Three-dimensional storage multi-axis linkage intelligent goods shelf position dynamic allocation scheduling system

Abstract

The present application relates to the technical field of stereoscopic storage automation, in particular to a stereoscopic storage multi-axis linkage intelligent goods shelf position dynamic allocation and scheduling system, which comprises a task analysis module, an environment perception module, a goods position recommendation module, a scheduling optimization module and a plan execution module. The task analysis module receives and analyzes various storage operation instructions in real time. The environment perception module dynamically perceives the real-time state of the whole warehouse area. The goods position recommendation module generates multiple candidate goods position sets for the goods to be stored according to the goods attributes and storage requirements, in combination with the real-time state of the whole warehouse area, by using an improved collaborative filtering algorithm based on the correlation of goods storage characteristics. The scheduling optimization module evaluates each candidate set according to the preset optimization target, selects the final goods position, and generates a detailed storage execution plan containing the collaborative operation path of multiple tunnel stacker cranes. The plan execution module schedules the equipment execution plan and monitors and updates the state in real time.

Description

Technical Field

[0001] This invention relates to the field of automated storage and retrieval systems, and more particularly to a dynamic allocation and scheduling system for multi-axis linkage intelligent racking locations in automated storage and retrieval systems. Background Technology

[0002] In automated storage and retrieval systems (AS / RS), the allocation and scheduling of storage locations is a core aspect. Current technologies often rely on static rules or strategies for location allocation, such as fixed locations, zoning by product category, or simple principles like proximity-based entry and first-in-first-out (FIFO). These methods define rules during the planning phase but lack flexibility in actual operation. Another common approach is allocation based on real-time availability of storage locations. While this considers immediate availability, the decision-making process is singular, typically focusing only on the shortest distance or the existence of available spaces, without fully considering the storage characteristics of the goods, the storage relationships between goods, and the coordination and path planning issues when multiple storage and retrieval devices operate in parallel. This leads to bottlenecks in warehouse space utilization, goods turnover efficiency, and overall equipment operating efficiency. An unreasonable storage location layout can increase retrieval paths for similar or related goods, resulting in low efficiency for subsequent outbound or inventory checks. Furthermore, without global collaborative path planning, multiple stacker cranes performing tasks are prone to operational conflicts and congestion, failing to fully utilize their advantages and limiting overall inbound and outbound throughput. Therefore, there is an urgent need for an intelligent storage location allocation method that can deeply integrate cargo characteristics and warehouse dynamics and optimize the collaborative scheduling of multiple devices, in order to solve the problems of inflexible static rules, local optima caused by single-objective optimization, and efficiency bottlenecks caused by independent operation of multiple devices. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of existing technologies and propose a dynamic allocation and scheduling system for multi-axis linkage intelligent shelving locations in three-dimensional warehousing.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: a multi-axis linkage intelligent racking system for automated warehousing with dynamic allocation and scheduling of storage locations, comprising: The task parsing module receives and parses inbound task orders, outbound task orders, and inventory count instructions issued by the warehouse management system in real time. The inbound task order contains the physical attributes and storage requirements of the goods to be received, and the outbound task order contains the location code and demand priority of the goods to be shipped. The environmental perception module, based on the electronic map of the automated warehouse and the shelf status sensors, dynamically perceives the real-time occupancy status, load-bearing capacity, and cargo storage and retrieval history of all storage locations in the entire warehouse area. The storage location recommendation module generates multiple candidate storage location sets for the goods to be stored based on the physical attributes and storage requirements of the goods to be stored, and in combination with the real-time occupancy status and load-bearing capacity of storage locations throughout the warehouse area, using an improved collaborative filtering recommendation algorithm. The improved collaborative filtering recommendation algorithm is optimized based on the correlation of the storage characteristics of the goods. The scheduling optimization module evaluates multiple candidate storage locations based on preset optimization goals, selects the final storage location from the multiple candidate storage locations based on the evaluation results, and generates a detailed inbound execution plan that includes the collaborative operation path of multiple aisle stacker cranes. The planning and execution module schedules the stacker cranes and conveyor belt systems in the aisle to execute the warehousing execution plan, and monitors the equipment operating status and changes in the occupancy of storage locations in real time during the execution process, and updates the real-time occupancy status of storage locations.

[0005] As a further aspect of the present invention, the improved collaborative filtering recommendation algorithm is optimized based on the correlation of cargo storage features, and its working principle includes: Construct a cargo storage feature matrix. The rows of the matrix represent the historical stored goods in the warehouse, and the columns of the matrix represent the storage feature dimensions of the goods. The storage feature dimensions include the size, weight, storage temperature requirements, moisture protection level, owner code, and expected storage period of the goods. Calculate the matching degree between the storage feature vector of the goods to be put into storage and each row vector in the storage feature matrix of the goods, and filter out the set of historical stored goods with a matching degree higher than a preset threshold. Extract the location codes of all goods that have been stored in the historical storage goods set, and count the frequency of use and cumulative storage duration of each location by goods in the historical storage goods set; Based on the frequency of use of storage locations by historical stored goods and the cumulative storage duration, calculate the recommended confidence level for each storage location for goods to be received. The recommendation confidence level is dynamically adjusted based on the real-time occupancy status and load capacity of the current storage location. The adjustment factors include whether the storage location is currently vacant and the ratio of the current load capacity to the maximum load capacity. Based on the dynamically adjusted recommendation confidence level, all storage locations are sorted in descending order. Several storage locations that rank highly and meet the basic storage requirements of the goods to be stored are selected to form a preliminary set of candidate storage locations.

[0006] As a further aspect of the present invention, the step of calculating the matching degree between the storage feature vector of the goods to be stored and each row vector in the goods storage feature matrix, and filtering out a set of historically stored goods with a matching degree higher than a preset threshold, includes: Numericalize and normalize the various storage characteristics of the goods to be stored, forming a standard storage characteristic vector of the goods to be stored. The same numerical and normalization processing is applied to the storage feature vectors of historical stored goods represented by each row in the cargo storage feature matrix. Calculate the cosine similarity between the standard warehousing feature vector of the goods to be put into storage and the processed standard feature vector of each row in the goods warehousing feature matrix; Set a matching threshold and compare the calculated cosine similarity with the matching threshold; Filter out all historical stored goods whose cosine similarity is greater than or equal to the matching degree threshold, and use the set of historical stored goods as the set of historical stored goods with a matching degree higher than the preset threshold.

[0007] As a further aspect of the present invention, the evaluation of multiple candidate storage locations based on a preset optimization objective includes: The optimization objectives include stabilizing the rack center of gravity, predicting future outbound efficiency, and optimizing the energy consumption of stacker cranes operating in aisles. To optimize the stability of the rack center of gravity, the offset of the rack center of gravity coordinate in three-dimensional space is calculated after the goods to be stored are placed in each of the candidate storage locations, and it is evaluated whether the offset exceeds the set safety range. To optimize future outbound efficiency, based on historical cargo storage and retrieval records, the historical average outbound response time of the aisles and floors of the candidate storage locations is analyzed. Combined with the expected storage period of the goods to be stored and the outbound frequency of the owner's code, the future outbound operation time cost is predicted. To optimize the energy consumption of collaborative operation of stacker cranes in aisle areas, this study simulates and calculates the total running distance and number of start-stop operations of multiple stacker cranes when performing the task of storing goods in various alternative storage locations, taking into account the current position status and task queue of multiple stacker cranes in aisle areas. Weighting coefficients were assigned to the center of gravity offset, predicted outbound time cost, total running distance, and number of start-stop cycles, and a comprehensive evaluation score was calculated for each candidate storage location. The overall evaluation scores of all storage locations belonging to the same set of candidate storage locations are aggregated and calculated to obtain the overall evaluation score of each set of candidate storage locations.

[0008] As a further aspect of the present invention, the step of calculating the offset of the center of gravity coordinate of the shelf in three-dimensional space after placing the goods to be stored into each of the candidate storage locations in the shelf center of gravity stability optimization objective includes: Establish a three-dimensional geometric model and a mass distribution model for each set of shelves in the automated warehouse. The mass distribution model records the weight of the goods currently stored in each location on the shelf and the position of the center of gravity of the goods themselves. Obtain the weight of the goods to be put into storage and the position of the preset center of gravity mark on the packaging box; For each candidate storage location in the set of alternative storage locations, the weight and center of gravity information of the goods to be stored are virtually added to the mass distribution model of the shelf to which the candidate storage location belongs. Recalculate the center of gravity coordinates of the entire rack in the three-dimensional coordinate system after virtual addition, and compare them with the design center of gravity coordinates of the rack in the unloaded state. Calculate the Euclidean distance between the virtual added center of gravity coordinates and the design center of gravity coordinates as the center of gravity offset. The calculated center of gravity offset of all candidate storage locations is compared with the maximum safe offset threshold allowed by the rack model, and candidate storage locations whose offset exceeds the safe threshold are marked.

