A warehouse four-way vehicle path planning method based on multi-floor discrete map and safety interval reservation

Abstract

The application discloses a warehouse four-way vehicle path planning method based on a multi-floor discrete map and a safety interval reservation, belongs to the technical field of intelligent warehouse path planning, and solves the path planning problem of four-way vehicles under the conditions of multi-floor, multiple obstacles and dynamic changes of shelf states in a stereoscopic warehouse. The method comprises the following steps: performing hierarchical gridding modeling on a warehouse space, and integrating the same-layer passing relationship, the cross-layer connecting relationship, the regional attribute and the directional passing rule into a same map model; combining a reservation table mechanism with safety interval search, so that the node occupation, the edge occupation and the opposite conflict can be simultaneously constrained in the path generation process, and the waiting mechanism and the turning cost are uniformly integrated into the search framework; and the shelf occupation state change, the temporary occupation of the parked vehicle and the task stage state are uniformly converted into the node accessible constraint and the local directional passing constraint, and are injected into the current path planning process. The application is suitable for the path planning of the four-way vehicle when the four-way vehicle executes an entry task in a multi-floor stereoscopic warehouse scene.

Description

Technical Field

[0001] This application relates to the field of intelligent warehouse path planning technology, and in particular to four-way vehicle path planning for multi-story warehouses. Background Technology

[0002] Four-way vehicles are crucial execution equipment in automated material handling systems (AS / RS), enabling them to perform tasks such as rack storage and retrieval, aisle transfer, and cross-level transport within the warehouse's rail network. As AS / RS scales expand and operational scenarios become increasingly complex, four-way vehicle path planning has become a key technical issue affecting the operational efficiency, resource utilization, and operational safety of warehousing systems. Research reveals that while existing four-way vehicle scheduling and path planning technologies have made some progress—for example, patent CN119863101A discloses a four-way shuttle system scheduling method for multi-business scenario integration, used to handle task allocation and obstacle avoidance coordination in multi-level racking, single-lift, and multi-vehicle obstacle avoidance scenarios—it has not yet formed a unified discrete map modeling and spatiotemporal collaborative path search mechanism for multi-level warehouse environments. Chinese patent CN118618778A, published on September 10, 2024, discloses a four-way vehicle path planning method for high-rise racking, used to achieve task allocation through a two-dimensional grid map and constraint rules. While existing technologies include path planning for four-way vehicles, they do not adequately consider cross-level connections, dynamic state injection, and collaborative search of reservation tables and safety zones in multi-story warehouses. Chinese patent CN121032352A, published on November 28, 2025, discloses an intelligent warehouse four-way vehicle scheduling method to alleviate operational conflicts caused by overlapping vehicle paths through path occupancy time analysis and priority adjustment. However, it still lacks sufficient consideration for cross-level path construction, dynamic constraint updates, and refined spatiotemporal conflict avoidance during multi-vehicle operation in multi-story warehouses. Overall, existing technologies do not adequately consider the comprehensive consideration of multi-story access relationships, warehouse-specific access rules, dynamic state constraints, and spatiotemporal collaborative collision avoidance for multiple vehicles in automated warehouse environments. Therefore, there is still room for further optimization of four-way vehicle path planning methods in complex automated warehouse scenarios. Summary of the Invention

[0003] The purpose of this invention is to solve the technical problems of path planning, cross-level path construction and dynamic obstacle avoidance of four-way vehicles in an automated warehouse environment under conditions of multiple floors, multiple obstacles and dynamic changes in shelf status. It provides a warehouse four-way vehicle path planning method based on multi-floor discrete maps and safety interval reservation.

[0004] The technical solution adopted by this invention to solve the above problems is: a warehouse four-way vehicle path planning method based on multi-story discrete maps and safety interval reservation, the method comprising:

[0005] Step 1: Construct a discrete map model of a multi-story warehouse;

[0006] Step 2: Receive the route planning request and determine the planning start point and planning end point;

[0007] Step 3: Determine whether the planning start point and the planning end point are on the same floor. If the task start point and the task end point are on the same floor, then execute same-floor path planning; if they are on different floors, then use cross-floor path planning.

[0008] Step 4: Construct dynamic constraints and inject them into the current path planning process, including: obtaining dynamic state information in the warehouse environment, which includes at least the shelf occupancy status, the parking status of other four-way vehicles, and the vehicle operation status corresponding to the current task stage; and dynamically correcting the node accessibility, local directional traffic relationships, and temporary obstacle set in the basic discrete map based on the dynamic state information.

[0009] Step 5: Perform spatiotemporal conflict avoidance path search based on the reservation table and safe intervals, including: for each path stage at the same level, construct a node reservation table and an edge reservation table to record the occupancy information of the planned path in the spatial-temporal dimension; then generate the corresponding safe interval set according to the occupancy time of each node in the node reservation table; subsequently, using the current position, the current available safe interval, the previous direction of movement, and the current arrival time as the search state, gradually expand the adjacent reachable nodes to obtain the stage path that meets the requirements of spatiotemporal conflict avoidance.

[0010] Step 6: Connect the same-floor path stages with the elevator cross-floor jump stages in sequence to form a complete path planning result.

[0011] Further, step 1 includes:

[0012] Each floor is divided into a grid consisting of several columns and several rows;

[0013] Any location node ,in, Indicates the floor number. Indicates column number, Indicates the row number. Indicates the area type to which the node belongs; the area type includes at least the shelving area. Track area Picking work area Elevator area and charging area ;

[0014] For each node, establish a location code, area mapping, shelf occupancy status, and directional access attributes; the directional access attributes are used to describe the passability of the four-way vehicle in the four directions of up, down, left, and right at the current node;

[0015] Represent the entire multi-story warehouse map as a graph model:

[0016]

[0017] in, For a set of nodes, Let it be the set of edges;

[0018]

[0019] in, This represents the set of access edges between nodes on the same floor. This represents the set of cross-floor connecting edges established between different floors via elevators;

[0020] For same-floor access relationships, if two nodes are located on the same floor and are adjacent in both column and row coordinates, then...

[0021]

[0022] in, For nodes The floor number of the floor where it is located. For nodes The floor number of the floor where it is located. For nodes Column number, For nodes Column number, For nodes row number, For nodes The row number;

[0023] At the same time, the current node Towards adjacent nodes If the directional passage attribute is set to allow passage, and the target node is allowed to enter under the current vehicle operating status, then at the node... With nodes Establish common edges at the same level between them;

[0024] For cross-floor connections, they are only allowed to be established between elevator area nodes;

[0025] If two nodes are located on adjacent floors, and their column number and row number are the same, and simultaneously satisfy the following conditions:

[0026]

[0027] in, For nodes Region type, For nodes The region type, then in the node With nodes Establish cross-layer connection edges between them;

[0028] Each floor uses a unified column and row coordinate system, and the elevator area nodes maintain a positional correspondence between different floors.

