A conflict avoidance method for a four-way shuttle vehicle

Abstract

The application discloses a kind of four-way shuttle vehicle conflict to move away from method, comprising: constructing global space-time diagram, record each node in future time mark Equipment occupancy information;For the first shuttle vehicle Path planning and mapping to global space-time diagram are carried out Conflict detection;When detecting that task end point is occupied by second shuttle vehicle without task when expected to arrive moment, target moving away point is selected from pre-set candidate moving away point set, and the screening process includes static topology screening, runtime filtering, Manhattan distance sorting, and based on the state of the load Directional connected graph is constructed and strong connected component is found;Generate moving away from task to make second shuttle vehicle drive to target moving away point release task end point and update space-time diagram.The application detects end point occupancy conflict actively, selects optimal moving away point based on the adjacent relationship between nodes, load state and historical data, realizes the immediate release of end point resource and global efficiency optimization.

Description

Technical Field

[0001] This invention belongs to the field of warehousing and logistics technology, and in particular relates to a method for resolving conflicts involving four-way shuttle vehicles. Background Technology

[0002] In a four-way shuttle warehouse system, when multiple shuttles share the same outlet or task endpoint, the equipment that arrives and completes its task first will continue to occupy the endpoint position if it does not receive a new task in a timely manner, forcing subsequent equipment to wait. Existing technologies rely on indirectly removing the equipment by assigning it a new task, but the timely allocation of new tasks cannot be guaranteed, leading to uncertain waiting and potentially causing cascading blockages or even partial deadlocks. Furthermore, existing solutions only remove the occupying equipment when handling occupancy conflicts, without considering the impact of the relocation action on the efficiency of subsequent tasks on the occupying equipment. Summary of the Invention

[0003] Purpose of the invention: In order to overcome the shortcomings of the existing technology, the present invention provides a four-way shuttle conflict relocation method, which actively detects the destination occupancy conflict, and selects the optimal relocation point based on the adjacency relationship between nodes, cargo status and historical data, so as to realize the instant release of destination resources and global efficiency optimization.

[0004] Technical solution: To achieve the above objectives, the present invention provides a four-way shuttle conflict resolution method, comprising the following steps:

[0005] Step S1: Construct a global spatiotemporal graph, which records the equipment occupancy information of each node in the warehousing system at each future time step;

[0006] Step S2: In response to assigning a task to the first shuttle, plan a first path from the current position of the first shuttle to the task endpoint;

[0007] Step S3: Map the first path to the global spatiotemporal graph and perform spatiotemporal conflict detection;

[0008] Step S4: When it is detected that the destination of the task is occupied by the second shuttle at the specific time when the first shuttle is expected to arrive, determine the task status of the second shuttle;

[0009] Step S5: If the second shuttle is in a no-task state, then take the task endpoint as the reference point and perform the following screening steps in sequence to obtain the target departure point;

[0010] Step S5.1: Traverse all nodes in the map and add nodes that meet the first filtering condition as candidate transfer points to the candidate transfer point set; the first filtering condition includes: the node is not disabled; the node is a storage node; the node has one and only two adjacent nodes, one of which is a storage node and the other is a lane node; and both adjacent nodes are not disabled and are not hoist nodes.

[0011] Step S5.2: Perform runtime filtering on the candidate departure point set to obtain a temporary set; the runtime filtering is used to exclude candidate departure points that are not available at the current time.

[0012] Step S5.3: Hotspot Conflict Prediction: If there is a node in the historical time period corresponding to the current time period that is occupied at a frequency f... occupy Based on the candidate transfer point set in step S5.1, the set hotspot frequency threshold is applied to filter all candidate points to obtain the available candidate transfer point set for the current time period.

[0013] Step S5.4: Calculate the Manhattan distance from the task endpoint to each departure point in the temporary set, and sort them in ascending order based on the Manhattan distance;

[0014] Step S5.5: Displacement Compensation Assessment: Based on the calculation results, distance filtering is performed using the expected increase in empty runs ΔC function, based on the function ΔC = D(p new ,p next ) - D (p orig ,p next The results show all possible values ​​(ΔC) for the existing candidate nodes, and only candidate nodes whose ΔC falls within the specified range are retained.

[0015] Step S5.6: Based on the cargo loading status of the second shuttle, determine the corresponding directed connected graph and find all strongly connected components in it;

[0016] Step S5.7: Sequentially determine whether each departure point belongs to the same strongly connected component as the task endpoint; take the first departure point that meets the condition as the target departure point;

[0017] Step S5.8: If there are no candidate nodes at this time, the device location nodes in the Idle state are extracted and filtered based on steps S5.4 to S5.7; if there are nodes that meet the requirements, the node where the new device is located is used as the starting point, and filtered through steps S5.4 to S5.7 to find the available relocation point for the device.

[0018] Step S5.8.1: If there are multiple filtering results, the sum of the Manhattan distances of the two devices to be moved from their respective target moving points shall be used as the first sorting condition, and the absolute value of the difference between the Manhattan distances of the two devices shall be used as the second sorting condition to sort the resulting moving point schemes based on the two devices.

[0019] Step S5.8.2: For the set of two separation points obtained in step 5.8.1, repeat steps S5.6~5.7. The first one that satisfies both conditions is the double separation point scheme.

[0020] Step S6: Generate a transfer task and send it to the second shuttle so that it can drive to the target transfer point and release the task endpoint; after the second shuttle arrives at the target transfer point, update the device occupancy information of the task endpoint at the corresponding time in the global spatiotemporal graph.