[0009] As a further aspect of the present invention, the optimization of energy consumption for the coordinated operation of stacker cranes in aisle areas, combined with the current position status and task queue of multiple stacker cranes, simulates and calculates the total operating distance and number of start-stop cycles of the multiple stacker cranes when performing the task of storing goods into various alternative storage locations, including: Obtain the real-time three-dimensional coordinates of all stacker cranes in the current automated warehouse and their current task queue information; For each alternative storage location, based on the preset stacker crane task allocation strategy, simulate the allocation of a stacker crane from currently idle or soon-to-be-idle stacker cranes for the task of transporting goods from the pickup point to the alternative storage location. When simulating task allocation, consider the expected position and status of the stacker crane after it has completed all tasks in its existing task queue; Based on the expected starting position of the stacker crane assigned the task in the simulation and the position of the alternative storage location, a three-dimensional movement path from the starting point to the final storage location is planned in the navigation grid of the automated warehouse. Calculate the geometric length of the planned path in three-dimensional space, which is the theoretical running distance for the stacker crane to perform this task; Meanwhile, based on the number of times the stacker crane changes its direction of movement in the horizontal, vertical, and fork extension directions in the planned path, the number of times the stacker crane starts and stops on the path is simulated and calculated. Each state transition from stationary to moving or from moving to stationary is counted as one start and stop. For all stacker cranes that are simulated to be assigned to the warehousing task, their theoretical running distance and number of start-stops are summed to obtain the simulated total running distance and simulated total number of start-stops of multiple stacker cranes when storing goods in the alternative storage location.

[0010] As a further aspect of the present invention, the step of selecting the final storage location from a set of multiple candidate storage locations based on the evaluation results and generating a detailed inbound execution plan that includes the collaborative operation paths of multiple aisle stacker cranes includes: Compare the overall evaluation scores of multiple alternative storage location sets, and select the alternative storage location set with the highest overall evaluation score as the preferred set; In the preferred set, the final storage location is selected according to the preset emergency rules. The emergency rules include selecting the storage location with the shortest path when the inbound task is urgent, or selecting the storage location with the most stable center of gravity when the task is stable. After determining the final storage location, based on the real-time location, status, and task queue of all current aisle stacker cranes, assign appropriate aisle stacker cranes to the handling tasks of goods to be stored. To plan the three-dimensional motion path of the assigned stacker crane from the picking point through the transfer point to the final storage location, the path planning needs to consider the kinematic constraints of the stacker crane and avoid conflicts with the paths of other stacker cranes. Integrate the motion paths, sequences, and time nodes of all stacker cranes participating in the collaborative operation to generate a detailed inbound execution plan. The plan includes the motion command sequence, speed curve, and expected start and end times of each stacker crane.

[0011] As a further aspect of the present invention, the planning of a three-dimensional movement path for the allocated stacker crane from the picking point through the transfer point to the final storage location includes: The three-dimensional space of the automated warehouse is discretized into a navigation grid composed of nodes, which are located at the center of the storage location, the intersection of the aisle, and the docking point of the equipment. The improved A* search algorithm, which takes into account dynamic obstacles, is used to search the navigation grid. Dynamic obstacles include the space occupied by other stacker cranes performing tasks and their predetermined paths. During the search process, different cost weights are assigned to the horizontal movement, vertical lifting, and fork extension and retraction of the stacker crane to reflect the differences in energy consumption and time for different axial movements. Once a preliminary path is found, it is smoothed and optimized to convert it into a smooth motion trajectory with continuous speed and acceleration of each drive shaft of the stacker crane. Collision detection is performed on the smoothed motion trajectory to ensure that at any point in time, the stacker crane performing the task maintains a safe distance from other stacker cranes and fixed facilities.

[0012] As a further aspect of the present invention, the scheduling aisle stacker crane and conveyor belt system execute the warehousing execution plan, and monitor the equipment operating status and changes in cargo space occupancy in real time during the execution process, including: The detailed inbound execution plan is broken down into independent equipment control instructions, and these instructions are issued to the corresponding stacker crane controllers and conveyor system controllers in chronological order. It receives real-time feedback from the stacker crane controller on motor current, encoder position, photoelectric sensor status information, as well as speed feedback and cargo presence detection information from the conveyor belt system. The feedback equipment status information is compared with the expected status in the inbound execution plan in real time to detect whether there is a positional deviation, timeout or equipment failure. When the stacker crane in the aisle completes the goods storage and retrieval operation and leaves the storage location, the occupancy status of the corresponding storage location in the automated warehouse electronic map is immediately updated, changing from "occupied" to "idle", or from "idle" to "occupied", and the goods code is recorded. If a device malfunction or a serious delay in task execution is detected during real-time monitoring, a dynamic rescheduling process will be triggered.

[0013] As a further aspect of the present invention, the triggering of the dynamic rescheduling process includes: Immediately suspend the current task of the affected stacker crane or conveyor section in the aisle and mark its status as "faulty" or "unavailable"; Assess the impact of the failure on the currently executing inbound execution plan and subsequent queued tasks, and identify all tasks that are blocked or cannot be executed as originally planned; Reacquire the status of currently available equipment resources and the real-time occupancy status of all storage locations; For the identified affected tasks, based on the reacquired resource and status information, and taking into account the estimated recovery time of the faulty equipment, the storage locations, equipment, and planned routes are reassigned to them. Generate a new local execution plan and coordinate with other normal equipment to work together with the new plan. Issue control commands to bypass the fault point or resume operation after waiting.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are as follows: An improved collaborative filtering algorithm, optimized based on the correlation of cargo storage characteristics, is employed for warehouse location recommendation, departing from the traditional allocation logic based on fixed rules or simple space occupancy. This algorithm analyzes historical warehousing data to identify groups of goods with similarities or correlations in physical attributes, storage requirements, and access frequency. When recommending new warehouse locations, the algorithm considers not only the real-time availability and load capacity of the locations but also the correlation between the target goods and existing goods in the warehouse as a key weight. This allows the system to spatially cluster or place goods with compatible storage requirements, similar turnover patterns, or frequent simultaneous requests in locations with related logic. The result is a more optimized cargo layout structure from both a global and long-term operational perspective. This layout reduces storage space waste caused by conflicting cargo attributes and enables centralized management of similar goods. More importantly, it reduces the overall movement path of storage and retrieval equipment for subsequent operations such as related outbound shipments and batch inventory checks, improving operational continuity and allowing the warehouse's static storage layout to directly serve dynamic operational efficiency.

[0015] A scheduling optimization mechanism based on multi-alternative scheme evaluation and multi-device collaborative path planning was introduced, enabling a leap from single-point storage location selection to global operation plan generation. The scheduling optimization module receives multiple sets of alternative storage locations from the storage location recommendation module, each set representing a feasible space allocation scheme. The module evaluates each scheme based on preset composite optimization objectives such as minimum total operation time, balanced equipment load, and minimum energy consumption. The evaluation process simulates the paths, time, and conflict resolution strategies required for multiple aisle stacker cranes to collaboratively complete current and potential queuing tasks under the given storage location allocation scheme. By comparing the evaluation results, the system selects the overall optimal final storage location and simultaneously generates a detailed, time-sequential multi-machine collaborative operation execution plan. This plan precisely defines the action sequence, travel path, and avoidance logic at intersections for each stacker crane. The effect is that the storage location allocation decision for a single inbound task is embedded into the context of continuous dynamic operations throughout the warehouse for optimization. This avoids the problem of local optima in a single task leading to subsequent system efficiency degradation. When multiple devices operate under a clear collaborative plan, parallel operation can be maximized, reducing idle waiting time caused by path intersection conflicts and significantly shortening the completion time of tasks. This improves the overall throughput capacity and equipment utilization rate of the automated warehouse during peak periods. Attached Figure Description

[0016] Figure 1 The flowchart is a process for the dynamic allocation and scheduling system of multi-axis linkage intelligent racking in three-dimensional warehousing described in this invention. Figure 2 A flowchart illustrating the working principle of the improved collaborative filtering recommendation algorithm; Figure 3 Flowchart for multi-objective optimization evaluation of candidate storage locations. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0018] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0019] See Figure 1 This invention provides a dynamic allocation and scheduling system for multi-axis linkage intelligent shelving locations in automated warehousing, specifically including: The task parsing module receives and parses various operational instructions from the warehouse management system in real time. These instructions include inbound task orders, outbound task orders, and inventory counting instructions. Inbound task orders contain the physical attributes and specific storage requirements of the goods to be received, while outbound task orders specify the location code and priority of the goods to be shipped. The environmental perception module, based on a pre-built electronic map of the automated warehouse, continuously receives data from status sensors deployed throughout the shelves to dynamically perceive the real-time occupancy status, current load capacity, and historical storage and retrieval records of all locations within the entire warehouse area. The location recommendation module, based on the inbound goods information provided by the task parsing module and combined with the real-time location status of all locations in the warehouse area provided by the environmental perception module, uses an improved collaborative filtering recommendation algorithm to generate multiple sets of alternative locations for the inbound goods. This algorithm, based on the traditional collaborative filtering concept, is specifically optimized according to the correlation between the storage characteristics of the goods. The scheduling optimization module comprehensively evaluates the set of candidate storage locations generated by the storage location recommendation module based on multiple preset optimization objectives. Based on the evaluation results, it selects a final storage location from the multiple sets and generates a detailed inbound execution plan. This plan includes the path arrangement for the collaborative operation of multiple aisle stacker cranes. The plan execution module is responsible for scheduling the specific aisle stacker cranes and conveyor belt systems to execute the inbound plan. During execution, it monitors the operating status of the equipment and changes in storage location occupancy in real time, and updates the system's real-time storage location occupancy status accordingly.