[0029] Furthermore, step 1 also includes:

[0030] Let the operating state of the four-way vehicle be denoted as , Including no-load operation status and cargo operation status Then the accessibility of a node is represented as

[0031]

[0032] The accessibility of a node is determined by the node area type, shelf occupancy status, and directional access attributes.

[0033] Further, step 2 includes:

[0034] The path planning request includes at least the vehicle's current location, the task start point location, the task end point location, and the current running status;

[0035] The input position code is parsed and converted into corresponding discrete nodes to obtain the current position node. Task start node and the task endpoint ;

[0036] The validity of the current location node, the task start node, and the task end node is verified. The validity verification includes at least: whether the corresponding floor is within the allowed range of the warehouse, whether the column number and row number are within the map boundary range, and whether the corresponding location has a valid node in the current warehouse discrete map. If any location does not meet the above conditions, the current path planning request is directly determined to be invalid and a planning failure result is returned.

[0037] Furthermore, in step 3, the cross-floor path includes three stages: the first stage is the same-floor path from the current location to the elevator entrance of the current floor; the second stage is the cross-floor jump between elevator nodes; and the third stage is the same-floor path from the elevator exit of the target floor to the task endpoint. Cross-floor movement is only allowed between elevator area nodes.

[0038] When there are multiple available elevator nodes, the path cost of different cross-floor path schemes is calculated separately, and the cross-floor path scheme is selected according to the path cost. The path cost includes the same-floor path cost from the current position to the elevator entrance, the elevator cross-floor cost, and the same-floor path cost from the elevator exit of the target floor to the destination.

[0039] Furthermore, step 4 includes:

[0040] For the shelf occupied status, when a shelf node is in an in-stock state, the local access attribute of the node itself or its adjacent nodes towards the shelf is restricted; when the shelf node is out of stock, the access relationship in the corresponding direction is restored.

[0041] For parked vehicle constraints, the system reads the current positions of vehicles in the other four directions and considers the positions of parked vehicles as temporary obstacles in this planning process; let the set of detected parked vehicle positions be... The current planned location of the vehicle is The starting point of the task is Therefore, the actual set of dynamic obstacles injected in this plan is:

[0042]

[0043] If the destination of the task is already occupied by a parked vehicle, the current task is determined to be unexecutable.

[0044] The node's allowed entry rules are adjusted based on the operating status of the four-way vehicle;

[0045] The dynamic constraints include: before each path planning begins, generating a corresponding dynamic constraint layer based on the real-time status and overlaying it onto the discrete map.

[0046] Furthermore, step 5 includes:

[0047] Suppose a planned path is Its corresponding arrival timeline is ,according to and Write the node occupancy time into the node reservation table, write the movement time between adjacent nodes into the edge reservation table, and impose occupancy restrictions on reverse edges at the same time.

[0048] For any node x, generate a set of safe intervals for that node based on the time it has been occupied. ,node Each safe zone represents a node that was not occupied by other vehicles within the corresponding time range;

[0049] During state expansion, if the current state is located at node The current arrival time is The adjacent nodes to be expanded are If the current direction of movement differs from the previous direction, the vehicle can wait at the current node until the earliest departure time that satisfies the constraints is found. If the current direction of movement is different from the previous direction, it means that the vehicle has turned at the current node, requiring additional turning time. Correspondingly, the arrival time of the target node can be represented as...

[0050]

[0051] in, Indicates the candidate departure time. Indicates the base travel time. This indicates the additional time for the turn; the extension is considered valid only if the target node is accessible within the safe zone corresponding to the arrival time, and neither the current edge nor the reverse edge is reserved at that time.

[0052] The path search ends when the target node arrival condition is met, and the search results of the current path stage are output. The search results include at least the node sequence of the stage, the arrival time of each node, and the node reservation information and edge reservation information corresponding to the path of the stage. If the candidate state set is empty during the search process, or there is no valid state that can be further expanded under the current reservation constraint, safe interval constraint, and edge occupancy constraint, the current stage search is terminated, and it is determined that there is no feasible path that meets the requirements of spatiotemporal conflict avoidance in the stage.

[0053] Based on the aforementioned spatiotemporal constraints, a heuristic search strategy is further introduced, prioritizing the expansion of candidate states with smaller expected total costs. Its heuristic function is:

[0054]

[0055] in, and Represents the planar coordinates of the current node. and This represents the planar coordinates of the target node on the current floor.

[0056] Secondly, the present invention provides a warehouse four-way vehicle routing system based on a multi-floor discrete map and safety interval reservation. The system has program modules corresponding to the steps of the warehouse four-way vehicle routing method based on a multi-floor discrete map and safety interval reservation as described above, and executes the steps in the warehouse four-way vehicle routing method based on a multi-floor discrete map and safety interval reservation when running.

[0057] Thirdly, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program stored in the memory, it executes the steps of a warehouse four-way vehicle path planning method based on a multi-floor discrete map and safety interval reservation as described above.

[0058] Fourthly, the present invention provides a computer-readable storage medium for storing a computer program that executes a warehouse four-way vehicle path planning method based on a multi-floor discrete map and safety interval reservation as described above.

[0059] The beneficial effects of this invention are:

[0060] This invention proposes a cross-level path planning method for four-way vehicles in warehouses based on a safety interval and spatiotemporal reservation coordination mechanism. This method addresses the shortcomings of existing technologies in avoiding spatiotemporal conflicts and ensuring low path executability under conditions of multiple vehicles operating in parallel. By combining a reservation table mechanism with a safety interval search, it can simultaneously constrain node occupancy, edge occupancy, and opposing conflicts during path generation. Furthermore, it integrates waiting mechanisms and turning costs into a unified search framework, thereby improving the success rate of path planning results in multi-vehicle operation scenarios and reducing vehicle waiting and channel conflicts. This invention also preferably provides a discrete map construction method suitable for multi-story automated warehouses. This method performs layered grid modeling of the warehouse space and integrates same-level access relationships, cross-level connection relationships, regional attributes, and directional access rules into a single map model. This addresses the shortcomings of existing technologies in representing multi-story warehouse structures, thereby improving the deployment usability of this method in complex automated warehouse scenarios. This invention preferably employs a constraint injection and real-time accessibility update mechanism oriented towards the dynamic state of warehouses. This mechanism transforms changes in shelf occupancy, temporary vehicle occupancy, and task phase states into node accessibility constraints and local directional accessibility constraints, which are then injected into the current path planning process. This ensures that the path planning results can promptly respond to changes in shelf status and parked vehicles, improving on-site execution availability. Through this technical solution, this invention enables the effective construction, dynamic adjustment, and spatiotemporal collaborative optimization of four-way vehicle paths in multi-story warehouse environments, providing a feasible path planning method for the efficient and stable operation of four-way vehicles in complex warehouse scenarios.