[0021] Furthermore, the relocation compensation assessment in step S5.5 includes:

[0022] Obtain the first position coordinates of the second shuttle when it arrives at the target departure point after completing the departure;

[0023] Obtain the original position coordinates of the second shuttle car without it being moved;

[0024] Obtain the starting position coordinates of the next task predicted to be assigned to the second shuttle from the system cache;

[0025] Calculate the expected increase in empty runs ΔC due to the second shuttle being relocated:

[0026] ΔC =D(p new ,p next ) - D (p orig ,p next );

[0027] Where, p new Let p be the first position coordinate. orig p represents the original position coordinates. next Let D be the coordinates of the starting point of the next task, and let D be the Manhattan distance function.

[0028] The filtering condition for the departure point is to use the expected increase in empty runs ΔC as a condition that is less than or equal to the preset compensation threshold.

[0029] Furthermore, the hotspot conflict prediction in step S5.3 includes:

[0030] Obtain the frequency f of the task endpoint being occupied within a preset historical time period. occupyThe frequency of occupation is the ratio of the average duration of continuous occupation of the task endpoint by the shuttle within a unit of time to the unit of time.

[0031] If the occupied frequency f occupy If the frequency exceeds the hotspot frequency threshold, the task endpoint is determined to be a hotspot conflict node.

[0032] The global candidate transfer points are filtered based on hotspots, so that all candidate transfer points are far away from hotspots.

[0033] Furthermore, for step S5.8, obtain the set S of all shuttles in the warehousing system that are in an idle state without any tasks. idle ;

[0034] If the second shuttle belongs to the shuttle set S idle If so, the second shuttle car is given priority as the departure device, and with the goal of minimizing the global task completion time, a target departure point is assigned to the second shuttle car from the set of candidate departure points;

[0035] If the second shuttle cannot complete the relocation request independently, then from the shuttle set S idle The third shuttle car is selected as an auxiliary transfer device and the transfer is performed in a chain transfer manner. The chain transfer means that the third shuttle car is first transferred to the first transfer point, and then the second shuttle car is transferred to the second transfer point. The chain transfer is limited to one layer and no recursive attempts are made.

[0036] Further, mapping the first path to the global spatiotemporal graph specifically includes:

[0037] Obtain the path length of each node in the first path, and calculate the estimated arrival time of each node in the first path based on the preset speed of the first shuttle.

[0038] The device identifier of the first shuttle is recorded in the device location information set of the corresponding node at the corresponding time in the global spatiotemporal graph.

[0039] Furthermore, the conditions for runtime filtering include:

[0040] The candidate departure point is not the endpoint of the task;

[0041] The candidate departure points were not disabled;

[0042] No goods were placed at the candidate departure point;

[0043] The candidate departure point was not locked or occupied by any device.

[0044] Furthermore, when sorting based on the Manhattan distance, the method also includes:

[0045] Candidate transfer points whose Manhattan distance is less than a preset minimum transfer distance threshold are filtered out.

[0046] Furthermore, determining the corresponding directed connected graph based on the cargo loading status of the second shuttle vehicle specifically involves:

[0047] Based on the cargo status of the second shuttle, a subgraph is dynamically constructed that contains only the nodes and edges that are allowed to pass under the current cargo status, and this subgraph is used as the corresponding directed connected graph.

[0048] Furthermore, it also includes:

[0049] When the second shuttle arrives at the target departure point, it releases its occupation of the task endpoint and removes the device identifier of the second shuttle from the set of device location information of the task endpoint at the corresponding time.

[0050] Furthermore, it also includes:

[0051] If the second shuttle is in a mission state or fails to select the target departure point, the first shuttle enters the waiting queue and periodically re-performs the spatiotemporal conflict detection until the mission endpoint is released.

[0052] Beneficial Effects: This invention improves the conflict resolution method for four-way shuttles from passive waiting to active relocation, eliminating uncertain delays caused by waiting for new task assignments. Candidate relocation point screening limits locations to end-storage positions, ensuring that equipment does not occupy lanes or create new obstacles after parking; runtime filtering ensures that relocation points are truly available at the current moment; strong connectivity component verification combined with cargo status dynamically constructs a subgraph to ensure the reachability of the relocation path; relocation compensation assessment quantifies the impact of relocation on subsequent tasks, avoiding local relocation from harming overall efficiency; hotspot conflict prediction employs an enhanced strategy of expanding the clearing radius for frequently congested endpoints, reducing the probability of repeated conflicts. This significantly reduces the risk of deadlock and congestion probability in multi-shuttle scheduling scenarios, improving the overall operational efficiency of the warehousing system. Attached Figure Description

[0053] Figure 1 This is a schematic diagram of a structure where multiple devices have the same shipping point.

[0054] Figure 2 This is a structural diagram of the candidate departure point. Detailed Implementation

[0055] The invention will now be further described with reference to the accompanying drawings.

[0056] A method for resolving conflicts involving a four-way shuttle includes the following steps:

[0057] Step S1: Construct a global spatiotemporal graph, which records the equipment occupancy information of each node in the warehousing system at each future time step.

[0058] Step S2: In response to assigning a task to the first shuttle, plan a first path from the current position of the first shuttle to the task endpoint.

[0059] Step S3: Map the first path to the global spatiotemporal graph and perform spatiotemporal conflict detection.

[0060] More specifically, in step S3, mapping the first path to the global spatiotemporal graph includes: obtaining the path length of each node in the first path, and calculating the estimated arrival time of each node in the first path based on the preset speed of the first shuttle; recording the device identifier of the first shuttle in the device location information set of the corresponding node at the corresponding time in the global spatiotemporal graph. This transforms the static path planning result into a location occupancy prediction with a time series, enabling subsequent conflict detection to be accurate to the same node and the same time, rather than relying solely on the same node for a vague judgment, thus avoiding false alarms due to time zone differences.

[0061] Step S4: When it is detected that the destination of the task is occupied by the second shuttle at the specific time when the first shuttle is expected to arrive, the task status of the second shuttle is determined.

