Logistics storage task distribution system based on AI

Abstract

The invention relates to the technical field of intelligent scheduling, in particular to a logistics storage task distribution system based on AI. According to the method, through fine calculation of energy consumption data of the equipment in the storage area and dynamic adjustment of the path energy consumption gradient, the material carrying path is optimized, the use frequency of the high-energy-consumption path is reduced, and the energy efficiency ratio of equipment operation is increased. In combination with density analysis of task execution time, task execution conditions in different time periods are monitored, the concentration degree of tasks in peak periods is identified, potential congestion risks are predicted based on trend analysis of task density, a task execution window is reasonably adjusted, and operation delay caused by excessive concentration of tasks is reduced. And analyzing the execution sequence of the storage tasks, and adjusting the tasks with close dependency relationships so that the tasks can be executed at reasonable time points. In the dynamic task allocation process, simulation optimization is carried out by using a reinforcement learning method, so that task allocation is more intelligent, and the overall operation cooperation capability of the warehouse is improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent scheduling technology, and in particular to an AI-based logistics warehousing task allocation system. Background Art

[0002] The field of intelligent scheduling technology covers multiple industries such as logistics, manufacturing, and traffic management, and aims to improve operational efficiency through algorithm optimization and resource scheduling. The core content of this technology includes warehousing task allocation, route optimization, and resource coordination. In the field of logistics and warehousing, intelligent scheduling is mainly used in cargo storage planning, order picking route optimization, and transportation vehicle scheduling.

[0003] Among them, the logistics warehousing task allocation system refers to a system that automatically allocates warehousing tasks such as material handling, picking, replenishment, and shelving according to factors such as the storage location of goods, warehousing task priority, available manpower and equipment status during warehousing operations. The system allocates various warehousing operations through warehousing task scheduling rules, determines the best warehousing task executors based on warehouse layout, warehousing task urgency, and the current status of operators, and matches appropriate handling tools or transportation equipment based on the characteristics of warehousing tasks.

[0004] In the process of warehousing operations, path selection mainly relies on fixed rules or simple shortest path algorithms, which fail to fully consider the dynamic changes of equipment energy consumption, resulting in long-term high-load operation of some paths, increasing energy consumption and equipment loss. For the management of task execution time, there is a lack of real-time monitoring of task density, making it difficult to identify the peak period of operation in a timely manner. Tasks are easily concentrated in a specific time period, affecting the balance of the overall warehousing operation. The processing of task dependencies is relatively fixed, and it fails to make flexible adjustments based on the actual execution needs between tasks, resulting in some tasks being delayed while waiting for dependent tasks to be completed, reducing the overall smoothness of task execution. The task allocation method is relatively static, usually allocated according to preset rules, and it is difficult to make dynamic adjustments based on real-time operation data, resulting in some tasks failing to match the optimal execution object, affecting the overall efficiency of warehousing operations. In addition, the decision-making process of task execution is relatively simple, lacking intelligent optimization methods based on historical task data and real-time warehousing status, making it difficult for task allocation to reach the optimal state, resulting in reduced utilization of operation resources. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and propose an AI-based logistics warehousing task allocation system.

[0006] In order to achieve the above objectives, the present invention adopts the following technical solutions: The AI-based logistics warehousing task allocation system includes:

[0007] The energy consumption gradient path avoidance module obtains the equipment energy consumption data of each path in the storage area, calculates the gradient distribution of equipment operation energy consumption on the path, adjusts the path according to the gradient distribution, and obtains the path energy consumption optimization result;

[0008] The task execution time monitoring module obtains the corresponding storage task time in the path energy consumption optimization result, calculates the execution density of the storage task in the time dimension, marks the congestion risk according to the execution density, and obtains the task execution time distribution information;

[0009] The task congestion adjustment module obtains the execution density trend of the storage task in the task execution time distribution information, calculates the congestion index of the storage task with congestion risk, adjusts the execution time window of some storage tasks, and obtains the task congestion adjustment result;

[0010] The task consistency optimization module extracts the execution order constraints between each storage task from the task congestion adjustment result, adjusts the earliest executable time of the downstream storage task, and obtains the task dependency optimization result;

[0011] The dynamic warehousing task allocation module uses a deep Q network to perform simulation optimization based on the warehousing task execution order in the task dependency optimization result, dynamically allocates task execution objects and execution time, and obtains task allocation optimization results.