[0061] This invention is applicable to path planning for four-way vehicles performing tasks such as inbound, outbound, transfer, empty vehicle movement, loaded vehicle movement, and parking in multi-story automated warehouses. Attached Figure Description

[0062] To more clearly illustrate the technical solution of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0063] Figure 1 This is a schematic diagram of the multi-layer warehouse layout and cross-layer path planning in the embodiments of the invention.

[0064] Figure 2 This is a flowchart of one embodiment of the method of the present invention.

[0065] Figure 3 This is a flowchart of the path search algorithm based on the reservation table and the safety interval in this invention. Detailed Implementation

[0066] To enable those skilled in the art to better understand the present disclosure, the technical solutions of the present disclosure will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of the present disclosure, and not all embodiments. Based on the embodiments of the present disclosure, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present disclosure.

[0067] A specific implementation method for warehouse four-way vehicle routing based on multi-story discrete maps and safety zone reservations includes:

[0068] Step S1: Construct a multi-story discrete map model of the warehouse.

[0069] The automated warehouse environment is discretized into a multi-level, layered grid map, and a discrete node model suitable for four-way vehicle path planning is established. In one embodiment, the warehouse includes several floors, each floor is divided into several columns and rows of grids, thereby uniformly representing the entire warehouse space as a set of computable discrete locations. For any location node, a... To represent, where, Indicates the floor number. Indicates column number, Indicates the row number. This indicates the area type to which the node belongs. Therefore, any location within the storage space can be uniquely identified by its floor, column, row, and area type.

[0070] Regarding the division of regional attributes, each node is assigned a clear meaning related to warehousing operations. The regional types include at least shelving areas. Track area Picking work area Elevator area and charging area Among them, the shelving area Further distinguish between two states: occupied and unoccupied locations, to characterize the impact of shelf occupancy on local accessibility; track area Used to indicate the main path for the four-way vehicle to travel along the track; picking station area Used to indicate the location of work interaction; elevator area Used to indicate the location where the four-way vehicle performs cross-floor switching between different floors; charging area This is used to indicate vehicle parking and charging locations. Through the above-described area division, nodes not only have geometric location attributes but also operational function attributes.

[0071] Based on this, a location code, area mapping, shelf occupancy status, and directional access attributes are established for each node. The directional access attributes describe the passability of a four-way vehicle in the current node in four directions: up, down, left, and right, and can be denoted as...

[0072]

[0073] in, , , , These represent the access permission status of a node in the up, down, left, and right directions, respectively. This setting allows for simultaneous description of the node's location, the area it belongs to, and the directions it can move in, enabling the discrete map model to reflect the directional constraints within the storage corridor.

[0074] Furthermore, the entire multi-story warehouse map is represented as a graph model.

[0075]

[0076] in, For a set of nodes, This is a set of edges. The set of edges includes two categories: edges that connect within the same layer and edges that connect across layers.

[0077]

[0078] in, This represents the set of access edges between nodes on the same floor. This represents the set of cross-floor connecting edges established between different floors via elevators.

[0079] For same-floor access relationships, if two nodes are located on the same floor and are adjacent in both column and row coordinates, then...

[0080]

[0081] At the same time, the current node Towards adjacent nodes If the directional passage attribute allows passage, and the target node is allowed to enter under the current vehicle operating state, then at the node... With nodes Establish same-level access edges between them. This rule describes the movement relationships of four-way vehicles within a single floor along the track area, accessible cargo area, and other permitted access areas.

[0082] For cross-floor connections, only connections between elevator area nodes are allowed. If two nodes are located on adjacent floors, their column and row numbers are identical, and both satisfy the following conditions...

[0083]

[0084] Then at the node With nodes Cross-level connection edges are established between them. Therefore, four-way vehicles can only transfer between floors via elevator nodes that are aligned vertically on different floors, and cannot move directly between ordinary nodes. In this way, the vertical traffic relationships in the warehouse map are explicitly constrained within the elevator shafts, making the construction of cross-level paths more consistent with the actual warehouse structure.

[0085] In one implementation, each floor uses a unified column and row coordinate system, and elevator area nodes maintain a positional correspondence between different floors. Based on this setup, the elevator entrance on the current floor and the elevator exit on the adjacent floor can form a clear mapping, thus making the multi-floor warehouse map no longer a simple superposition of multiple two-dimensional planes, but a layered map structure that simultaneously includes planar passage relationships and vertical connection relationships. During path planning, the four-way vehicle can either perform intra-floor movement within a certain floor or switch to an adjacent floor via elevator nodes to continue moving when certain conditions are met.

[0086] To reflect the impact of shelf status and vehicle operating status on map accessibility, this invention further overlays node accessibility constraints on the basic graph model. Let the operating status of a four-way vehicle be denoted as... ,in, Indicates no-load operation status. If we represent the cargo loading and unloading status, then node accessibility can be represented as:

[0087]

[0088] Whether entry is permitted is determined by the node area type, shelf occupancy status, and directional access attributes. For nodes with goods, accessibility is related to the loading status, task stage, and local access rules; for elevator area nodes, they serve both as connections between nodes on the same floor and as for cross-floor switching. Introducing these access constraints allows the map model to reflect the actual availability of different nodes in the warehouse environment under different business conditions. When the shelf occupancy status changes, the local access attributes of that node and its adjacent nodes can be updated synchronously; when parked vehicles or temporary obstacles are detected, new inaccessible constraints can be superimposed on the discrete map model. Therefore, the map model constructed in this step is not a simple static geometric mesh, but a multi-floor warehouse-specific discrete map model that integrates floor structure, area function, directional rules, shelf status, and cross-floor mapping relationships.

[0089] By employing the above method, a discrete map of a multi-story warehouse is constructed, enabling the expression and calculation of intra-floor track connections, inter-floor elevator connections, and regional constraints within the warehouse scenario within a unified model. This provides foundational map support for subsequent planning origin-endpoint analysis, candidate inter-floor passage selection, dynamic obstacle injection, and spatiotemporal path search based on reservation tables and safety zones. This map construction method improves the deployment usability of path planning methods in actual multi-story automated warehouse operation scenarios.

