning system of the embodiment of the present application realizes intelligent upgrading of the whole process of warehouse scene path planning through modular collaborative design. The map construction module fuses laser SLAM and visual multi-source data, performs spatio-temporal alignment and feature extraction, generates a high-precision reference base map in combination with the Cartographer SLAM algorithm, constructs a structured road network containing traffic attributes through a topological editing tool, and provides a high-fidelity environmental basis for planning. The data mapping module establishes accurate association between nodes and storage locations, realizes real-time synchronization of inventory data in cooperation with the Redis cluster of the inventory synchronization module, and the transaction mechanism and the hybrid persistence strategy guarantee data consistency and high availability, supporting the obstacle identification module to accurately mark occupied nodes in the warehouse. The path planning module is based on node data structure and priority queue scheduling, quickly generates a globally optimal path through the heuristic algorithm, and directly adapts to the unmanned forklift motion control system after simplification and verification. The path updating module dynamically monitors the driving state, triggers re-planning and realizes on-demand adjustment through similarity judgment, ensuring the optimality of the path in a dynamic scene. For low-temperature narrow lane scenes, the steering optimization module utilizes the empty space at the bottom of the shelf to realize steering in narrow spaces, improving the adaptability of special scenes. The modules of the system are connected through standardized interfaces, and the whole-process closed-loop design significantly improves the path planning efficiency, dynamic response speed and operation safety, providing reliable technical support for various warehouses, especially cold-chain warehouse unmanned operation.

[0059] It should be noted that, for the foregoing embodiments, in order to simply describe, they are all expressed as a series of action combinations. However, those skilled in the art should know that the present application is not limited to the order of the actions described, because according to the present application, certain steps can be performed in other orders or simultaneously.

[0060] Those skilled in the art can clearly understand the method according to the above-mentioned embodiments can be realized by means of software and necessary general hardware platform, and of course, it can also be realized by hardware equipment. Based on such understanding, the technical solutions of the present application can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes a plurality of instructions for making a terminal device (computer, AGV body) execute the method steps described in the embodiments of the present application.

[0061] The above only describes the preferred embodiments of the present application, and it should be noted that, for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, which should also be considered as the protection scope of the present application.

Claims

1. A path planning method for an unmanned forklift, characterized in that: include: Construct a base map of the warehouse scene and generate a road network topology. The road network topology is constructed with nodes and edges as core elements. The nodes represent the key coordinate positions of the warehouse area, and the edges represent the path segments that can be passed by unmanned forklifts. Establishing a mapping association relationship between the information of each node in the road network topology and the storage location data of the warehouse management system, and storing the mapping association relationship in a data storage medium; The unmanned forklift control system and the warehouse management system maintain a real-time shared inventory data set through a distributed cluster. When an update event occurs to the warehouse location data of the warehouse management system, the updated inventory data is synchronized to the distributed cluster. The unmanned forklift control system obtains the in-stock location information from the real-time shared inventory data set through a cache client interface, determines an occupied node in the warehouse based on the mapping association, and marks the occupied node in the warehouse as an obstacle node; A path search algorithm is used to perform a path planning operation. During the node traversal process of the path search algorithm, an avoidance process is performed on the coordinate points marked as obstacle nodes to generate a target planning path. During the driving process of the unmanned forklift, after completing a preset number of path segments, the inventory information is obtained from the real-time shared inventory data set and the obstacle node mark is updated. The planned path is regenerated through the path search algorithm. If there is a difference between the regenerated planned path and the current execution path, a path update instruction is issued.

2. The method according to claim 1, characterized in that The method of constructing a warehouse scene reference base map and generating a road network topology structure includes: using a laser SLAM device in combination with a visual acquisition device to obtain warehouse environment data, and using the Cartographer SLAM algorithm to generate a two-dimensional raster map as a scene reference base map; based on the scene reference base map, generating the road network topology structure through a map editing tool, wherein the road network topology structure is annotated with the travel direction attributes and width parameter thresholds of each path segment.

3. The method according to claim 1, characterized in that The method of establishing a mapping association relationship between the node information in the road network topology structure and the storage location data of the warehouse management system includes: importing the node parameters of the road network topology structure and the storage location parameters of the warehouse management system, building a one-to-one mapping association relationship, configuring a unique identification code for each mapping association relationship, and setting a joint index structure for the storage location code and the node code on the data storage medium.