[0062] The task endpoint is the final delivery location of the task. When the task endpoint is occupied at a specific time, the first shuttle cannot complete the delivery of the task. This differs from the occupation of intermediate nodes in the path: intermediate nodes can be rerouted; the endpoint cannot be rerouted and must wait for the endpoint resources to be released or guided to be released. Therefore, endpoint occupation is used as a specific condition to trigger a relocation. Determining the task status of the second shuttle is to distinguish whether it has a task being executed or is in an idle state. If a task is being executed, its occupation of the endpoint is normal operation and will not be interrupted; if there is no idle task, its occupied endpoint should be relocated to release resources.

[0063] Step S5: If the second shuttle is in a no-task state, then take the task endpoint as the reference point and perform the following screening steps in sequence to obtain the target departure point.

[0064] The goal of the selection process is to find the most suitable location for the second shuttle to temporarily stop from all nodes on the map. This location must simultaneously meet several constraints: spatially dockable, currently dockable, path-reachable, and without interfering with the operation of other equipment after being moved. To this end, this invention establishes a progressively advancing five-step selection process, with each step narrowing the candidate pool until the optimal solution is reached, as follows:

[0065] Step S5.1: Traverse all nodes in the map and add nodes that meet the first filtering condition as candidate departure points to the candidate departure point set. The first filtering condition is a topological filtering based on the map's static attributes. The first filtering condition includes:

[0066] (1.1) The node is not disabled, excluding nodes in the map that have been marked as unavailable by the system due to failure, maintenance or other reasons.

[0067] (1.2) The node is a storage location node, ensuring that the transfer point is located within the rack space rather than on the aisle passageway. Storage location nodes are positions used to store goods or for equipment to dock, and their dimensions are matched to the shuttle vehicle; while aisle nodes are paths for shuttle vehicle passage. If equipment is docked on an aisle node, it will block the passageway, preventing other equipment from passing. Therefore, setting the transfer point as a storage location node ensures that the transfer action does not obstruct the aisle.

[0068] (1.3) The core of the selection criteria is that the node has exactly two adjacent nodes. An adjacent node is a node that can be directly reached from the current node along the passable direction on the map. As shown in the figure, the reason why this invention requires exactly two adjacent nodes, rather than one, three, or four, is that exactly two neighbors mean that the node is in the middle or end of a "chain". Furthermore, one of the adjacent nodes is a storage location node and the other is an aisle node, ensuring that the node is located at the end of the shelf and adjacent to the aisle: one side of the storage location node is another storage location inside the shelf, and the other side of the aisle node is a passable aisle. After the equipment enters the storage location at the end of the aisle and stops, it does not occupy the aisle space and has a single aisle exit for exiting. If a node has more than two adjacent nodes, it may be located at an aisle intersection or in the middle of the shelf, which will affect the passage in multiple directions or make it difficult to exit after stopping. If there are only zero or one adjacent nodes, the node is an isolated node or the end of a dead end, and the equipment cannot exit after entering. If all adjacent nodes are storage location nodes, then this node is located deep inside the shelving. After entering, the equipment needs to pass through other storage locations to reach the aisle, but these other storage locations may be occupied by goods or other equipment, making passage practically impossible. If all adjacent nodes are aisle nodes, then this node itself is located in the aisle, and stopping at it will block the aisle.

[0069] (1.4) Furthermore, neither of the two adjacent nodes is disabled and neither is a hoist node, ensuring the smooth flow of the entry and exit paths and avoiding conflicts with vertical transportation equipment such as hoists.

[0070] Therefore, the candidate departure points selected through the above step S5.1 meet the requirements of not obstructing traffic and facilitating entry and exit on the map.

[0071] Step S5.2: Perform runtime filtering on the candidate departure point set to obtain a temporary set; the runtime filtering is used to exclude candidate departure points that are not available at the current time.

[0072] The first filtering criterion is based on static map attributes, while runtime filtering is based on the dynamic state at the time of task execution. The relationship between the two is as follows: the first filtering criterion specifies which nodes are structurally suitable as detour points, while runtime filtering specifies whether these nodes are actually available at the current moment. This two-level filtering separation ensures accurate filtering and efficient computation.

[0073] Furthermore, executing the first screening condition corresponds to the first screening, which is a global screening during the startup phase. Its purpose is to identify candidate nodes that will not affect passage. Since this is a global screening, subsequent screenings are based on this, saving time and improving operational efficiency (it is only executed once at the beginning; subsequent runtime filtering does not need to start from scratch each time). Executing runtime filtering corresponds to the second screening. Runtime filtering is based on the dynamic state during runtime, performing a second screening on the results of the first screening. This second screening aims to filter based on specific time points, the availability of candidate nodes, and the accuracy of their status, among other dynamic factors.

[0074] More specifically, the conditions for runtime filtering include:

[0075] (2.1) The candidate departure point is not the endpoint of the task.

[0076] (2.2) The candidate departure points are not disabled.

[0077] (2.3) No goods are placed at the candidate relocation point. Specifically, real-time information is obtained by querying the storage location status table of the warehouse management system to exclude storage locations that are already occupied by goods.

[0078] (2.4) The candidate departure point is not locked or occupied by any device. This was obtained by querying the device occupancy information of the node at the expected arrival time in the global spatiotemporal graph, excluding nodes that, although not occupied, have been allocated to other devices.

[0079] By filtering through the above conditions (2.1) to (2.4), it is ensured that the final selected transfer point is indeed available at the current moment.

[0080] Step S5.3: Hotspot Conflict Prediction: If there is a node in the historical time period corresponding to the current time period that is occupied at a frequency f... occupy Then, based on the candidate transfer point set in step S5.1, the set hotspot frequency threshold is applied to filter all candidate points to obtain the available candidate transfer point set for the current time period.