[0090] Step S2: Receive the route planning request and determine the planning start point and planning end point.

[0091] The system receives a four-way vehicle path planning request, which includes at least the vehicle's current position, task start position, task end position, and current running status. The system first parses the input position code and converts it into corresponding discrete nodes, obtaining the current position node. Task start node and the task endpoint .

[0092] After location parsing is complete, the current location, task start point, and task end point are validated for validity. This validation includes at least the following: whether the corresponding floor is within the warehouse's permitted area, whether the column and row numbers are within the map boundaries, and whether the corresponding location has a valid node in the current discrete warehouse map. If any location does not meet these conditions, the current path planning request is deemed invalid, and a planning failure result is returned.

[0093] Furthermore, the node access constraints and passage rules used for subsequent route planning are determined based on the current business status of the four-way vehicle. When the task belongs to inbound, outbound, or transfer tasks, the overall task can be further divided into an empty operation phase and a loaded operation phase, so that subsequent route planning can adopt corresponding passage constraints according to different phases. In this way, external task requests can be uniformly converted into computable inputs on a discrete map, and a foundation can be provided for subsequent same-level route planning, cross-level route construction, and dynamic constraint injection.

[0094] Step S3: Determine whether the planning start point and planning end point are located on the same floor, and construct the cross-floor path.

[0095] After obtaining the task start and end points, the first step is to determine if they are on the same floor. If they are, same-floor path planning is performed directly within that floor. If they are on different floors, a segmented cross-floor path construction method is adopted. Specifically, the cross-floor path is divided into three stages: the first stage is the same-floor path from the current location to the elevator entrance on the current floor; the second stage is the cross-floor jump between elevator nodes; and the third stage is the same-floor path from the elevator exit on the target floor to the task end point. This method decomposes the originally complex multi-floor path planning problem into a combined problem of "same-floor movement – ​​cross-floor jump – same-floor movement," thereby improving the clarity and executability of cross-floor path construction.

[0096] Furthermore, cross-floor movement is only permitted between elevator zone nodes. This means that four-way vehicles can only switch between floors via vertically aligned elevator nodes on different floors, and cannot move directly between ordinary nodes. Therefore, the cross-floor path construction process is always based on the cross-floor connection relationships described by the multi-floor discrete map model, ensuring that the cross-floor paths remain consistent with the actual warehouse structure.

[0097] Preferably, when multiple available elevator nodes exist, the path costs of different cross-floor path schemes can be calculated separately, and the scheme with the better cost can be selected as the cross-floor path scheme. Let the path cost from the current position to the elevator entrance on the same floor be... The cost of elevators moving between floors is The cost of the same-floor path from the elevator exit on the target floor to the destination is... Then the total cost of the corresponding cross-layer solution can be expressed as:

[0098]

[0099] Using the above methods, the system can organize and construct cross-level paths in multi-level warehousing scenarios while meeting cross-level access constraints.

[0100] Step S4: Construct dynamic constraints and inject them into the current path planning process.

[0101] Before performing path search, dynamic status information of the warehouse environment is acquired. This dynamic status information includes at least the shelf occupancy status, the parking status of other four-way vehicles, and the vehicle operation status corresponding to the current task stage. Based on this dynamic status information, the accessibility of nodes, local directional traffic relationships, and temporary obstacle sets in the basic discrete map are dynamically corrected, so that subsequent path planning can reflect the real-time accessibility of the warehouse environment.

[0102] This step is one of the key improvements of this invention. Unlike methods that rely solely on static maps for path searching, this invention does not directly plan paths on a fixed topology map. Instead, it transforms the real-time business status of the warehouse environment into dynamic constraints in the current planning process and injects them into the base discrete map before path searching. Therefore, the base map not only represents the static spatial structure of the warehouse but also reflects the impact of changes in shelf occupancy, temporary vehicle occupancy, and task phase changes on node accessibility and local traversability, making the path planning results more consistent with actual warehouse operation scenarios.

[0103] Regarding shelf occupancy status, when a shelf node is in an in-stock state, the local accessibility of that node or its adjacent nodes towards the shelf is restricted; when the shelf node is out of stock, the accessibility in the corresponding direction is restored. In this way, changes in shelf status can be directly mapped to updates to directional access rules in the local map, ensuring that subsequent path planning results are consistent with the actual warehousing scenario. Furthermore, for the shelf node itself, its accessibility can be adjusted synchronously based on the current task stage and vehicle operating status to prevent vehicles from mistakenly treating in-stock locations as intermediate access nodes under unauthorized conditions.

[0104] For parked vehicle constraints, the system reads the current positions of vehicles in the other four directions and treats the positions of parked vehicles as temporary obstacles in this planning process. Let the set of detected parked vehicle positions be... The current planned location of the vehicle is The starting point of the task is Therefore, the actual set of dynamic obstacles injected in this plan is:

[0105]

[0106] In other words, when generating the obstacle set, the location of the currently planned vehicle and the starting point that the current task must reach are removed to avoid misclassifying necessary nodes as obstacles. If the task endpoint is already occupied by a parked vehicle, the current task can be directly determined to be unexecutable.

[0107] Furthermore, the rules for allowing entry to nodes can differ depending on the operating state or task stage of the four-way vehicle. For example, the entry conditions for some nodes may differ between the empty and loaded stages; the occupancy status of relevant shelf nodes may also change before and after picking up or placing goods. Therefore, this invention incorporates the vehicle's operating state, task stage state, and shelf state into the dynamic constraint modeling process, enabling path planning to no longer rely solely on a static map, but rather to perform a search based on a real-time passable map coupled with the current warehouse state, current task state, and current vehicle state.

[0108] Preferably, the dynamic constraints employ a method where the base map remains unchanged while constraints are overlaid during the current planning process. That is, the original warehouse topology is not directly modified; instead, a corresponding dynamic constraint layer is generated based on the real-time status before each path planning begins and overlaid onto the base discrete map. This approach maintains the consistency of the base map model while facilitating repeated use in different tasks, allowing different tasks to obtain path planning results that match the current environment under different time conditions.

[0109] By incorporating changes in shelf status, parked vehicle status, and vehicle operation status into the current path planning process, the subsequently generated paths can proactively avoid impassable areas and temporary obstacles in the current environment, and dynamically adjust node accessibility according to the task stage, thereby improving the actual executability, environmental adaptability, and warehousing scenario specificity of the path results.