[0020] In one embodiment of the present invention, see [reference] Figure 2 A cargo storage feature matrix is ​​constructed, where rows represent historically stored goods in the warehouse, and columns represent various storage feature dimensions of the goods. These dimensions include cargo size, weight, storage temperature requirements, moisture protection level, owner code, and expected storage period. The matching degree between the storage feature vector of the goods to be received and the vectors in each row of the cargo storage feature matrix is ​​calculated, and a set of historically stored goods with a matching degree higher than a preset threshold is selected. This process includes numericalizing and normalizing the various storage features of the goods to be received to form their standard storage feature vectors. The same numericalizing and normalizing process is applied to the storage feature vectors of the historically stored goods represented in each row of the cargo storage feature matrix. The cosine similarity between the standard storage feature vector of the goods to be received and the processed standard feature vectors in each row of the matrix is ​​calculated. A matching degree threshold is set, and the calculated cosine similarity is compared with this threshold. All historically stored goods with a cosine similarity greater than or equal to the matching degree threshold are selected, forming a set of historically stored goods.

[0021] Extract the location codes of all goods previously stored in the historical storage goods set, and calculate the frequency of use and cumulative storage duration of each location for goods in the set. Based on the frequency of use and cumulative storage duration, calculate the recommended confidence level for each location for the goods currently awaiting warehousing. Dynamically adjust the recommended confidence level by considering the real-time occupancy status and load capacity of the current location, with adjustment factors including whether the location is currently idle and the ratio of current load capacity to maximum load capacity. Sort all locations in descending order based on the dynamically adjusted recommended confidence level, and select several locations that rank highly and meet the basic storage requirements of the goods awaiting warehousing to form a preliminary candidate location set.

[0022] In its implementation, the storage location recommendation module executes an improved collaborative filtering recommendation algorithm. This algorithm is optimized based on the correlation of goods storage features. The algorithm's working principle includes constructing a goods storage feature matrix. The rows of the matrix represent historically stored goods in the warehouse, and the columns represent the storage feature dimensions of the goods. These dimensions include goods size, weight, storage temperature requirements, moisture protection level, owner code, and expected storage period. The matching degree between the storage feature vector of the goods to be received and each row vector in the goods storage feature matrix is ​​calculated. A set of historically stored goods with a matching degree higher than a preset threshold is selected. Storage location codes for all goods in the historically stored goods set are extracted, and the frequency and cumulative storage time of each storage location used by the historically stored goods set are statistically analyzed. Based on the frequency and cumulative storage time of storage locations used by the historically stored goods set, the recommendation confidence level for each storage location for the goods to be received is calculated. The recommendation confidence level is dynamically adjusted based on the current real-time occupancy status and load capacity of the storage location. Adjustment factors include whether the storage location is currently idle and the ratio of its current load capacity to its maximum load capacity. All storage locations are sorted in descending order based on the dynamically adjusted recommendation confidence level. Several storage locations that rank highly and meet the basic storage requirements of the goods to be stored are selected to form a preliminary set of candidate storage locations.

[0023] In some embodiments, calculating the matching degree between the storage feature vector of the goods to be received and the vectors of each row in the storage feature matrix includes quantifying and normalizing the various storage features of the goods to be received to form a standard storage feature vector. The same quantification and normalization process is then performed on the storage feature vectors of the historical stored goods represented by each row in the storage feature matrix. The cosine similarity between the standard storage feature vector of the goods to be received and the processed standard feature vectors of each row in the storage feature matrix is ​​calculated. A matching degree threshold is set, and the calculated cosine similarity is compared with the matching degree threshold. All historical stored goods with a cosine similarity greater than or equal to the matching degree threshold are selected, and the set of historical stored goods is taken as the set of historical stored goods with a matching degree higher than the preset threshold. It can be understood that cosine similarity... The calculation formula is:

[0024] in: This represents the standard warehousing feature vector of goods to be received into the warehouse. A standard warehousing feature vector representing historically stored goods. This represents the total number of warehouse feature dimensions. and Representing vectors respectively and In the Normalized values ​​across multiple warehouse feature dimensions. The matching threshold is set to a fixed value or dynamically adjusted based on warehouse operation data. For example, a higher matching threshold is set when there are many warehouse feature dimensions to improve screening accuracy, while a lower threshold is set when there are fewer warehouse feature dimensions to broaden the screening range. Data comparison shows that when the matching threshold is increased, the size of the selected historical storage goods set decreases, but the similarity of the goods in the set with the storage features of the goods to be received is higher; when the matching threshold is decreased, the size of the selected historical storage goods set increases, but it may include goods with low similarity, affecting the accuracy of subsequent recommendation confidence calculations.

[0025] In practice, after extracting the location codes of all goods previously stored in the historical storage goods set, the frequency of use and cumulative storage duration of each location by goods in the historical storage goods set are calculated. Frequency refers to the total number of times a location is used by goods in the historical storage goods set, and cumulative storage duration refers to the total length of time a location is occupied by these goods. When calculating the recommended confidence level for each location for goods to be received, a weighted summation method is used, for example, the recommended confidence level... ,in This represents the normalized frequency value. This represents the normalized cumulative storage duration value. and It is a weighting coefficient and satisfies When dynamically adjusting the recommendation confidence level, the adjustment factors include whether the storage location is currently available and the ratio of the current load capacity to the maximum load capacity. The adjustment formula is as follows: ,in It is an idle indicator factor, indicating when the storage space is idle. And when the storage space is occupied , It is the load ratio factor. , This indicates the current load-bearing capacity of the storage location. This indicates the maximum load capacity of the storage location. Optionally, the load ratio factor can be adjusted to... To prevent division by zero errors, the corrected recommended confidence level Reduce the number of occupied or overloaded storage spaces to zero to ensure that unavailable storage spaces are not recommended.

[0026] All storage locations are sorted in descending order based on dynamically adjusted recommendation confidence levels. Several locations that rank highly and meet the basic storage requirements for the goods to be stored are selected. These basic requirements include that the storage location size is not smaller than the goods' size and that the maximum load-bearing capacity of the storage location is not less than the goods' weight. In some embodiments, the top-ranked locations are selected after sorting. The selected storage locations form a preliminary pool of potential storage locations. The value is set according to the warehouse size and task requirements, for example, in a large warehouse. To provide sufficient options in a small warehouse To reduce computational load. Data comparison shows that when When the value increases, the set of alternative storage locations includes more locations, increasing the selection space for the scheduling optimization module, but the computational load increases; when When the value decreases, the set of candidate storage locations shrinks, improving computational efficiency, but potentially missing high-quality locations. It's understandable that the improved collaborative filtering recommendation algorithm optimizes storage location selection through the correlation of storage features, combined with real-time dynamic adjustments to provide a high-quality candidate set for subsequent scheduling optimization. Throughout the process, all terms are expressed in full, such as the cargo storage feature matrix and recommendation confidence level, avoiding abbreviations.