[0081] Step S5.4: Calculate the Manhattan distance from the task endpoint to each departure point in the temporary set, and sort them in ascending order based on the Manhattan distance.

[0082] Manhattan distance is calculated as the sum of the absolute values ​​of the distances between two points along the X-axis and the Y-axis. This distance metric aligns with the actual travel distance of the shuttle on the rectangular grid map, as shuttles can only travel horizontally or vertically. Sort points by Manhattan distance from smallest to largest, prioritizing those closer to the destination in subsequent screening processes. This minimizes the shuttle's departure distance and shortens the destination release time.

[0083] Preferably, when sorting based on the Manhattan distance, the method further includes filtering out candidate departure points whose Manhattan distance is less than a preset minimum departure distance threshold. This threshold is set to avoid moving the second shuttle only a very short distance—for example, to an adjacent storage location—leaving it still near the destination, which could continue to hinder subsequent tasks or create new local conflicts with the arriving first shuttle. By setting a minimum departure distance threshold, the effectiveness and thoroughness of the departure process are ensured.

[0084] Step S5.5: Displacement Compensation Assessment: Based on the calculation results, distance filtering is performed using the expected increase in empty runs ΔC function, based on the function ΔC = D(p new ,p next ) - D (p orig ,p next The algorithm obtains all ΔC values ​​of the existing candidate nodes and retains only the candidate nodes whose ΔC values ​​are within the specified range.

[0085] Step S5.6: Based on the cargo loading status of the second shuttle, determine the corresponding directed connected graph and find all strongly connected components in it.

[0086] When a four-way shuttle travels within a warehousing system, its passage along a particular path depends on whether it is currently carrying cargo. When the shuttle is empty, its smaller size allows it to pass through all types of paths, including narrow dedicated aisles. However, when the shuttle is loaded, the cargo size may exceed the width limits of some paths, or some paths may be prohibited for cargo transport due to safety regulations. Therefore, the nodes and edges in the globally directed connected graph do not represent passage in all cargo states; different cargo states need to be modeled separately.

[0087] Specifically, determining the corresponding directed connected graph based on the cargo status of the second shuttle involves: dynamically constructing a subgraph containing only nodes and edges allowed to pass under the current cargo status, and using this subgraph as the corresponding directed connected graph. If the second shuttle is in an empty state, the global directed connected graph is used as the directed connected graph; if the second shuttle is in a cargo status, the global directed connected graph is traversed, and nodes and edges corresponding to dedicated paths that only allow empty passage are removed to generate a directed connected subgraph for the cargo status.

[0088] After the directed connected graph is determined, the Tarjan algorithm or Kosaraju algorithm is used to find all strongly connected components. A strongly connected component is a maximal subset of the directed graph where there is a bidirectional reachable path between any two nodes. Two nodes belonging to the same strongly connected component means that it is possible to reach the other node from either node and return. Applying strongly connected component detection to departure point selection has the advantage that if the departure point and the task endpoint belong to the same strongly connected component, the second shuttle can reach the departure point from the task endpoint and can return to the task endpoint or continue to other locations when needed, without being unable to turn back due to entering a dead end. Strongly connected component detection guarantees bidirectional reachability, rather than unidirectional reachability.

[0089] Step S5.7: Sequentially determine whether each departure point belongs to the same strongly connected component as the task endpoint; take the first departure point that meets the condition as the target departure point.

[0090] Following the Manhattan distance determined in step S5.4 in ascending order, each departure point is checked one by one to see if it belongs to the same strongly connected component identified in step S5.6 as the task endpoint. The first departure point that simultaneously satisfies the condition of belonging to the same strongly connected component and passes the runtime filtering in step S5.2 is determined as the target departure point. This filtering process combines the conditions of spatial proximity and topological reachability and returnability, thereby ensuring both departure efficiency and that the departure path is truly reachable and returnable under the current cargo loading status.

[0091] The selection of departure points starts from all nodes on the map, and proceeds through static topology screening, runtime dynamic filtering, distance sorting, and strong connectivity component reachability verification based on cargo status awareness. This process is progressive, and finally, the target departure point is obtained that is structurally unobstructed, available at the current moment, the closest in distance, and bidirectionally reachable under the current cargo status.

[0092] Step S5.8: If there are no candidate nodes at this time, extract the location nodes of the devices in the Idle state (excluding the devices currently located at the end point) and filter them based on steps S5.4 to S5.7; if there are nodes that meet the requirements, take the node where the new device is located as the starting point, filter it through steps S5.4 to S5.7, and find the available transfer point for the device.

[0093] Step S5.8.1: If there are multiple filtering results, the sum of the Manhattan distances of the two devices to be moved from their respective target move points shall be used as the first sorting condition, and the absolute value of the difference between the Manhattan distances of the two devices shall be used as the second sorting condition to sort the resulting move point schemes based on the two devices.

[0094] Step S5.8.2: For the set of dual separation points (separation of two devices) obtained in step 5.8.1, re-execute steps S5.6~5.7. The first one that satisfies both conditions is the dual separation point scheme.

[0095] Step S6: Generate a transfer task and send it to the second shuttle (and the third shuttle, if present in the plan) so that it can drive to the target transfer point and release the task endpoint; after the second shuttle arrives at the target transfer point, update the device occupancy information of the task endpoint at the corresponding time in the global spatiotemporal graph.