[0012] As a further solution of the present invention, the path energy consumption optimization results include path equipment energy consumption gradient distribution, energy consumption threshold comparison results, and path energy consumption adjustment; the task execution time distribution information includes warehouse task execution density, congestion risk time period mark, and time dimension task distribution trend; the task congestion adjustment results include congestion index calculation results, warehouse task diversion, and task execution time window adjustment; the task dependency optimization results include task execution order constraints, dependent task execution order conflict detection, and downstream task earliest executable time adjustment; the task allocation optimization results include warehouse task resource requirements, execution time schedule, path energy consumption characteristic data, and task execution object allocation.

[0013] As a further solution of the present invention, the energy consumption gradient path avoidance module includes:

[0014] The path energy consumption data acquisition submodule acquires the equipment energy consumption data of each path in the storage area at a specified time, calls the equipment energy consumption record value, extracts the equipment operation time and power consumption value, and combines the time series data to organize the time segment energy consumption data of the equipment on all paths to obtain the path equipment energy consumption data set;

[0015] The energy consumption gradient calculation submodule is based on the path device energy consumption data set and adopts the formula:

[0016]

[0017] Calculate the energy consumption gradient distribution G of the equipment on the path p , get the path energy consumption gradient distribution value;

[0018] Among them, E t represents the energy consumption of the device at time t, Represents the average device energy consumption of the path, D t Represents the running time difference of the devices on the path, and n represents the total number of time slices;

[0019] The path energy consumption optimization submodule compares the energy consumption gradient with the preset path energy consumption threshold according to the path energy consumption gradient distribution value, selects the path whose energy consumption gradient exceeds the threshold at consecutive moments, adjusts the path, reallocates the path selection of some storage tasks, optimizes the equipment's operating time window, and obtains the path energy consumption optimization result.

[0020] As a further solution of the present invention, the task execution time monitoring module includes:

[0021] The task time acquisition submodule obtains the storage task time information corresponding to the path energy consumption optimization result, extracts the start time, execution duration and completion time of each task, and sorts out the task time distribution data in each time period according to the task scheduling situation on the path to obtain the task time distribution data set;

[0022] The execution density calculation submodule is based on the task time distribution data set and uses the formula:

[0023]

[0024] Calculate the task execution density H within the time period t′ t′ ;

[0025] Among them, N t′ represents the number of tasks executed within time t′, T i represents the execution duration of task i, W t′ represents the available execution window of the task within time t′, G i represents the interval time between tasks i;

[0026] The congestion risk judgment submodule compares the distribution density with a preset density threshold according to the task execution density value, filters the time period when the task density exceeds the set threshold, and marks the filtered time period with congestion risk to obtain task execution time distribution information.

[0027] As a further solution of the present invention, the task congestion adjustment module includes:

[0028] The task density trend analysis submodule obtains the storage task execution density trend in the task execution time distribution information, extracts the task density values ​​of multiple time periods, arranges the task density change curve according to the time series, analyzes the growth, stability or decline trend of the task density, and obtains the task execution density trend data;

[0029] The congestion index calculation submodule is based on the task execution density trend data and adopts the formula:

[0030]

[0031] Calculate the congestion index value I c ;

[0032] Among them, H t′ represents the task density at time t′, max(H t′ ) and min(H t′ ) represent the maximum and minimum values ​​of the task density in the time window, N represents the total number of time windows for task statistics, T t′ represents the total execution time of all tasks within time t′, G t′ represents the average time interval between tasks;

[0033] The task diversion and time adjustment submodule compares the congestion index value with the preset storage task congestion threshold, filters the time period when the congestion index exceeds the threshold, adjusts the tasks in the filtered time period, reallocates some tasks to other time windows, adjusts the task execution time and path, and obtains the task congestion adjustment result.

[0034] As a further solution of the present invention, the task consistency optimization module includes:

[0035] The execution order constraint extraction submodule extracts the execution order constraints between each storage task from the task congestion adjustment result, obtains the adjusted task execution time, the dependency relationship and path between tasks, analyzes the before-after correlation between tasks, and obtains a task execution order list;

[0036] The dependency priority adjustment submodule determines, based on the task execution order list, whether the storage task time adjustment causes an execution order conflict between dependent storage tasks. If there is a conflict, the earliest executable time of the downstream storage task is adjusted based on the storage task dependency priority to obtain the task dependency optimization result.