[0027] In one embodiment of the present invention, see [reference] Figure 3 The scheduling optimization module evaluates multiple candidate storage locations based on preset optimization objectives, including rack center of gravity stability, future outbound efficiency prediction, and energy consumption for aisle stacker crane collaborative operation. For the rack center of gravity stability optimization objective, it calculates the offset of the rack's center of gravity coordinates in three-dimensional space after placing the goods to be stored in each candidate storage location, and assesses whether this offset exceeds the set safety range. For the future outbound efficiency prediction optimization objective, based on historical goods storage and retrieval records, it analyzes the historical average outbound response time of the aisles and floor heights of the candidate storage locations, and combines this with the expected storage period of the goods to be stored and the outbound frequency of the owner's code to predict the future outbound operation time cost. For the aisle stacker crane collaborative operation energy consumption optimization objective, it simulates and calculates the total running distance and start / stop count of multiple stacker cranes when executing the task of storing goods in each candidate storage location, considering the current position status and task queue of multiple aisle stacker cranes. Weighting coefficients are assigned to the calculated center of gravity offset, predicted outbound time cost, total operating distance, and number of start-stop cycles, respectively. Based on these coefficients, a comprehensive evaluation score is calculated for each candidate storage location. The comprehensive evaluation scores of all storage locations belonging to the same candidate storage location set are aggregated to obtain the overall evaluation score for each candidate storage location set.

[0028] In practical implementation, the scheduling optimization module evaluates multiple candidate storage location sets based on preset optimization objectives. These objectives include rack center of gravity stability, future outbound efficiency prediction, and energy consumption for aisle stacker crane collaborative operation. For the rack center of gravity stability optimization objective, the module calculates the offset of the rack's center of gravity coordinates in three-dimensional space after placing the goods to be stored in each candidate storage location, and assesses whether the offset exceeds the set safety range. For the future outbound efficiency prediction optimization objective, based on historical goods storage and retrieval records, the module analyzes the historical average outbound response time of the aisles and floor heights of the candidate storage locations, and combines this with the expected storage period of the goods to be stored and the outbound frequency of the owner's code to predict the future outbound operation time cost. For the aisle stacker crane collaborative operation energy consumption optimization objective, the module simulates and calculates the total running distance and start / stop count of multiple stacker cranes when executing the task of storing goods in each candidate storage location, considering the current position status and task queues of multiple aisle stacker cranes. Weighting coefficients were assigned to the center of gravity offset, predicted outbound time cost, total operating distance, and number of start-stop cycles, respectively, and a comprehensive evaluation score was calculated for each candidate storage location. The comprehensive evaluation scores of all storage locations belonging to the same candidate storage location set were aggregated to obtain the overall evaluation score for each candidate storage location set.

[0029] In some embodiments, the calculation process for optimizing the center of gravity stability of the shelving involves establishing a mass distribution model of the shelving and virtually adding cargo weight information. The center of gravity offset is obtained by comparing the Euclidean distance between the overall center of gravity coordinates of the shelving before and after the virtual addition. The safe range is preset based on the structural mechanical parameters of the shelving. For example, if the calculated center of gravity offset for location A is 0.15 meters, and the maximum safe offset threshold allowed by the shelving model is 0.2 meters, then location A passes the safety assessment. If the calculated center of gravity offset for location B is 0.25 meters, exceeding the maximum safe offset threshold, then location B is marked as an unqualified option. Data comparison shows that the smaller the offset, the less impact the location has on the overall stability of the shelving, but it may not be the location with the optimal outbound efficiency or the lowest energy consumption. Therefore, it is necessary to combine it with other optimization objectives for comprehensive evaluation. The analysis process for predicting and optimizing future outbound efficiency requires querying historical data to obtain the historical average outbound response time for each aisle and floor. The historical average outbound response time refers to the average time taken to complete an outbound task from that location over a past period. Goods with longer expected storage periods tend to be stored in areas with slightly higher average outbound response times to conserve resources in high-efficiency areas. Goods with high outbound frequency according to their owner codes are preferentially assigned to locations with lower average outbound response times. For example, candidate location C is located on the 5th floor of aisle X, with a historical average outbound response time of 45 seconds. Since the expected storage period for the goods to be received is long and the outbound frequency of their owner codes is low, the predicted future outbound operation time cost may be assessed as moderate. Candidate location D is located on the 2nd floor of aisle Y, with a historical average outbound response time of 30 seconds. Since the outbound frequency of the goods to be received is high, the predicted future outbound operation time cost is even lower.

[0030] In practical implementation, the simulation calculation for optimizing the energy consumption of collaborative operation of stacker cranes in aisle areas requires obtaining the real-time coordinates and task queues of all stacker cranes. A path planning algorithm is then used to simulate the three-dimensional motion path of each candidate storage location within the navigation grid after task allocation. The total running distance is the sum of the theoretical running distances of each assigned stacker crane, and the number of start-stop cycles is the sum of the number of times each stacker crane changes its direction of movement along its respective path. For example, simulating the storage of goods in candidate storage location E requires collaborative operation of stacker cranes M and N. The theoretical running distance of stacker crane M is calculated to be 85 meters with 6 start-stop cycles, and the theoretical running distance of stacker crane N is 60 meters with 4 start-stop cycles, resulting in a total running distance of 145 meters and a total of 10 start-stop cycles. Simulating the storage of goods in candidate storage location F requires only independent operation of stacker crane P, with a calculated theoretical running distance of 120 meters and 8 start-stop cycles, resulting in a total running distance of 120 meters and a total of 8 start-stop cycles. Data comparison shows that although location E has a longer total running distance, its location may make the rack center of gravity more stable; location F has a shorter total running distance, but it may lead to a larger shift in the rack center of gravity, reflecting the trade-off between different optimization objectives.

[0031] It is understandable that calculating the comprehensive evaluation score for each candidate storage location requires normalizing indicators with different dimensions and units and then weighting and summing them to arrive at the comprehensive evaluation score. The calculation formula is:

[0032] in: This indicates the offset of the center of gravity. This indicates the predicted outbound time cost. This represents the simulated total running distance. This represents the total number of start-stop cycles in the simulation. (Function) , , , These are the normalization functions for the corresponding indicators, used to map the original indicator values ​​to a uniform scoring range, such as [0,1]. Higher values ​​indicate better performance for that indicator. Weighting coefficients. , , , It is a pre-defined non-negative constant, and satisfies The specific values ​​of the weighting coefficients reflect the degree of emphasis placed on different optimization objectives. For example, in the early stages of warehouse operation, more emphasis might be placed on equipment safety, resulting in higher weighting coefficients. The focus is on emphasizing the stability of the shelving's center of gravity; during periods of stable warehouse operation, efficiency may be prioritized, and higher values ​​may be set. and The value emphasizes outbound efficiency and operational energy consumption. Optionally, the normalization function can take the form of linear scaling or a Gaussian function, for example... ,in It is the maximum allowable center of gravity offset, when the offset When it is 0 The score is 1, when the offset is achieve hour The score is 0.

[0033] In some embodiments, the comprehensive evaluation scores of all locations belonging to the same set of candidate locations are aggregated and calculated. This aggregation calculation typically employs an averaging or maximum-value method to obtain the overall evaluation score for each set of candidate locations. For example, candidate location set G contains three locations with comprehensive evaluation scores of 0.85, 0.78, and 0.92, respectively. Using the averaging method, the overall evaluation score for set G is 0.85. Candidate location set H contains four locations with comprehensive evaluation scores of 0.70, 0.88, 0.75, and 0.90, respectively. Using the averaging method, the overall evaluation score for set H is 0.8075. The overall evaluation score of set G is higher than that of set H. Optionally, in specific scenarios, the maximum-value method can also be used, where the overall evaluation score of the set is equal to the highest comprehensive evaluation score among its locations. This method is suitable for scheduling strategies with a strong preference for a single optimal location. Data comparison shows that using the average value method reflects the overall average level of the set, which is beneficial for selecting a set of storage locations with more balanced overall quality. Using the maximum value method may select a set containing one excellent storage location but with other locations being less desirable, and the final selection of storage locations will then focus on that excellent location. It is understandable that the overall evaluation score provides a direct quantitative basis for comparison when selecting the preferred set from multiple candidate storage location sets.

[0034] In one embodiment of the present invention, when calculating the center of gravity offset of a rack, a three-dimensional geometric model and a mass distribution model of each set of racks in the automated warehouse are established. The mass distribution model records the weight of the goods currently stored in each location on the rack and the center of gravity position of the goods themselves. The weight of the goods to be received and the preset center of gravity mark position on their packaging are obtained. For each candidate location in the set of alternative locations, the weight and center of gravity position information of the goods to be received are virtually added to the mass distribution model of the rack to which the candidate location belongs. The center of gravity coordinates of the entire rack in the three-dimensional coordinate system after the virtual addition are recalculated and compared with the design center of gravity coordinates of the rack in the unloaded state. The Euclidean distance between the virtual added center of gravity coordinates and the design center of gravity coordinates is calculated as the center of gravity offset. The center of gravity offsets corresponding to all candidate locations are compared with the maximum safe offset threshold allowed by the rack model, and candidate locations whose offsets exceed the safe threshold are marked.