[0110] Step S5: Perform a spatiotemporal conflict avoidance path search based on the reservation table and the safe interval.

[0111] For each path stage at the same level, a node reservation table and an edge reservation table are constructed to record the occupancy information of the planned path in the spatial-temporal dimension. Then, a set of corresponding safe intervals is generated based on the occupancy time of each node in the node reservation table. Subsequently, using the current position, the current available safe interval, the previous direction of movement, and the current arrival time as the search state, the adjacent reachable nodes are gradually expanded to obtain a stage path that meets the requirements of avoiding spatiotemporal conflicts.

[0112] This step is the main algorithmic innovation of this invention. In existing methods, the reservation table is usually only used to record whether certain locations are occupied at a certain time. However, this invention does not use the reservation table alone, nor does it use the safety interval alone. Instead, it unifies node reservation, edge reservation, reverse edge conflict constraints, safety interval generation, waiting mechanism, and turning cost into the same search framework. This expands the path planning process from the traditional spatial reachability judgment to a collaborative path planning process oriented towards spatial-temporal dual-dimensional constraints. In other words, subsequent vehicles no longer only determine whether they can reach a certain node during the search process, but also whether they can safely reach the node at a certain time without conflicting with other vehicles on nodes, edges, or opposite channels.

[0113] Suppose a planned path is Its corresponding arrival timeline is Based on the path and corresponding timeline, the node occupancy time can be written into the node reservation table, the movement time between adjacent nodes can be written into the edge reservation table, and occupancy restrictions can be imposed on reverse edges at the same time, thereby avoiding node conflicts, edge conflicts, and opposite conflicts at the same time in the subsequent path search process.

[0114] Furthermore, for any node x, a set of safe intervals for that node is generated based on the time it has been occupied. ,node Each safe interval represents a node that is not occupied by other vehicles within the corresponding time range. This setting transforms the discrete occupancy information in the reservation table into continuous available time periods that can be directly used in search decisions, thereby improving the efficiency of spatiotemporal path search.

[0115] During state expansion, if the current state is located at node The current arrival time is The adjacent nodes to be expanded are If the current direction of movement differs from the previous direction, the vehicle can wait at the current node until the earliest departure time that satisfies the constraints is found. If the current direction of movement is different from the previous direction, it means the vehicle has turned at the current node, incurring an additional turning time cost. Correspondingly, the arrival time of the target node can be represented as...

[0116]

[0117] in, Indicates the candidate departure time. Indicates the base travel time. This indicates the additional time required for the turn. The expansion is considered valid only if the target node is accessible within the safe zone corresponding to the arrival time, and neither the current edge nor the reverse edge is reserved at that time. The path search ends when the target node arrival condition is met, and the search results for the current path stage are output. The search results include at least the node sequence for that stage, the arrival time of each node, and the node and edge reservation information corresponding to the path in that stage. If the candidate state set is empty during the search process, or if there are no valid states that can be further expanded under the current reservation constraints, safe zone constraints, and edge occupancy constraints, the search for the current stage is terminated, and it is determined that there is no feasible path that meets the spatiotemporal conflict avoidance requirements for that stage.

[0118] To improve search efficiency, a heuristic search strategy is further introduced based on the aforementioned spatiotemporal constraints. This strategy prioritizes expanding candidate states with smaller expected total costs. The heuristic function can be expressed as follows:

[0119]

[0120] in, and Represents the planar coordinates of the current node. and This represents the planar coordinates of the target node on the current floor.

[0121] By incorporating node occupancy constraints, edge occupancy constraints, reverse edge conflict avoidance, safe interval generation, waiting mechanism, and turning cost into a unified search framework, the path planning results can better adapt to the actual operational needs of four-way parallel operation, local congestion, and dynamic obstacle conditions in multi-story warehouse scenarios. This improves the success rate of path planning in multi-vehicle parallel scheduling, reduces node and reciprocal conflicts, and facilitates deployment and implementation in warehouse systems.

[0122] Step S6: Output the path planning results.

[0123] The path planning stages on each floor are sequentially concatenated with the elevator cross-floor transfer stages to form a complete path planning result. For each stage's planning result, a validity check is first performed. This check includes at least: whether adjacent nodes are truly passable, whether the path incorrectly passes through shelves with goods, whether the path enters the obstacle position where the parked vehicle is located, and whether the path conflicts with other reserved paths at nodes or in opposite directions. If a path in a certain stage fails the check, other feasible routes are searched again; if no feasible path can be found within the allowed range, a planning failure result is returned.

[0124] After each stage of the path is validated, it is assembled according to the task execution order to form a complete path. For tasks such as inbound, outbound, and transfer, the status of the starting and ending shelves can be updated synchronously based on the picking and placing actions during path assembly to ensure that the warehouse status used in subsequent task planning remains up-to-date.

[0125] Furthermore, to facilitate scheduling, execution, and visualization, the complete path can be further divided according to the continuous direction of movement, breaking the entire path down into several path segments with the same continuous direction. If the complete path is represented as...

[0126]

[0127] in, to If the path segments are concatenated in sequence, then each path segment should output at least the starting node, ending node, direction of movement, and node sequence information.

[0128] Simultaneously, the total number of steps, total number of turns, and number of elevator uses are counted for the entire path. The final output should include at least: task number, vehicle number, current location, task start point, task end point, complete path, segmented path, total number of steps, total number of turns, and number of elevator uses. If necessary, the arrival time or reservation information of each node can also be output synchronously so that the subsequent scheduling system can write it into the reservation table and use it for the collaborative planning of the next four-way vehicle.

[0129] In summary, this implementation proposes a spatiotemporal collaborative path search method based on a reservation table and a safety interval to address the shortcomings of existing multi-vehicle path planning methods, which mainly rely on static spatial reachability judgments and lack the ability to collaboratively avoid node conflicts, edge conflicts, and reciprocal conflicts. It integrates node reservation, edge reservation, reverse edge conflict constraints, safety interval generation, waiting mechanisms, and turning costs into a unified search framework. Through this approach, path planning under multi-vehicle parallel operation conditions is extended from a simple spatial search to a spatial-temporal collaborative search, improving the executability, conflict avoidance capability, and operational stability of path planning.

[0130] This implementation proposes a discrete map construction method suitable for multi-story automated warehouses to address the shortcomings of existing path planning methods in representing multi-story warehouse structures, describing cross-floor access relationships clearly, and unifying the modeling of warehouse-specific directional rules. It integrates floor structure, intra-floor access relationships, cross-floor elevator connections, area attributes, and directional access rules into a single map model. This approach achieves a unified representation and calculation of multi-story warehouse environments, providing fundamental support for cross-floor path construction and path search, and improving the adaptability of path planning methods to complex automated warehouse scenarios.