4. The method according to claim 1, characterized in that When an update event occurs in the warehouse location data of the warehouse management system, synchronizing the updated inventory data to the distributed cluster includes: deploying a data synchronization component in the warehouse management system, and when the warehouse location data changes, the data synchronization component synchronizes the updated inventory data to the distributed cluster in a preset data format through a transaction processing mechanism, and the distributed cluster adopts a hybrid persistence strategy.

5. The method according to claim 1, characterized in that The path planning operation performed using a path search algorithm includes: defining a node data structure based on a programming language, wherein the node data structure includes coordinate parameters, a cost function value, and a parent node pointer; using a priority queue to store a set of nodes to be explored, and performing node expansion in ascending order of the cost function value; and calculating the heuristic function of the cost function using a Manhattan distance algorithm.

6. The method according to claim 1, characterized in that It also includes a low-temperature storage narrow aisle turning optimization process: the unmanned forklift control system obtains the load status information of the unmanned forklift, and when it is determined that the unmanned forklift is in a no-load state, the in-stock and out-of-stock position information of the first floor of the shelves on both sides of the aisle are retrieved in the path planning stage, and the out-of-stock area is selected as the turning node; the unmanned forklift is controlled to lower the fork height during the retreat process, so that the fork is inserted into the empty space on the first floor of the shelf to complete the turning operation; the upper end surface of the first floor of the shelf is kept flush with the ground, and when there is no goods stored on the first floor of the shelf, a turning gap is formed between the upper end surface of the first floor of the shelf and the lower end surface of the second floor of the shelf.

7. An unmanned forklift path planning system, characterized in that: include: A map construction module constructs a base map of the warehouse scene and generates a road network topology. The road network topology is constructed with nodes and edges as core elements. The nodes represent key coordinate locations in the warehouse area, and the edges represent traversable paths for unmanned forklifts. a data mapping module connected to the map construction module via a communication interface, receiving node parameters of the road network topology structure, establishing a mapping association relationship between the node parameters and the storage location data of the warehouse management system, and storing the mapping association relationship in a data storage medium; An inventory synchronization module is connected to the warehouse management system through a communication interface to maintain a real-time shared inventory data set. When an update event occurs to the warehouse location data of the warehouse management system, the updated inventory data is synchronized to the distributed cluster; an obstacle identification module, connected to the distributed cluster of the inventory synchronization module via a communication interface to obtain inventory location information, and connected to the data mapping module via a fourth communication interface to determine an occupied node in the inventory based on the mapping association relationship, and mark the occupied node in the inventory as an obstacle node; A path planning module is connected to the obstacle identification module through a communication interface, receives obstacle node marking information, and uses a path search algorithm to perform path planning operations to generate a target planned path; The path update module is connected to the path planning module through the communication interface, receives the target planned path, and triggers the path planning module to regenerate the planned path every time a preset number of path segments are completed during the driving process of the unmanned forklift. If there is a difference between the regenerated planned path and the current execution path, a path update instruction is sent to the unmanned forklift through the communication interface.

8. The system according to claim 7, characterized in that Also included is a low-temperature storage diversion optimization module, including: The communication interface is used to connect with the unmanned forklift control system to obtain the load status information of the unmanned forklift; when it is determined that the unmanned forklift is in an unloaded state, the communication interface is used to retrieve the in-stock and out-of-stock position information of the first floor of the shelves on both sides of the aisle during the path planning stage, and the out-of-stock area is selected as the turning node; the communication interface is used to control the unmanned forklift to lower the height of the fork during the retreat process, so that the fork is inserted into the empty space on the first floor of the shelf to complete the turning operation; the upper end surface of the first floor of the shelf is kept flush with the ground, and when there is no goods stored on the first floor of the shelf, a turning gap is formed between the upper end surface of the first floor of the shelf and the lower end surface of the second floor of the shelf.

9. An unmanned forklift path planning device, characterized in that: include A processor and a memory, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented; AGV body, wherein the system described in any one of claims 7-8 is integrated into the AGV body.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images