[0035] When simulating the energy consumption of stacker cranes operating collaboratively in aisles, the real-time 3D coordinates of all stacker cranes in the automated warehouse and their current task queue information are obtained. For each candidate storage location, based on a preset stacker crane task allocation strategy, a stacker crane is simulated to be allocated from currently idle or soon-to-be-idle stacker cranes to transport goods from the picking point to the candidate storage location. When simulating task allocation, the expected position and state of the stacker crane after completing all tasks in its existing task queue are considered. Based on the expected starting position of the stacker crane to be assigned the task and the position of the candidate storage location, a 3D motion path from the starting point to the final storage location is planned in the navigation grid of the automated warehouse. The geometric length of this planned path in 3D space is calculated as the theoretical running distance for the stacker crane to execute this task. Simultaneously, based on the number of changes in the stacker crane's movement direction in the horizontal, vertical, and fork extension directions along the planned path, the number of starts and stops of the stacker crane on the path is simulated. Each state transition from stationary to moving or from moving to stationary is counted as one start and stop. For all stacker cranes that are simulated to be assigned to the warehousing task, their theoretical running distance and number of start-stops are summed to obtain the simulated total running distance and simulated total number of start-stops of multiple stacker cranes when storing goods in the alternative storage location.

[0036] In practical implementation, calculating the center of gravity offset requires establishing a 3D geometric model and a mass distribution model for each set of shelves in the automated warehouse. The mass distribution model records the weight of the goods currently stored in each location on the shelf and the center of gravity position of the goods themselves, obtaining the weight of the goods to be received and the preset center of gravity mark position on their packaging. For each candidate location in the set of alternative locations, the weight and center of gravity position information of the goods to be received are virtually added to the mass distribution model of the shelf to which the candidate location belongs. The center of gravity coordinates of the entire shelf in the 3D coordinate system after the virtual addition are recalculated and compared with the design center of gravity coordinates of the shelf in the unloaded state. The Euclidean distance between the virtual added center of gravity coordinates and the design center of gravity coordinates is calculated as the center of gravity offset. The calculated center of gravity offsets corresponding to all candidate locations are compared with the maximum safe offset threshold allowed by the shelf model, and candidate locations with offsets exceeding the safe threshold are marked. When calculating the center of gravity offset, the center of gravity coordinates of the shelf after the virtual addition are... The centroid coordinates are designed by calculating the centroid of all goods on the shelf (including stored goods and virtually added goods awaiting warehousing). It is the theoretical geometric center of the shelving when it is unloaded and structurally uniform, and the offset of the center of gravity. The calculation formula is:

[0037] in: and They are all three-dimensional coordinate vectors. This represents the Euclidean length (modulus) of the calculated vector. The maximum allowable safe offset threshold for the shelf model is a preset scalar value. When the calculated center of gravity offset Greater than In such cases, the corresponding candidate storage location is marked as an unstable storage location and will not participate in subsequent comprehensive evaluations, or its evaluation score will be reduced accordingly. For example, the design center of gravity coordinates of a set of shelving units. The dimensions are (2.5, 1.8, 6.0) meters. The weight of the goods to be stored is 200 kg. The preset center of gravity marker position relative to the shelf coordinate system is (1.2, 0.5, 3.1) meters. After this goods are virtually added to a candidate storage location, the new overall center of gravity coordinates are calculated based on the existing goods distribution on that shelf. If the distance is (2.7, 1.9, 6.2) meters, then the center of gravity offset is... Meters. If the maximum safe offset threshold for this type of shelving... If the center of gravity offset is 0.4 meters, the storage location passes the stability check; if the center of gravity offset of another storage location is calculated to be 0.4 meters, then the storage location is marked. See Table 1 for exemplary data on center of gravity offset calculation and safety comparison for a set of four candidate storage locations.

[0038] Table 1: Comparison of Calculation and Safety of Candidate Cargo Location Center of Gravity Offset

[0039] In some embodiments, when calculating the energy consumption of the simulated aisle stacker crane collaborative operation, it is necessary to obtain the real-time three-dimensional coordinate positions of all aisle stacker cranes in the current automated warehouse and their current task queue information. For each candidate storage location, based on a preset stacker crane task allocation strategy, it is simulated that a stacker crane is allocated from currently idle or soon-to-be-idle stacker cranes to the task of transporting goods from the picking point to the candidate storage location. When simulating task allocation, the expected position and state of the stacker crane after completing all tasks in its existing task queue are considered. Based on the expected starting position of the stacker crane to be assigned the task and the position of the candidate storage location, a three-dimensional motion path from the starting point to the final storage location is planned in the navigation grid of the automated warehouse. The geometric length of the planned path in three-dimensional space is calculated as the theoretical running distance of the stacker crane to perform this task. At the same time, based on the number of times the stacker crane changes its movement direction in the horizontal, vertical, and fork extension directions in the planned path, the number of starts and stops of the stacker crane on the path is simulated and calculated. Each state transition from stationary to moving or from moving to stationary is counted as one start and stop. For all stacker cranes simulated and assigned to inbound tasks, their theoretical running distance and number of starts and stops are summed to obtain the simulated total running distance and simulated total number of starts and stops for multiple stacker cranes when storing goods into alternative storage locations. For example, simulating storing goods into alternative storage location A-10-05, the system detects that stacker crane #1 is currently performing a task and is expected to arrive at location P1 (empty) in 10 seconds. Stacker crane #2 is currently idle at location P2. The preset allocation strategy is to select the equipment that is expected to start the task earliest, so the task is simulated and assigned to stacker crane #1. From the expected starting position P1 of stacker crane #1 to the picking point, and then to the final storage location A-10-05, the planned three-dimensional motion path has a geometric length of 125 meters. Path analysis shows that the stacker crane turns 3 times in the horizontal direction, starts and stops 2 times in the vertical direction, and extends and retracts the forks 1 time, for a total of 6 simulated starts and stops. If this task requires another stacker crane to assist in pallet handling, assuming stacker crane #3 is assigned, its theoretical operating distance is 80 meters and the number of start-stop operations is 4, then the simulated total operating distance for storing the pallet at location A-10-05 is 205 meters, and the simulated total number of start-stop operations is 10. Data comparison shows that different alternative locations, due to their different spatial positions, result in different stacker crane combinations, planned path lengths, and complexities in the simulation, thus producing varying simulated total operating distances and simulated total number of start-stop operations.

[0040] In practice, the simulation calculation process is static and forward-looking, based on snapshot information at the current moment, and does not actually drive the equipment to move. It can be understood that the theoretical running distance is the sum of the geometric lengths of the paths, reflecting the cost of spatial displacement; the number of start-stop cycles is an estimate of the frequency of equipment acceleration and deceleration, reflecting the cost of action switching. Both together constitute an indirect quantification of the energy consumption of collaborative operation. Optionally, in addition to the "earliest available" strategy, the preset stacker crane task allocation strategy can also include a "shortest path" strategy, that is, selecting the stacker crane with the shortest path from the current location to the task pickup point during simulation allocation, without considering the completion time of its current task queue. Different allocation strategies will lead to different simulation results. For example, for the same alternative storage location, using the "earliest available" strategy might allocate stacker crane #1 (long path but quick availability), with a larger simulated total running distance; using the "shortest path" strategy might allocate stacker crane #2 (shorter path but requires waiting), with a smaller simulated total running distance. In some embodiments, the simulation only considers the stacker crane primarily performing handling tasks. Alternatively, the avoidance paths of stacker cranes that may be affected and are engaged in other operations can also be included in the calculation of the total running distance and number of starts and stops to more comprehensively assess the collaborative impact. Optionally, the node spacing of the navigation grid used when planning the three-dimensional motion path is fixed, for example, one node every 1 meter in the horizontal direction and one node every 0.5 meters in the vertical direction. The theoretical running distance is the sum of the straight-line distances of each path segment. When calculating the number of changes in motion direction, only when the change in direction angle between path segments exceeds a set threshold (e.g., 5 degrees) is it counted as a change in direction to prevent minor vibrations from being mistakenly counted as starts and stops. It can be understood that the simulated total running distance and the simulated total number of starts and stops provide quantitative inputs regarding energy consumption and efficiency for the subsequent comprehensive evaluation score calculation.

[0041] In one embodiment of the present invention, the scheduling optimization module compares the overall evaluation scores of multiple candidate storage location sets and selects the candidate storage location set with the highest overall evaluation score as the preferred set. Within the preferred set, the final storage location is selected according to preset urgency rules. These urgency rules include selecting the storage location with the shortest path when the inbound task is urgent, or selecting the storage location with the most stable center of gravity when the task is stable. After determining the final storage location, based on the real-time location, status, and task queue of all current aisle stacker cranes, a suitable aisle stacker crane is assigned to the handling task of the goods to be inbound.