[0131] This implementation proposes a constraint injection and real-time accessibility update mechanism for dynamic warehouse conditions. This addresses the shortcomings of existing path planning methods, which often rely on static maps and struggle to reflect changes in shelf occupancy, temporary vehicle occupancy, and task phase changes on path feasibility. The mechanism transforms changes in shelf occupancy, temporary vehicle occupancy, and task phase states into node accessibility constraints and local directional accessibility constraints, injecting these constraints into the current path planning process. This approach enables generated paths to proactively adapt to the dynamic warehouse environment, improving the practicality, environmental adaptability, and engineering application value of the path planning results.

[0132] Example:

[0133] like Figures 1 to 3 This embodiment provides a method for cross-floor path planning of four-way vehicles in warehouses based on a safety zone and spatiotemporal reservation coordination mechanism. This method is applicable to path planning for four-way vehicles performing tasks such as inbound, outbound, relocation, empty vehicle movement, loaded vehicle movement, and parking in multi-story automated warehouses. The system generates an executable path that satisfies traffic constraints, dynamic obstacle avoidance constraints, turning constraints, spatiotemporal conflict avoidance constraints, and cross-floor constraints based on the warehouse map, shelf occupancy status, vehicle current location, task start point, task end point, parked vehicle location, and elevator node information.

[0134] In this embodiment, the warehouse adopts a multi-layered three-dimensional layout, including several floors, each represented by a grid of columns and rows. Each floor includes shelving areas, rail areas, picking stations, elevator areas, and charging stations. The elevator area maintains vertical alignment between different floors to support four-way vehicles performing cross-floor switching along elevator nodes. In this way, the entire warehousing environment can be represented as a layered discrete map that simultaneously includes intra-floor access relationships and cross-floor connection relationships.

[0135] In this embodiment, the system preferably reads warehouse map data from a cached database; when the cached database is currently unavailable or no valid map data is read, the system then reads map information from a relational database. The read data includes at least the floor number, column number, row number, area type, location code, shelf occupancy status, and access attributes in the four directions (up, down, left, and right) for each location node. Based on this data, the system constructs a discrete warehouse node model and establishes connections between nodes to form the basic map required for subsequent path planning. The implementation method is consistent with the construction idea of ​​the multi-floor discrete warehouse map model described above in this invention, but this embodiment further clarifies the source of the map data, the field composition, and the loading method.

[0136] In this embodiment, any warehouse location node can be uniquely identified by its floor number, column number, row number, and area type. For adjacent nodes on the same floor, if passage is permitted from the current location towards the adjacent node, and the adjacent node is permitted to enter under the current vehicle operating state, then a same-floor passage relationship is established between them. For nodes on different floors, a cross-floor connection relationship is established only if both are elevator nodes and maintain consistency in their column and row numbers. Thus, the constructed map is not a simple superposition of multiple two-dimensional planes, but a unified hierarchical map structure that simultaneously includes planar passage relationships and vertical cross-floor relationships.

[0137] In the task illustrated in this embodiment, the initial position of the four-way vehicle is located in the track area of ​​the starting floor, the task start point is located at the operation node of the picking floor, and the task end point is located at the target node of the placing floor. Therefore, it is a typical cross-floor transportation task. After receiving the task request, the system first parses the floor, column number, and row number corresponding to the current position, task start point, and task end point according to the location code. Then, it checks whether the location is within the warehouse floor range, whether it is within the map boundary range, and whether the corresponding node exists in the current map. If any location cannot be parsed, or the parsing result does not exist in the map, the system directly returns a failure message; only when all locations pass the validity check will the system continue to execute subsequent path planning.

[0138] Preferably, the system can also divide the planning process into stages based on the task type. If the task is an inbound, outbound, or transfer task, the entire task is divided into an empty-load operation stage and a loaded-load operation stage. That is, the four-way vehicle first moves empty from its current position to the task start point, and then moves loaded from the task start point to the task end point. If the task is an empty-load movement task, the entire process is planned as empty. If the task is a loaded-load movement task, the planning is executed as loaded from the current position. If the task is a parking task, the system plans to the designated parking position as empty. Through the above division, the subsequent path planning process can better conform to the warehousing business process.

[0139] In this embodiment, the system does not directly perform path search on a fixed static map. Instead, it transforms the real-time business status of the warehouse environment into dynamic constraints in the current planning process and injects them into the basic discrete map before path search. First, the system dynamically updates the local accessibility relationships on the map based on the current shelf occupancy status. When a shelf node is currently stocked, the system restricts the local accessibility attributes of itself and adjacent nodes towards that shelf to prevent vehicles from mistakenly passing through stocked shelves as ordinary passageways. When the shelf node is currently empty, the accessibility relationships in the corresponding direction are restored. Simultaneously, to prevent vehicles from making longitudinal crossings in the shelf area that do not conform to the warehouse structure, the longitudinal accessibility attributes of shelf nodes can be kept as impassable, allowing vehicles to enter or park only in the directions permitted by the rules. Through the above processing, the basic map can be dynamically updated as the shelf status changes, making the planning results closer to the actual warehouse scenario.

[0140] In addition to the shelf status, the system also reads the operating status of other four-way vehicles. If a certain other four-way vehicle is not currently performing a task, it is considered to be in a parked state, and its current position is used as a temporary obstacle node in this path planning. Preferably, when generating temporary obstacles, the current planned vehicle's own position and the current task start position are excluded to avoid misjudging necessary nodes as obstacles; if the task endpoint is already occupied by a parked vehicle, the system can directly determine that the current task is not executable and return a corresponding prompt. In this way, the system can dynamically adjust the path planning results according to the actual distribution of parked vehicles without changing the original map structure. Therefore, the basic map can not only represent the static spatial structure of the warehouse, but also reflect the impact of changes in shelf occupancy status, temporary occupancy of parked vehicles, and changes in task stage on node accessibility and local accessibility, thus making the path planning results more consistent with the actual warehousing operation scenario.