[0096] A relocation task is generated, and the location coordinates and path information of the target relocation point are encoded into an executable command sequence for the second shuttle, which is then sent to the second shuttle's onboard control system. After the second shuttle executes the relocation task and departs from the task endpoint, the endpoint resource changes from an occupied state to an idle state. At this time, the device identifier of the second shuttle is removed from the device location information set of the task endpoint at the corresponding time, ensuring that the state of the global spatiotemporal map is consistent with the actual situation. This update operation provides an accurate data foundation for subsequent periodic conflict detection, preventing the endpoint from being misjudged as occupied for a long time due to residual data, and avoiding subsequent equipment being unable to use the endpoint resource due to incorrect information.

[0097] More specifically, the present invention further includes: when the second shuttle arrives at the target departure point, releasing its occupation of the task endpoint, and removing the device identifier of the second shuttle from the device location information set of the task endpoint at the corresponding time moment. This step corresponds to the update operation in step S6, ensuring that the release of the endpoint's occupation status and the update of the spatiotemporal map are completed synchronously.

[0098] The invention further includes: if the second shuttle is in a task-intensive state, or if the target relocation point cannot be selected, the first shuttle enters a waiting queue and periodically re-performs the spatiotemporal conflict detection until the task endpoint is released. This step provides a fault-tolerant path: when active relocation cannot be executed due to the occupied equipment having a task or no available relocation point, it switches to a passive waiting mode to ensure that the system does not crash due to relocation failure. The periodic retry mechanism enables the first shuttle to be awakened in the first detection cycle after the endpoint resource is released and continue to execute the task, avoiding indefinite suspension.

[0099] The relocation compensation assessment in step S5.5 includes: obtaining the first position coordinates of the second shuttle when it arrives at the target relocation point after completing the relocation; obtaining the original position coordinates of the second shuttle when it is not relocated; obtaining the starting position coordinates of the next task predicted to be assigned to the second shuttle in the system cache; and calculating the expected increase in empty runs ΔC caused by the relocation of the second shuttle.

[0100] ΔC =D(p new ,p next ) - D (p orig ,p next ).

[0101] Where, p new Let p be the first position coordinate. orig p represents the original position coordinates. next Let be the starting coordinates of the next task, and D be the Manhattan distance function.

[0102] While the relocation process resolves the issue of the current destination being occupied, the second shuttle's new location differs from its original position. The system cache may contain the next task assigned to the second shuttle but not yet executed. After the second shuttle completes its relocation, if the next task is triggered, the second shuttle needs to travel from its new location to the task's starting point. If it hasn't been relocated, the second shuttle travels from its original location to the task's starting point. The difference between the distance from the new location and the original location to the task's starting point is the additional empty travel distance incurred due to the relocation.

[0103] D(p new ,pnext D(p) represents the expected travel distance from the new location to the starting point of the next mission after the second shuttle has been relocated; orig ,p next The distance traveled by the second shuttle from its original position to the next task starting point is the expected distance if the second shuttle is not reassigned. The difference between the two, ΔC, is the additional empty travel distance caused by the reassignment. This calculation does not rely on predictions of future random events, but only on currently known deterministic information: tasks allocated but not yet dispatched in the system cache. Therefore, it is deterministic and repeatable, and can be accurately calculated before the reassignment decision is made.

[0104] The filtering condition for the departure point is that the expected increase in empty runs ΔC is less than or equal to a preset compensation threshold. Specifically:

[0105] (1) If the expected increase in empty runs ΔC is greater than the preset compensation threshold, the currently selected target departure point is abandoned, and the next departure point is selected from the temporary set to re-execute the departure compensation assessment step.

[0106] (2) If the expected increase in empty runs ΔC is less than or equal to the preset compensation threshold, the currently selected target departure point will be confirmed as the final departure location and a departure task will be generated.

[0107] The preset compensation threshold is a parameter configured by the system administrator based on actual business needs. When the expected increase in empty runs exceeds this threshold, it indicates that the efficiency loss of relocating the second shuttle to the target departure point is too great, and it is not worth sacrificing the efficiency of the second shuttle's subsequent tasks to release the current destination. At this time, the current departure point is abandoned, and the next candidate departure point with a greater distance is selected for re-evaluation until a departure point with an acceptable ΔC is found. This invention upgrades the departure decision from the traditional immediate availability to a completely new global efficiency optimization by introducing a departure compensation evaluation mechanism. That is, it does not execute the decision as soon as the first available departure point is found, but makes a decision after evaluating the comprehensive impact of the departure on the subsequent tasks of the second shuttle, avoiding damage to the overall system efficiency due to local departures.

[0108] For the hotspot conflict prediction in step S5.3, the following are included:

[0109] Step S5.3.1: Obtain the occupancy frequency f of the task endpoint within a preset historical time period. occupy The occupancy frequency is the ratio of the average duration of continuous occupancy of the task endpoint by the shuttle within a unit of time to the unit of time.

[0110] In this step, the occupation frequency reflects the congestion level of the task endpoint over a past period. If an endpoint is frequently occupied by equipment for extended periods, it is more likely to become a conflict hotspot. This frequency information is calculated by statistically analyzing equipment arrival, dwell, and departure events recorded in the system's operation logs. A higher occupation frequency indicates greater throughput pressure on the endpoint and a higher frequency of conflicts. This historical statistical information reflects the congestion patterns of different nodes during long-term operation, thus serving as the data basis for subsequent differentiated relocation strategies.

[0111] Step S5.3.2: Perform hotspot-based filtering on all candidate transfer points globally, ensuring that all candidate transfer points are far from hotspots, as detailed below:

[0112] 1) If the occupied frequency f occupy If the frequency exceeds the hotspot frequency threshold, the task endpoint is determined to be a hotspot conflict node.

[0113] The hotspot frequency threshold is based on the system's historical average occupancy rate or a parameter configured by the administrator. When the occupancy frequency exceeds this threshold, the system automatically marks the endpoint as a hotspot conflict node. Through threshold detection, the continuous variable (occupancy frequency) is transformed into a discrete state (hotspot / non-hotspot), enabling the system to automatically switch to an enhanced de-occupancy strategy when preset conditions are met.