[0042] The system plans a three-dimensional movement path for assigned stacker cranes from the pickup point, through the transfer point, to the final storage location. During planning, the three-dimensional space of the automated warehouse is discretized into a navigation mesh composed of nodes located at the storage location center, aisle intersections, and equipment docking points. An improved A* search algorithm considering dynamic obstacles is used to search the navigation mesh. Dynamic obstacles include other stacker cranes performing tasks and the space occupied by their predetermined paths. During the search, different cost weights are assigned to the horizontal movement, vertical lifting, and fork extension / retraction of the stacker crane to reflect the energy consumption and time differences of different axial movements. After a preliminary path is found, it is smoothed and optimized, transforming it into a smooth trajectory of continuous velocity and acceleration of each drive axis of the stacker crane. Collision detection is performed on the smoothed trajectory to ensure that at any given time, the stacker crane performing the task maintains a safe distance from other stacker cranes and fixed facilities. Integrate the movement paths, sequences, and time nodes of all stacker cranes participating in the collaborative operation to generate a detailed inbound execution plan. This plan includes the action command sequence, speed curve, and expected start and end times of each stacker crane.

[0043] In practical implementation, the scheduling optimization module selects the final storage location from multiple candidate storage location sets based on the evaluation results and generates a detailed inbound execution plan that includes the collaborative operation paths of multiple aisle stacker cranes. The overall evaluation scores of multiple candidate storage location sets are compared, and the set with the highest overall evaluation score is selected as the preferred set. Within the preferred set, the final storage location is selected according to preset emergency rules. These rules include selecting the storage location with the shortest path when the inbound task is urgent, or selecting the storage location with the most stable center of gravity when the task is stable. After determining the final storage location, based on the real-time location, status, and task queue of all aisle stacker cranes, suitable aisle stacker cranes are assigned to the handling tasks of the goods to be inbound. A three-dimensional motion path is planned for the assigned aisle stacker crane from the pickup point through the transfer point to the final storage location. Path planning must consider the kinematic constraints of the stacker crane and avoid conflicts with other stacker crane paths. Integrate the motion paths, sequences, and time nodes of all stacker cranes participating in the collaborative operation to generate a detailed inbound execution plan. The plan includes the motion command sequence, speed curve, and expected task start and end time for each stacker crane. See Table 2.

[0044] Table 2: Location Selection Table within the Preferred Location Set

[0045] In some embodiments, the selection of the final storage location relies on preset urgency rules. The urgency of the task is assigned by the warehouse management system when the task is issued or dynamically determined by the system based on the task queue load. For example, when an inbound task is marked as "high" urgency, the system applies the "shortest path" rule, comparing the simulated total running distance of each storage location in the preferred set Alpha. Storage location D-08-10, at 195 meters, has the shortest path and is therefore selected as the final storage location. When an inbound task is marked as "low" urgency, the system applies the "most stable center of gravity" rule, comparing the center offset of each storage location in the preferred set Beta. Storage location E-12-03, at 0.15 meters, has the smallest center of gravity offset and is therefore selected as the final storage location. Data comparison shows that the application of urgency rules allows the storage location selection strategy to adapt to different operational pressure scenarios, prioritizing equipment operating efficiency when rapid inbound operations are urgently needed, and prioritizing the long-term stability of warehousing facilities during routine operations.

[0046] In practical implementation, planning the three-dimensional movement path of the assigned stacker cranes from the pickup point through the transfer point to the final storage location requires discretizing the three-dimensional space of the automated warehouse into a navigation grid composed of nodes. Nodes are located at the storage location center, aisle intersections, and equipment docking points. An improved A-search algorithm considering dynamic obstacles is used to search for paths within the navigation grid. Dynamic obstacles include other stacker cranes performing tasks and the space occupied by their predetermined paths. During the search process, different cost weights are assigned to the horizontal movement, vertical lifting, and fork extension / retraction of the stacker crane to reflect the energy consumption and time differences of different axial movements. After a preliminary path is found, it is smoothed and optimized, converting it into a smooth motion trajectory with continuous velocity and acceleration of each drive axis of the stacker crane. Collision detection is performed on the smoothed trajectory to ensure that at any given time, the stacker crane performing the task maintains a safe distance from other stacker cranes and fixed facilities. It can be understood that the movement cost from the current node to an adjacent node in the improved A-search algorithm... The calculation formula is:

[0047] in: This represents the Euclidean distance of horizontal movement. It is the cost weight of horizontal movement; This represents the Euclidean distance for vertical rise and fall. It is the cost weight for vertical rise and fall; This represents the Euclidean distance of the fork extension / retraction. This refers to the cost weighting of fork extension and retraction. Since the maximum speed and acceleration differ along different axes of the stacker crane—for example, horizontal movement is usually the fastest—the cost weighting... It can be set to 1.0; vertical rise and fall are relatively slow, and the cost weight is relatively high. It can be set to 1.5; the fork extension and retraction is the slowest and energy consumption is the highest, so the cost weighting is lower. It can be set to 2.0, which will make the algorithm prioritize horizontal movement and reduce vertical and forklift movement during pathfinding. During the search, grid nodes occupied by dynamic obstacles will be temporarily marked as impassable, and additional penalty costs will be imposed on paths passing through their neighboring areas to encourage detours.

[0048] In some embodiments, path smoothing optimization is achieved through spline curve interpolation algorithms, transforming the polygonal path formed by connecting navigation grid nodes into a smooth curve. The smoothed motion trajectory must satisfy the maximum speed and acceleration constraints of each drive axis of the stacker crane. Collision detection is based on the space-time cube occupied by the equipment and the predetermined path. The bounding box spatial position of each stacker crane at each moment is calculated, and it is checked whether there is any intersection between them and between them and the fixed rack. The safety distance is usually set to 0.5 meters. For example, to plan a path for stacker crane #1 from the picking point P1(0,0,0) to the transfer point M1(10,2,1), and then to the final storage location F1(15,5,3), the improved A* algorithm, after considering dynamic obstacles (such as the area (12-14,3-4,2-3) that stacker crane #2 will occupy from t=5 seconds to t=10 seconds), may plan a path that bypasses the area. The initial path is P1->(8,1,0)->M1->(13,4,2)->F1. After smoothing the path, a continuous trajectory is generated. Collision detection verifies that at t=8 seconds, the expected position of stacker crane #1 and the predetermined position of stacker crane #2 meet the safe distance requirement. Optionally, transfer points are usually selected as aisle intersections or specific equipment buffer positions to change the direction of movement or coordinate the intersection of multiple equipment. It can be understood that when generating a detailed inbound execution plan, the sequence of action instructions includes steps such as "move to the picking point", "extend forks to pick up", "lift the load", "move horizontally to the transfer point", "adjust vertically", "move horizontally to the target column", "extend forks to release", and "retract forks to return". The speed curve specifies the curve of the speed of each drive shaft changing with time in each step, and the expected task start and end times are timestamps calculated based on the path length, speed curve, and equipment acceleration and deceleration capabilities.

[0049] In one embodiment of the invention, the planning and execution module decomposes the detailed inbound execution plan into independent equipment control commands and sends these commands sequentially to the corresponding aisle stacker crane controller and conveyor system controller. It receives real-time feedback from the aisle stacker crane controller regarding motor current, encoder position, and photoelectric sensor status information, as well as speed feedback and goods presence detection information from the conveyor system. The module compares the feedback equipment status information with the expected status in the inbound execution plan in real time to detect any positional deviations, timeouts, or equipment malfunctions. When the aisle stacker crane completes a goods storage / retrieval operation and leaves the storage location, it immediately updates the occupancy status of the corresponding storage location on the automated warehouse electronic map, changing it from "occupied" to "idle," or vice versa, and records the goods code. If equipment malfunctions or severe task execution delays are detected during real-time monitoring, a dynamic rescheduling process is triggered.

[0050] When the dynamic rescheduling process is triggered, immediately suspend the current tasks of the affected stacker cranes or conveyor sections in the aisles and mark their status as "faulty" or "unavailable." Assess the impact of the fault on the currently executing inbound execution plan and subsequent queued tasks, identifying all tasks that are blocked or cannot be executed as originally planned. Reacquire the status of currently available equipment resources and the real-time occupancy status of all storage locations. For the identified affected tasks, based on the reacquired resource and status information and considering the estimated recovery time of the faulty equipment, reallocate storage locations, equipment, and planned routes to them. Generate a new local execution plan and coordinate with other normal equipment to work in conjunction with the new plan, issuing control commands to bypass the fault point or resume operations after waiting.