[0141] After dynamic constraint injection is completed, the system performs cross-floor path construction. In this embodiment, the task start point and task end point are located on different floors, so a segmented cross-floor planning method is adopted. The system determines the elevator entrance node on the current floor and calculates the same-floor path from the current position to the elevator entrance; it performs layer-by-layer jumps based on the vertical correspondence between elevator nodes on adjacent floors; and it calculates the same-floor path from the elevator exit to the task end point on the target floor. Preferably, when there are multiple available elevator nodes, the path costs of different cross-floor schemes are compared, and the scheme with the better cost is selected as the cross-floor path scheme to complete the cross-floor path construction in the multi-floor warehouse scenario.

[0142] For each stage of the same-level path, the system further establishes a node reservation table and an edge reservation table to record the occupancy information of the planned path in the spatial-temporal dimensions. The node reservation table records when a node is occupied by a planned vehicle, and the edge reservation table records whether a certain moving edge is occupied at a certain moment. Furthermore, when the system records that a vehicle moves from a location node to an adjacent location node, it also simultaneously marks the corresponding reverse moving edge as unavailable at the same time, thereby avoiding simultaneous head-on conflicts between two four-way vehicles in a narrow passage. Thus, the system can not only avoid simultaneous node occupancy conflicts but also avoid head-on passage crossing conflicts.

[0143] After the node reservation table is established, the system further generates corresponding safety intervals based on the occupied times of each node. A safety interval refers to a continuous time period within which a node is accessible but not occupied by other vehicles. During the subsequent path search process, vehicles are only allowed to enter a node if their expected arrival time falls within the target node's safety interval. If the current time does not meet the entry conditions, the vehicle is allowed to wait at the current node until the earliest possible departure time that satisfies the constraints is found. By introducing safety intervals, the path planning process no longer only determines whether a path is connected, but also whether it is safe to traverse under the current time conditions.

[0144] During path search within the same layer, the system uses the current position, current safe zone, previous direction of movement, and current arrival time as the search state, progressively expanding to surrounding adjacent nodes. If the target node is inaccessible at the current moment, or if the current edge is occupied by another vehicle, the system prioritizes waiting. If the current direction of movement is inconsistent with the previous direction, the system determines that the vehicle has turned at the current position and adds corresponding turning time. The total time a vehicle takes from the current node to the next node includes the basic movement time, waiting time, and turning time. Through these settings, the planning results better reflect the movement characteristics of a real four-way vehicle.

[0145] In this embodiment, the system does not use a reservation table or a safe zone in isolation. Instead, it integrates node reservation, edge reservation, reverse edge conflict avoidance, safe zone generation, waiting mechanisms, and turning time into a unified search framework. This expands the path planning process from traditional spatial accessibility judgment to a collaborative path planning process oriented towards both spatial and temporal constraints. Consequently, subsequent vehicles must satisfy not only spatial connectivity requirements but also temporal accessibility requirements during the search process, significantly improving the executability and stability of path planning under conditions of multi-vehicle parallel operation.

[0146] Preferably, during the same-layer path search process, the system can also introduce a heuristic search strategy to prioritize expanding candidate states with lower expected total costs, thereby improving path search efficiency. In this embodiment, the heuristic search strategy is mainly used to quickly filter candidate expansion directions that are closer to the target node, thereby reducing invalid searches and improving the overall computational efficiency of same-layer path planning and cross-layer path planning.

[0147] After each stage of path search is completed, the system performs a validity check on the path at each stage. The check includes at least: whether adjacent nodes are truly passable, whether the path incorrectly passes through stocked shelves, whether the path enters the obstacle location where a parked vehicle is located, and whether the path conflicts with other reserved paths at any node or in the opposite direction. If a path at a certain stage fails the check, other feasible paths are searched again; if no feasible path can be found within the allowed range, a planning failure result is returned.

[0148] Once each stage of the path passes verification, the system sequentially concatenates the same-floor path stages with the elevator cross-floor stages according to the task execution order, forming a complete execution path. Furthermore, to facilitate scheduling by the higher-level system and vehicle control execution, the system can further divide the complete path according to the continuous direction of movement, splitting the entire path into several path segments. Each path segment includes at least the starting node, ending node, direction of movement, node sequence, path length, and whether it involves an elevator. Finally, the system output includes at least the task number, vehicle number, task type, current position, task start point, task end point, obstacle vehicle position, total number of steps, total number of turns, elevator usage count, complete path, and segmented path information; if necessary, it can also simultaneously save the arrival time and reservation information of each node for subsequent querying and multi-vehicle collaborative scheduling.

[0149] Through the above implementation methods, the system can realize unified modeling of multi-story warehouse environments, real-time injection of dynamic constraints, and spatiotemporal collaborative path search based on reservation tables and safety intervals. This enables the generation of executable paths that meet the requirements of cross-story passage, dynamic obstacle avoidance, and multi-vehicle spatiotemporal conflict avoidance in complex automated warehouse scenarios, thereby improving the efficiency, operational stability, and engineering adaptability of four-way vehicle path planning.

Claims

1. A warehouse four-way vehicle route planning method based on multi-story discrete maps and safety interval reservation, characterized in that, The method includes: Step 1: Construct a discrete map model of a multi-story warehouse; Step 2: Receive the route planning request and determine the planning start point and planning end point; Step 3: Determine whether the planning start point and the planning end point are on the same floor. If the task start point and the task end point are on the same floor, then execute same-floor path planning; if they are on different floors, then use cross-floor path planning. Step 4: Construct dynamic constraints and inject them into the current path planning process, including: obtaining dynamic state information in the warehouse environment, which includes at least the shelf occupancy status, the parking status of other four-way vehicles, and the vehicle operation status corresponding to the current task stage; and dynamically correcting the node accessibility, local directional traffic relationships, and temporary obstacle set in the basic discrete map based on the dynamic state information. Step 5: Perform spatiotemporal conflict avoidance path search based on the reservation table and safe intervals, including: for each path stage at the same level, construct a node reservation table and an edge reservation table to record the occupancy information of the planned path in the spatial-temporal dimension; then generate the corresponding safe interval set according to the occupancy time of each node in the node reservation table; subsequently, using the current position, the current available safe interval, the previous direction of movement, and the current arrival time as the search state, gradually expand the adjacent reachable nodes to obtain the stage path that meets the requirements of spatiotemporal conflict avoidance. Step 6: Connect the same-floor path stages with the elevator cross-floor jump stages in sequence to form a complete path planning result.