[0114] 2) For task endpoints identified as hotspot conflict nodes, after runtime filtering, an enhanced relocation strategy is executed, which includes: removing candidate relocation points that are less than a first preset radius threshold from the temporary set, causing the second shuttle to be relocated to a storage node further away from the task endpoint, and ensuring that no device in an idle state without a task is retained within the first preset radius threshold range around the task endpoint before the arrival of the first shuttle.

[0115] Conventional relocation strategies only require moving the device occupying the endpoint to any available relocation point, without restricting the distance of that relocation point from the endpoint. However, for hotspot conflict nodes, because their occupation frequency is significantly higher than that of ordinary nodes, even if the currently occupying device is moved, other idle devices may reoccupy the endpoint within a short period of time, causing the endpoint to be occupied again when subsequent devices arrive, resulting in repeated conflicts.

[0116] The enhanced relocation strategy of this invention removes candidate relocation points that are less than a first preset radius threshold from the endpoint, forcing the occupied device to be relocated to a more distant storage node. Therefore, no device in a task-free idle state exists within the preset radius around the endpoint. When the first shuttle arrives at the endpoint, it will not be re-occupied due to the presence of idle devices near the endpoint, thus effectively reducing the repeated conflict rate of hotspot conflict nodes.

[0117] For step S5.8, obtain the set S of all shuttles in the warehouse system that are in an idle state with no tasks. idle .

[0118] In traditional conflict resolution methods, when the destination is occupied by a second shuttle, the system only considers moving the second shuttle without evaluating whether other idle equipment is more suitable for relocation. However, the second shuttle may be closest to the destination, but its subsequent tasks are urgent or relocation is costly; while another idle equipment that is slightly farther away but has no subsequent tasks may be more suitable for relocation. Therefore, this invention establishes a global idle equipment resource view, including all shuttles in a task-free idle state in a unified evaluation scope, enabling relocation decisions to find the optimal solution globally.

[0119] If the second shuttle belongs to the shuttle set S idle If so, the second shuttle car is given priority as the departure device, and with the goal of minimizing the global task completion time, a target departure point is assigned to the second shuttle car from the set of candidate departure points;

[0120] If the second shuttle cannot complete the relocation request independently, then from the shuttle set S idle The third shuttle car is selected as an auxiliary transfer device and the transfer is performed in a chain transfer manner. The chain transfer means that the third shuttle car is first transferred to the first transfer point, and then the second shuttle car is transferred to the second transfer point. The chain transfer is limited to one layer and no recursive attempts are made.

[0121] The second shuttle cannot complete the relocation request independently, including but not limited to the following situations:

[0122] (1) The currently selected target transfer point is occupied by a third device, and the device is in an idle state with no task.

[0123] (2) The path of the second shuttle to the target transfer point is blocked by the third device.

[0124] (3) The transfer path of the second shuttle car conflicts with the current position of the third equipment.

[0125] The selection of candidate devices for relocation can be based on the following evaluation dimensions: the distance from the current location of each idle device to the destination, the impact of relocation on subsequent tasks, and the current loading status of each idle device. Minimizing the global task completion time means evaluating the comprehensive cost of different idle devices as relocation targets and selecting the device with the least impact on overall system efficiency for relocation. Through a global idle device pool mechanism, the relocation decision is transformed from a locally optimal approach of immediate occupancy and departure to a globally optimal approach that comprehensively considers the status of all idle devices, avoiding adverse effects on other high-priority tasks due to inappropriate relocation targets.

[0126] Therefore, this invention improves the conflict resolution method for four-way shuttles from passive waiting to active relocation, eliminating uncertain delays caused by waiting for new task assignments. Candidate relocation point screening limits locations to end-point storage locations, ensuring that equipment does not occupy lanes or create new obstacles after parking; runtime filtering ensures that relocation points are truly available at the current moment; strong connectivity component verification combined with cargo status dynamically constructs a subgraph to ensure the reachability of the relocation path; relocation compensation assessment quantifies the impact of relocation on subsequent tasks, avoiding local relocation from harming overall efficiency; hotspot conflict prediction employs an enhanced strategy of expanding the clearing radius for frequently congested endpoints, reducing the probability of repeated conflicts. This significantly reduces the risk of deadlock and congestion probability in multi-shuttle scheduling scenarios, improving the overall operational efficiency of the warehousing system.

[0127] The present invention also provides some other embodiments, as follows:

[0128] Example 1: As Figure 1 As shown, multiple devices have the same shipping point: When multiple devices, such as device A and device B, have the task of delivering goods to the shipping port node P, the device that arrives first, assuming it is B, will prevent device A from moving forward. This is because B occupies the destination P, and when no new task is assigned, B will usually wait in place, causing device A to wait as well, until node P is released by device B.

[0129] When the above scenario occurs, the required resource point (path endpoint) is occupied by other devices, which means that completing the current task requires waiting for the device that has occupied the relevant resource point to leave. In order to reduce the waiting time and avoid the chain reaction (such as deadlock) that may be caused by long waiting, the system will actively look for possible relocation points for the device and relocate it so that the relevant task can continue.

[0130] Example 2: Read the map and connectivity information, and store the relevant data in the following NodeInfo structure for convenient subsequent calculations. Here, NodeInfo is represented as a tuple as follows:

[0131] NodeInfo = (ID, x, y,z, attribute,adjacent).