[0051] In practical implementation, the planning and execution module schedules the stacker cranes and conveyor belt systems to execute the inbound execution plan. During execution, it monitors the equipment operating status and changes in storage location occupancy in real time. The detailed inbound execution plan is broken down into independent equipment control commands, which are then sent to the corresponding stacker crane controllers and conveyor belt system controllers in chronological order. The module receives real-time feedback from the stacker crane controllers on motor current, encoder position, and photoelectric sensor status information, as well as speed feedback and goods presence detection information from the conveyor belt system. This feedback is compared in real-time with the expected status in the inbound execution plan to detect any positional deviations, timeouts, or equipment malfunctions. When the stacker crane completes the goods storage / retrieval action and leaves the storage location, the occupancy status of the corresponding storage location in the automated warehouse electronic map is immediately updated, changing from "occupied" to "idle," or vice versa, and the goods code is recorded. If equipment malfunctions or severe task execution delays are detected during real-time monitoring, a dynamic rescheduling process is triggered.

[0052] In some embodiments, decomposing a detailed inbound execution plan into individual device control instructions involves translating a high-level sequence of actions into instructions recognizable by the low-level controller. For example, "move to pickup point" in an action instruction sequence is decomposed into target position coordinates and velocity curve parameters sent to the stacker crane drive. The generated control instructions are encapsulated and sent to the device according to predetermined timestamps. When comparing feedback information with the expected state in real time, the position deviation is calculated by comparing the actual position fed back by the encoder with the planned expected position. The calculation formula is:

[0053] in: This represents the actual three-dimensional coordinates fed back by the encoder of the stacker crane in the tunnel. This represents the expected three-dimensional coordinates at the current time point in the inbound execution plan. When the calculated position deviation... When the position deviation exceeds the allowable threshold (e.g., 0.05 meters), the system records a position deviation event. Continuous position deviations or excessively large deviations may trigger an alarm. Time monitoring is performed by comparing the expected completion timestamp of the command with the timestamp of the "action completed" signal fed back by the equipment. If the actual completion time is later than the expected completion time by more than a set tolerance (e.g., 5 seconds), it is recorded as a timeout event. Equipment malfunctions are determined by analyzing abnormal motor current, persistent abnormal sensor signals, or error codes reported by the controller. For example, if the stacker crane fork motor current continuously exceeds 150% of the rated value for more than 1 second, the system determines it as a motor overload fault.

[0054] In practice, triggering the dynamic rescheduling process involves immediately suspending the current tasks of the affected stacker cranes or conveyor belt segments and marking their status as "faulty" or "unavailable." The impact of the fault on the currently executing inbound execution plan and subsequent queued tasks is assessed, identifying all blocked or unexecuted tasks. The system reacquires the status of currently available equipment resources and the real-time occupancy status of all storage locations. For the identified affected tasks, based on the reacquired resource and status information and considering the estimated recovery time of the faulty equipment, storage locations, equipment, and planned paths are reassigned. A new local execution plan is generated, and other normal equipment is coordinated with the new plan. Control commands are issued to bypass the fault point or resume operation after waiting. For example, if stacker crane #3 is marked as "faulty" due to a photoelectric sensor malfunction during task execution, its current task "transporting pallets to storage location C-07-08" is immediately suspended. The system assessment finds that this fault blocks two subsequent outbound tasks in the current inbound execution plan that require storage location C-07-08 to be vacated, identifying these three tasks as affected tasks. The system reacquires the resource status, confirms that stacker cranes #1 and #2 are in normal condition and that storage location C-07-08 is still in an "occupied" state. Considering that the failure of stacker crane #3 is expected to take 15 minutes to repair, the system decides to reallocate a stacker crane and storage location to the suspended inbound task, assuming it is stacker crane #1 and storage location D-10-05. The system also replans the path for the two blocked outbound tasks to retrieve goods from other available storage locations, generates a new local execution plan involving stacker cranes #1 and #2, and issues it.

[0055] It is understandable that updating the occupancy status of storage locations on the automated warehouse's electronic map is a critical state synchronization operation. This must be performed immediately after the physical storage / retrieval action is confirmed to ensure that the environmental perception module provides accurate real-time occupancy status of storage locations across the entire warehouse area for subsequent tasks. In some embodiments, the completion of a goods storage / retrieval action is confirmed by both the "forks retracted and away from storage location" signal from the stacker crane controller and the "storage location status change" signal from the rack photoelectric sensors. Only when both signals meet the conditions does the system determine that the goods storage / retrieval is complete and trigger a state update. Data comparison shows that delays in state updates may lead to subsequent task planning based on incorrect information. For example, if a storage location is actually vacant but the map still shows it as occupied, the system will not assign new tasks to that location, reducing warehouse utilization. Conversely, if a storage location is actually still occupied but the map is incorrectly updated to vacant, subsequent tasks may attempt to store goods in an already occupied location, causing conflicts or accidents. Optionally, for conveyor belt systems, status monitoring is primarily achieved through speed feedback and cargo presence detection information. When the speed feedback of a certain section of the conveyor belt falls below a set value or the cargo presence detection signal abnormally times out, the system determines that the conveyor belt section is abnormal and may trigger dynamic rescheduling for tasks that depend on that section. The algorithm for replanning the path in the dynamic rescheduling process is similar to the initial planning, but the search space is limited by currently occupied cargo locations and available equipment. The generated new local execution plan needs to be time-coordinated with the unaffected subsequent parts of the original plan. It can be understood that the planning execution module and the dynamic rescheduling process together constitute a closed-loop control for inbound operations, ensuring that the system can maintain continuous operation in both cases of accurate plan execution and handling of unexpected anomalies.

[0056] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A multi-axis linkage intelligent racking system for automated warehousing with dynamic allocation and scheduling of storage locations, characterized in that: The system includes: The task parsing module receives and parses inbound task orders, outbound task orders, and inventory count instructions issued by the warehouse management system in real time. The inbound task order contains the physical attributes and storage requirements of the goods to be received, and the outbound task order contains the location code and demand priority of the goods to be shipped. The environmental perception module, based on the electronic map of the automated warehouse and the shelf status sensors, dynamically perceives the real-time occupancy status, load-bearing capacity, and cargo storage and retrieval history of all storage locations in the entire warehouse area. The storage location recommendation module generates multiple candidate storage location sets for the goods to be stored based on the physical attributes and storage requirements of the goods to be stored, and in combination with the real-time occupancy status and load-bearing capacity of storage locations throughout the warehouse area, using an improved collaborative filtering recommendation algorithm. The improved collaborative filtering recommendation algorithm is optimized based on the correlation of the storage characteristics of the goods. The scheduling optimization module evaluates multiple candidate storage locations based on preset optimization goals, selects the final storage location from the multiple candidate storage locations based on the evaluation results, and generates a detailed inbound execution plan that includes the collaborative operation path of multiple aisle stacker cranes. The planning and execution module schedules the stacker cranes and conveyor belt systems to execute the warehousing execution plan, and monitors the equipment operating status and changes in storage space occupancy in real time during the execution process, and updates the real-time storage space occupancy status. The improved collaborative filtering recommendation algorithm is optimized based on the correlation of goods storage features, and its working principle includes: Construct a cargo storage feature matrix. The rows of the matrix represent the historical stored goods in the warehouse, and the columns of the matrix represent the storage feature dimensions of the goods. The storage feature dimensions include the size, weight, storage temperature requirements, moisture protection level, owner code, and expected storage period of the goods. Calculate the matching degree between the storage feature vector of the goods to be put into storage and each row vector in the storage feature matrix of the goods, and filter out the set of historical stored goods with a matching degree higher than a preset threshold. Extract the location codes of all goods that have been stored in the historical storage goods set, and count the frequency of use and cumulative storage duration of each location by goods in the historical storage goods set; Based on the frequency of use of storage locations by historical stored goods and the cumulative storage duration, calculate the recommended confidence level for each storage location for goods to be received. The recommendation confidence level is dynamically adjusted based on the real-time occupancy status and load capacity of the current storage location. The adjustment factors include whether the storage location is currently vacant and the ratio of the current load capacity to the maximum load capacity. All storage locations are sorted in descending order based on the dynamically adjusted recommendation confidence level. Several storage locations that rank highly and meet the basic storage requirements of the goods to be stored are selected to form a preliminary set of candidate storage locations. The process of calculating the matching degree between the storage feature vector of the goods to be stored and each row vector in the goods storage feature matrix, and filtering out a set of historically stored goods with a matching degree higher than a preset threshold, includes: The various warehousing characteristics of the goods to be received are quantified and normalized to form a standard warehousing characteristic vector of the goods to be received. The same numerical and normalization processing is applied to the storage feature vectors of historical stored goods represented by each row in the cargo storage feature matrix. Calculate the cosine similarity between the standard warehousing feature vector of the goods to be put into storage and the processed standard feature vector of each row in the goods warehousing feature matrix; Set a matching threshold and compare the calculated cosine similarity with the matching threshold; Filter out all historical stored goods whose cosine similarity is greater than or equal to the matching degree threshold, and use the set of historical stored goods as the set of historical stored goods with a matching degree higher than the preset threshold.