2. The warehouse four-way vehicle route planning method based on multi-story discrete map and safety interval reservation as described in claim 1, characterized in that: Step 1 includes: Each floor is divided into a grid consisting of several columns and several rows; Any location node ,in, Indicates the floor number. Indicates column number, Indicates the row number. Indicates the area type to which the node belongs; the area type includes at least the shelving area. Track area Picking work area Elevator area and charging area ; For each node, establish a location code, area mapping, shelf occupancy status, and directional access attributes; the directional access attributes are used to describe the passability of the four-way vehicle in the four directions of up, down, left, and right at the current node; Represent the entire multi-story warehouse map as a graph model: in, For a set of nodes, Let it be the set of edges; in, This represents the set of access edges between nodes on the same floor. This represents the set of cross-floor connecting edges established between different floors via elevators; For same-floor access relationships, if two nodes are located on the same floor and are adjacent in both column and row coordinates, then... in, For nodes The floor number of the floor where it is located. For nodes The floor number of the floor where it is located. For nodes Column number, For nodes Column number, For nodes row number, For nodes The row number; At the same time, the current node Towards adjacent nodes If the directional passage attribute is set to allow passage, and the target node is allowed to enter under the current vehicle operating status, then at the node... With nodes Establish common edges at the same level between them; For cross-floor connections, they are only allowed to be established between elevator area nodes; If two nodes are located on adjacent floors, and their column number and row number are the same, and simultaneously satisfy the following conditions: in, For nodes Region type, For nodes The region type, then in the node With nodes Establish cross-layer connection edges between them; Each floor uses a unified column and row coordinate system, and the elevator area nodes maintain a positional correspondence between different floors.

3. The warehouse four-way vehicle route planning method based on multi-story discrete map and safety interval reservation as described in claim 1, characterized in that: Step 1 also includes: Let the operating state of the four-way vehicle be denoted as , Including no-load operation status and cargo operation status Then the accessibility of a node is represented as The accessibility of a node is determined by the node area type, shelf occupancy status, and directional access attributes.

4. The warehouse four-way vehicle route planning method based on multi-story discrete map and safety interval reservation as described in claim 1, characterized in that: Step 2 includes: The path planning request includes at least the vehicle's current location, the task start point location, the task end point location, and the current running status; The input position code is parsed and converted into corresponding discrete nodes to obtain the current position node. Task start node and the task endpoint ; The validity of the current location node, the task start node, and the task end node is verified. The validity verification includes at least: whether the corresponding floor is within the allowed range of the warehouse, whether the column number and row number are within the map boundary range, and whether the corresponding location has a valid node in the current warehouse discrete map. If any location does not meet the above conditions, the current path planning request is directly determined to be invalid and a planning failure result is returned.

5. The warehouse four-way vehicle route planning method based on multi-story discrete map and safety interval reservation as described in claim 1, characterized in that: In step 3, the cross-floor path includes three stages: the first stage is the same-floor path from the current location to the elevator entrance of the current floor; the second stage is the cross-floor jump between elevator nodes; and the third stage is the same-floor path from the elevator exit of the target floor to the task endpoint. Cross-floor movement is only permitted between elevator zone nodes; When there are multiple available elevator nodes, the path cost of different cross-floor path schemes is calculated separately, and the cross-floor path scheme is selected according to the path cost. The path cost includes the same-floor path cost from the current position to the elevator entrance, the elevator cross-floor cost, and the same-floor path cost from the elevator exit of the target floor to the destination.

6. The warehouse four-way vehicle route planning method based on multi-story discrete map and safety interval reservation as described in claim 1, characterized in that: Step 4 includes: For the shelf occupied status, when a shelf node is in an in-stock state, the local access attribute of the node itself or its adjacent nodes towards the shelf is restricted; when the shelf node is out of stock, the access relationship in the corresponding direction is restored. For parked vehicle constraints, the system reads the current positions of vehicles in the other four directions and considers the positions of parked vehicles as temporary obstacles in this planning process; let the set of detected parked vehicle positions be... The current planned location of the vehicle is The starting point of the task is Therefore, the actual set of dynamic obstacles injected in this plan is: If the destination of the task is already occupied by a parked vehicle, the current task is determined to be unexecutable. The node's allowed entry rules are adjusted based on the operating status of the four-way vehicle; The dynamic constraints include: before each path planning begins, generating a corresponding dynamic constraint layer based on the real-time status and overlaying it onto the discrete map.

7. The warehouse four-way vehicle route planning method based on multi-story discrete map and safety interval reservation as described in claim 1, characterized in that: Step 5 includes: Suppose a planned path is Its corresponding arrival timeline is ,according to and Write the node occupancy time into the node reservation table, write the movement time between adjacent nodes into the edge reservation table, and impose occupancy restrictions on reverse edges at the same time. For any node x, generate a set of safe intervals for that node based on the time it has been occupied. ,node Each safe zone represents a node that was not occupied by other vehicles within the corresponding time range; During state expansion, if the current state is located at node The current arrival time is The adjacent nodes to be expanded are If the current direction of movement differs from the previous direction, the vehicle can wait at the current node until the earliest departure time that satisfies the constraints is found. If the current direction of movement is different from the previous direction, it means that the vehicle has turned at the current node, requiring additional turning time. Correspondingly, the arrival time of the target node can be represented as... in, Indicates the candidate departure time. Indicates the base travel time. This indicates the additional time for the turn; the extension is considered valid only if the target node is accessible within the safe zone corresponding to the arrival time, and neither the current edge nor the reverse edge is reserved at that time. The path search ends when the target node arrival condition is met, and the search results of the current path stage are output. The search results include at least the node sequence of the stage, the arrival time of each node, and the node reservation information and edge reservation information corresponding to the path of the stage. If the candidate state set is empty during the search process, or there is no valid state that can be further expanded under the current reservation constraint, safe interval constraint, and edge occupancy constraint, the current stage search is terminated, and it is determined that there is no feasible path that meets the requirements of spatiotemporal conflict avoidance in the stage. Based on the aforementioned spatiotemporal constraints, a heuristic search strategy is further introduced, prioritizing the expansion of candidate states with smaller expected total costs. Its heuristic function is: in, and Represents the planar coordinates of the current node. and This represents the planar coordinates of the target node on the current floor.

8. A warehouse four-way vehicle routing system based on multi-story discrete maps and safety zone reservations, characterized in that: The system has a program module corresponding to the steps of any one of claims 1-7, and executes the steps in the warehouse four-way vehicle path planning method based on multi-floor discrete map and safety interval reservation when running.

9. A computer device, characterized in that: It includes a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program stored in the memory, it performs the steps of the warehouse four-way vehicle path planning method based on a multi-floor discrete map and safety interval reservation as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: The storage medium is used to store a computer program that executes a warehouse four-way vehicle path planning method based on a multi-floor discrete map and safety interval reservation, as described in any one of claims 1-7.

Images