[0132] Where ID is the node number, x, y, z represent the x, y, z coordinates of the node, respectively, attribute is a set of attribute markers for the node, such as attribute names like IsDisable, IsItem, IsStorage, IsElevator, IsPort, etc., with corresponding values ​​of true or false to describe the state of the resource point, and adjacent is the other endpoints of the directed edges originating from this node and their weights (path lengths). adjacent can contain the following directions: XPos, XNeg, YPos, YNeg, ZPos, Zneg. Here, X, Y, Z represent the directions along the X, Y, Z coordinate system, and Pos and Neg represent the positive or negative directions. At the same time, since x, y, z are the spatial coordinates of the node, this direction is also equivalent to starting from the current node and moving towards the front, back, left, and right layers (XPos, XNeg, YPos, YNeg) in the horizontal plane.

[0133] For example: Map node set V = {Node1, Node2, ..., Node...} n}, the valid node P is represented as:

[0134]

[0135] The attribute[IsDisable] indicates whether the node is disabled for use / access.

[0136] The attribute[IsItem] indicates whether there is any goods placed at this node.

[0137] The attribute[IsStorage] indicates whether the node is a shelf or an aisle.

[0138] The attribute[IsElevator] indicates whether the node is an elevator.

[0139] The attribute[IsPort] indicates whether the node is a conveyor belt.

[0140] The attribute [IsCheckNeighbor] indicates whether to check adjacent nodes when stopping.

[0141] adjacent[XPos]=(ID2,weight2) represents the ID of the positive neighbor node X and the path weight.

[0142] adjacent[XNeg]=(ID3,weight3) represents the ID of the negative neighbor node X and the path weight.

[0143] adjacent[YPos]=(ID4,weight4) represents the ID of the positive adjacent node in the Y direction and the path weight.

[0144] adjacent[YNeg]=(ID5,weight5) represents the ID of the negative neighbor node in the Y direction and the path weight.

[0145] Example 3: In order to perform relevant spatiotemporal detection, a spatiotemporal node is established to describe the corresponding changes of each device in the map over time, TSPoint=(ID,state[tick]). Here, the ID of the node is the same as the ID of the node at the same location in the map, and state is a set of device location information based on a time sequence. At a specific time, the state node has three possibilities: no device, one device, or more than one device.

[0146] Establish a global spacetime graph: W = {Point1, Point2, ..., Point...} n}

[0147] Therefore, a valid TSPoint node is represented as:

[0148] TSPoint=(ID,tick1:(D k ),...,tick m :(D i D j ),...,tick n :()).

[0149] Among them, D k Tick ​​represents the k-th device. m :(D i D j ) indicates that at time m, device D i D j It will appear at this location.

[0150] Example 4: To facilitate tracking the usage status of the transfer point, the transfer point structure is defined as follows:

[0151] Helipad = {DeviceId,Node}.

[0152] Here, DeviceId is the device ID, corresponding to the device ID using this departure point, and Node is the information of the node in the map corresponding to the departure point after NodeInfo is instantiated.

[0153] Example 5: Figure 2 As shown, when the planner starts running, it preprocesses the map nodes to obtain all possible candidate relocation points.

[0154] The selection principle for candidate transfer points is: nodes with high connectivity (allowing for quick access from the rack to the aisle to continue the handling task) and low likelihood of obstruction (obstruction refers to the relatively small possibility that the equipment at the transfer point will become an obstacle in the path of other equipment). Therefore, the selected nodes are storage locations (racks) that are adjacent to the aisle.

[0155] Example 6: When a new task planning path is received for the device:

[0156] Based on the connected graph, find the connected paths L{n1,n2,...,n...} that start from the current position and reach a specific stage of the task. x}, (x∈N, x≥1).

[0157] Begin the path planning and detection process:

[0158] Based on the existing spatiotemporal location information of the equipment, find the starting point of the equipment in the spatiotemporal map.

[0159] Traverse all nodes of the connected path and add the nodes to the corresponding positions in the spatiotemporal graph according to the order of time ticks.

[0160] Iterate through all nodes of the device in the spacetime graph until the endpoint.

[0161] Whether other devices are occupied at the task endpoint is determined by: count(TSPoint[n x [tick] x ])>1, where count is a function to find the number of elements in a set. If:

[0162] Yes, and the device currently has no task. Based on the destination, find the transfer point and send a transfer task to the device.

[0163] Yes, but the device is currently on a mission and does not need to be relocated.

[0164] No, there is no conflict, and there is no need for relocation.

[0165] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for resolving conflicts involving a four-way shuttle, characterized in that: Includes the following steps: Step S1: Construct a global spatiotemporal graph, which records the equipment occupancy information of each node in the warehousing system at each future time step; Step S2: In response to assigning a task to the first shuttle, plan a first path from the current position of the first shuttle to the task endpoint; Step S3: Map the first path to the global spatiotemporal graph and perform spatiotemporal conflict detection; Step S4: When it is detected that the destination of the task is occupied by the second shuttle at the specific time when the first shuttle is expected to arrive, determine the task status of the second shuttle; Step S5: If the second shuttle is in a no-task state, then take the task endpoint as the reference point and perform the following screening steps in sequence to obtain the target departure point; Step S5.1: Traverse all nodes in the map and add nodes that meet the first filtering condition as candidate transfer points to the candidate transfer point set; the first filtering condition includes: the node is not disabled; the node is a storage node; the node has one and only two adjacent nodes, one of which is a storage node and the other is a lane node; and both adjacent nodes are not disabled and are not hoist nodes. Step S5.2: Perform runtime filtering on the candidate departure point set to obtain a temporary set; the runtime filtering is used to exclude candidate departure points that are not available at the current time. Step S5.3: Hotspot Conflict Prediction: If there is a frequency f of node being occupied in the historical time period corresponding to the current time period. occupy Based on the candidate transfer point set in step S5.1, the set hotspot frequency threshold is applied to filter all candidate points to obtain the available candidate transfer point set for the current time period. Step S5.4: Calculate the Manhattan distance from the task endpoint to each departure point in the temporary set, and sort them in ascending order based on the Manhattan distance; Step S5.5: Displacement Compensation Assessment: Based on the calculation results, distance filtering is performed using the expected increase in empty runs ΔC function, based on the function ΔC = D(p new ,p next ) - D (p orig ,p next ) Obtain all ΔC values ​​for existing candidate nodes, and retain only candidate nodes whose ΔC falls within the specified range; Step S5.6: Based on the cargo loading status of the second shuttle, determine the corresponding directed connected graph and find all strongly connected components in it; Step S5.7: Sequentially determine whether each departure point belongs to the same strongly connected component as the task endpoint; take the first departure point that meets the condition as the target departure point; Step S5.8: If there are no candidate nodes at this time, the device location nodes in the Idle state are extracted and filtered based on steps S5.4 to S5.7; if there are nodes that meet the requirements, the node where the new device is located is used as the starting point, and filtered through steps S5.4 to S5.7 to find the available transfer point for the device. Step S5.8.1: If there are multiple filtering results, the sum of the Manhattan distances of the two devices to be moved from their respective target moving points shall be used as the first sorting condition, and the absolute value of the difference between the Manhattan distances of the two devices shall be used as the second sorting condition to sort the resulting moving point schemes based on the two devices. Step S5.8.2: For the set of two separation points obtained in step 5.8.1, repeat steps S5.6~5.