2. The automated storage and retrieval system for multi-axis linkage intelligent shelving with dynamic allocation and scheduling of storage locations as described in claim 1, is characterized in that, The evaluation of multiple candidate storage locations based on preset optimization objectives includes: The optimization objectives include stabilizing the rack center of gravity, predicting future outbound efficiency, and optimizing the energy consumption of stacker cranes operating in aisles. To optimize the stability of the rack center of gravity, the offset of the rack center of gravity coordinate in three-dimensional space is calculated after the goods to be stored are placed in each of the candidate storage locations, and it is evaluated whether the offset exceeds the set safety range. To optimize future outbound efficiency, based on historical cargo storage and retrieval records, the historical average outbound response time of the aisles and floors of the candidate storage locations is analyzed. Combined with the expected storage period of the goods to be stored and the outbound frequency of the owner's code, the future outbound operation time cost is predicted. To optimize the energy consumption of collaborative operation of stacker cranes in aisle areas, this study simulates and calculates the total running distance and number of start-stop operations of multiple stacker cranes when performing the task of storing goods in various alternative storage locations, taking into account the current position status and task queue of multiple stacker cranes in aisle areas. Weighting coefficients were assigned to the center of gravity offset, predicted outbound time cost, total running distance, and number of start-stop cycles, and a comprehensive evaluation score was calculated for each candidate storage location. The overall evaluation scores of all storage locations belonging to the same set of candidate storage locations are aggregated and calculated to obtain the overall evaluation score of each set of candidate storage locations.

3. The three-dimensional warehouse multi-axis linkage intelligent racking location dynamic allocation and scheduling system according to claim 2, characterized in that, The optimization objective for stabilizing the shelf's center of gravity involves calculating the offset of the shelf's center of gravity coordinates in three-dimensional space after placing the goods to be received into each of the candidate storage locations. This includes: Establish a three-dimensional geometric model and a mass distribution model for each set of shelves in the automated warehouse. The mass distribution model records the weight of the goods currently stored in each location on the shelf and the position of the center of gravity of the goods themselves. Obtain the weight of the goods to be put into storage and the position of the preset center of gravity mark on the packaging box; For each candidate storage location in the set of alternative storage locations, the weight and center of gravity information of the goods to be stored are virtually added to the mass distribution model of the shelf to which the candidate storage location belongs. Recalculate the center of gravity coordinates of the entire rack in the three-dimensional coordinate system after virtual addition, and compare them with the design center of gravity coordinates of the rack in the unloaded state. Calculate the Euclidean distance between the virtual added center of gravity coordinates and the design center of gravity coordinates as the center of gravity offset. The calculated center of gravity offset of all candidate storage locations is compared with the maximum safe offset threshold allowed by the rack model, and candidate storage locations whose offset exceeds the safe threshold are marked.

4. The three-dimensional warehouse multi-axis linkage intelligent racking location dynamic allocation and scheduling system according to claim 3, characterized in that, The energy consumption optimization target for the coordinated operation of stacker cranes in aisle areas is calculated by simulating the total running distance and number of start-stop operations of multiple stacker cranes when performing the task of storing goods into various alternative storage locations, taking into account the current position status and task queue of multiple stacker cranes. This includes: Obtain the real-time three-dimensional coordinates of all stacker cranes in the current automated warehouse and their current task queue information; For each alternative storage location, based on the preset stacker crane task allocation strategy, simulate the allocation of a stacker crane from currently idle or soon-to-be-idle stacker cranes for the task of transporting goods from the pickup point to the alternative storage location. When simulating task allocation, consider the expected position and status of the stacker crane after it has completed all tasks in its existing task queue; Based on the expected starting position of the stacker crane assigned the task in the simulation and the position of the alternative storage location, a three-dimensional movement path from the starting point to the final storage location is planned in the navigation grid of the automated warehouse. Calculate the geometric length of the planned path in three-dimensional space, which is the theoretical running distance for the stacker crane to perform this task; Meanwhile, based on the number of times the stacker crane changes its direction of movement in the horizontal, vertical, and fork extension directions in the planned path, the number of times the stacker crane starts and stops on the path is simulated and calculated. Each state transition from stationary to moving or from moving to stationary is counted as one start and stop. For all stacker cranes that are simulated to be assigned to the warehousing task, their theoretical running distance and number of start-stops are summed to obtain the simulated total running distance and simulated total number of start-stops of multiple stacker cranes when storing goods in the alternative storage location.

5. The three-dimensional warehouse multi-axis linkage intelligent racking location dynamic allocation and scheduling system according to claim 4, characterized in that, The final storage location is selected from multiple candidate storage locations based on the evaluation results, and a detailed inbound execution plan containing the collaborative operation paths of multiple aisle stacker cranes is generated, including: Compare the overall evaluation scores of multiple alternative storage location sets, and select the alternative storage location set with the highest overall evaluation score as the preferred set; In the preferred set, the final storage location is selected according to the preset emergency rules. The emergency rules include selecting the storage location with the shortest path when the inbound task is urgent, or selecting the storage location with the most stable center of gravity when the task is stable. After determining the final storage location, based on the real-time location, status, and task queue of all current aisle stacker cranes, assign appropriate aisle stacker cranes to the handling tasks of goods to be stored. To plan the three-dimensional motion path of the assigned stacker crane from the picking point through the transfer point to the final storage location, the path planning needs to consider the kinematic constraints of the stacker crane and avoid conflicts with the paths of other stacker cranes. Integrate the motion paths, sequences, and time nodes of all stacker cranes participating in the collaborative operation to generate a detailed inbound execution plan. The plan includes the motion command sequence, speed curve, and expected start and end times of each stacker crane.

6. The three-dimensional warehouse multi-axis linkage intelligent racking location dynamic allocation and scheduling system according to claim 5, characterized in that, The three-dimensional movement path planned for the allocated stacker crane from the pickup point through the transfer point to the final storage location includes: The three-dimensional space of the automated warehouse is discretized into a navigation grid composed of nodes, which are located at the center of the storage location, the intersection of the aisle, and the docking point of the equipment. The improved A* search algorithm, which takes into account dynamic obstacles, is used to search the navigation grid. Dynamic obstacles include the space occupied by other stacker cranes performing tasks and their predetermined paths. During the search process, different cost weights are assigned to the horizontal movement, vertical lifting, and fork extension of the stacker crane to reflect the differences in energy consumption and time for different axial movements. Once a preliminary path is found, it is smoothed and optimized to convert it into a smooth motion trajectory with continuous speed and acceleration of each drive shaft of the stacker crane. Collision detection is performed on the smoothed motion trajectory to ensure that at any point in time, the stacker crane performing the task maintains a safe distance from other stacker cranes and fixed facilities.

7. The three-dimensional warehouse multi-axis linkage intelligent racking location dynamic allocation and scheduling system according to claim 6, characterized in that, The scheduling aisle stacker crane and conveyor belt system execute the warehousing plan and monitor the equipment operating status and changes in storage space occupancy in real time during the execution process, including: The detailed inbound execution plan is broken down into independent equipment control instructions, and these instructions are issued to the corresponding stacker crane controllers and conveyor system controllers in chronological order. It receives real-time feedback from the stacker crane controller on motor current, encoder position, photoelectric sensor status information, as well as speed feedback and cargo presence detection information from the conveyor belt system. The feedback equipment status information is compared with the expected status in the inbound execution plan in real time to detect whether there is a positional deviation, timeout or equipment failure. When the stacker crane in the aisle completes the goods storage and retrieval operation and leaves the storage location, the occupancy status of the corresponding storage location in the automated warehouse electronic map is immediately updated, changing from "occupied" to "idle", or from "idle" to "occupied", and the goods code is recorded. If a device malfunction or a serious delay in task execution is detected during real-time monitoring, a dynamic rescheduling process will be triggered.

8. The three-dimensional warehouse multi-axis linkage intelligent racking location dynamic allocation and scheduling system according to claim 7, characterized in that, The process of triggering dynamic rescheduling includes: Immediately suspend the current task of the affected stacker crane or conveyor section in the aisle and mark its status as "faulty" or "unavailable"; Assess the impact of the failure on the currently executing inbound execution plan and subsequent queued tasks, and identify all tasks that are blocked or cannot be executed as originally planned; Reacquire the status of currently available equipment resources and the real-time occupancy status of all storage locations; For the identified affected tasks, based on the reacquired resource and status information, and taking into account the estimated recovery time of the faulty equipment, the storage locations, equipment, and planned routes are reassigned to them. Generate a new local execution plan and coordinate with other normal equipment to work together with the new plan. Issue control commands to bypass the fault point or resume operation after waiting.

Images