7. The first one that satisfies both conditions is the double separation point scheme. Step S6: Generate a transfer task and send it to the second shuttle so that it can drive to the target transfer point and release the task endpoint; after the second shuttle arrives at the target transfer point, update the device occupancy information of the task endpoint at the corresponding time in the global spatiotemporal graph.

2. The method for resolving conflicts involving a four-way shuttle as described in claim 1, characterized in that: The relocation compensation assessment in step S5.5 includes: Obtain the first position coordinates of the second shuttle when it arrives at the target departure point after completing the departure; Obtain the original position coordinates of the second shuttle car without it being moved; Obtain the starting position coordinates of the next task predicted to be assigned to the second shuttle from the system cache; Calculate the expected increase in empty runs ΔC due to the second shuttle being relocated: ΔC =D(p new ,p next ) - D (p orig ,p next ); Where, p new Let p be the first position coordinate. orig p represents the original position coordinates. next Let D be the coordinates of the starting point of the next task, and let D be the Manhattan distance function. The filtering condition is to use the expected increase in empty runs ΔC as a filter condition for the departure point.

3. The method for resolving conflicts involving a four-way shuttle as described in claim 1, characterized in that: For the hotspot conflict prediction in step S5.3, the following are included: Obtain the frequency f of the task endpoint being occupied within a preset historical time period. occupy The frequency of occupation is the ratio of the average duration of continuous occupation of the task endpoint by the shuttle within a unit of time to the unit of time. If the occupied frequency f occupy If the frequency exceeds the hotspot frequency threshold, the task endpoint is determined to be a hotspot conflict node. The global candidate transfer points are filtered based on hotspots, so that all candidate transfer points are far away from hotspots.

4. The method for resolving conflicts involving a four-way shuttle as described in claim 1, characterized in that: For step S5.8, obtain the set S of all shuttles in the warehouse system that are in an idle state with no tasks. idle ; If the second shuttle belongs to the shuttle set S idle If so, the second shuttle car is given priority as the departure device, and with the goal of minimizing the global task completion time, a target departure point is assigned to the second shuttle car from the set of candidate departure points; If the second shuttle cannot complete the relocation request independently, then from the shuttle set S idle The third shuttle car is selected as an auxiliary transfer device and the transfer is performed in a chain transfer manner. The chain transfer means that the third shuttle car is first transferred to the first transfer point, and then the second shuttle car is transferred to the second transfer point. The chain transfer is limited to one layer and no recursive attempts are made.

5. The method for resolving conflicts involving a four-way shuttle as described in claim 1, characterized in that: Mapping the first path to the global spatiotemporal graph specifically includes: Obtain the path length of each node in the first path, and calculate the estimated arrival time of each node in the first path based on the preset speed of the first shuttle. The device identifier of the first shuttle is recorded in the device location information set of the corresponding node at the corresponding time in the global spatiotemporal graph.

6. The method for resolving conflicts involving a four-way shuttle as described in claim 1, characterized in that: The conditions for runtime filtering include: The candidate departure point is not the endpoint of the task; The candidate departure points were not disabled; No goods were placed at the candidate departure point; The candidate departure point was not locked or occupied by any device.

7. The method for resolving conflicts involving a four-way shuttle as described in claim 1, characterized in that: When sorting based on the Manhattan distance, the following is also included: Candidate transfer points whose Manhattan distance is less than a preset minimum transfer distance threshold are filtered out.

8. A method for resolving conflicts involving a four-way shuttle as described in claim 1, characterized in that: The determination of the corresponding directed connected graph based on the cargo loading status of the second shuttle is specifically as follows: Based on the cargo status of the second shuttle, a subgraph is dynamically constructed that contains only the nodes and edges that are allowed to pass under the current cargo status, and this subgraph is used as the corresponding directed connected graph.

9. A method for resolving conflicts involving a four-way shuttle as described in claim 1, characterized in that: Also includes: When the second shuttle arrives at the target departure point, it releases its occupation of the task endpoint and removes the device identifier of the second shuttle from the set of device location information of the task endpoint at the corresponding time.

10. A method for resolving conflicts involving a four-way shuttle as described in claim 1, characterized in that: Also includes: If the second shuttle is in a mission state or fails to select the target departure point, the first shuttle enters the waiting queue and periodically re-performs the spatiotemporal conflict detection until the mission endpoint is released.

Images