[0037] As a further solution of the present invention, the dynamic storage task allocation module includes:

[0038] The task feature extraction submodule extracts the feature data of resource requirements, execution time, and path energy consumption of the warehousing tasks based on the execution order of the warehousing tasks in the task dependency optimization results, calls the task data records and real-time task flows, filters invalid feature items, and uses a deep Q network for simulation optimization to obtain optimized feature data of the warehousing tasks;

[0039] The task allocation optimization submodule is based on the warehouse task optimization feature data, calls task priority, equipment availability and operation window constraints, calculates the adaptability score of the task execution order, dynamically sorts the tasks according to their priorities, adjusts the execution time and execution objects of conflicting tasks, and obtains the task allocation optimization results.

[0040] Compared with the prior art, the advantages and positive effects of the present invention are:

[0041] In the present invention, by the refined calculation of the energy consumption data of the equipment in the storage area and the dynamic adjustment of the path energy consumption gradient, the material handling path is optimized, the use frequency of the high energy consumption path is reduced, and the energy efficiency ratio of the equipment operation is improved. Combined with the density analysis of the task execution time, the task execution status of different time periods is monitored, the concentration of peak tasks is identified, and based on the trend analysis of the task density, the potential congestion risk is predicted, the task execution window is reasonably adjusted, and the operation delay caused by excessive concentration of tasks is reduced. The execution order of the storage tasks is analyzed, and the tasks with close dependencies are adjusted so that they are executed at a reasonable time point to avoid the overall operation efficiency being affected by the conflict of the task execution order. In the process of dynamic task allocation, the reinforcement learning method is used for simulation optimization, and the task priority, equipment availability and operation window constraints are comprehensively considered to make the task allocation more intelligent and improve the overall operation coordination ability of the warehouse. By screening and optimizing the task feature data, it is ensured that the selection of the storage task execution object is more in line with the real-time operation situation, the occurrence of inefficient task allocation is reduced, and the rationality of task scheduling is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 is a system flow chart of the present invention;

[0043] Figure 2 This is a flow chart of the energy consumption gradient path avoidance module of the present invention;

[0044] Figure 3 It is a flowchart of the task execution time monitoring module of the present invention;

[0045] Figure 4 A flowchart of a task congestion adjustment module of the present invention;

[0046] Figure 5 A flowchart of the task consistency optimization module of the present invention;

[0047] Figure 6 This is a flow chart of the dynamic warehousing task allocation module of the present invention. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0049] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0050] See also Figure 1 , the AI-based logistics warehousing task allocation system includes:

[0051] The energy consumption gradient path avoidance module obtains the equipment energy consumption data of each path in the storage area, calculates the gradient distribution of equipment operation energy consumption on the path, adjusts the path according to the gradient distribution, and obtains the path energy consumption optimization result;

[0052] The task execution time monitoring module obtains the corresponding storage task time in the path energy consumption optimization result, calculates the execution density of the storage task in the time dimension, marks the congestion risk according to the execution density, and obtains the task execution time distribution information;

[0053] The task congestion adjustment module obtains the execution density trend of the storage task in the task execution time distribution information, calculates the congestion index of the storage task with congestion risk, adjusts the execution time window of some storage tasks, and obtains the task congestion adjustment result;

[0054] The task consistency optimization module extracts the execution order constraints between each storage task from the task congestion adjustment results, adjusts the earliest executable time of the downstream storage tasks, and obtains the task dependency optimization results;

[0055] The dynamic storage task allocation module uses a deep Q network to simulate and optimize the storage task execution order in the task dependency optimization result, dynamically allocates task execution objects and execution time, and obtains the task allocation optimization result;

[0056] The path energy consumption optimization results include path equipment energy consumption gradient distribution, energy consumption threshold comparison results, and path energy consumption adjustment. The task execution time distribution information includes warehouse task execution density, congestion risk time period mark, and time dimension task distribution trend. The task congestion adjustment results include congestion index calculation results, warehouse task diversion, and task execution time window adjustment. The task dependency optimization results include task execution order constraints, dependent task execution order conflict detection, and downstream task earliest executable time adjustment. The task allocation optimization results include warehouse task resource requirements, execution time schedule, path energy consumption characteristic data, and task execution object allocation.

[0057] See also Figure 2 , the energy consumption gradient path avoidance module includes:

[0058] The path energy consumption data acquisition submodule acquires the equipment energy consumption data of each path in the storage area at a specified time, calls the equipment energy consumption record value, extracts the equipment operation time and power consumption value, and combines the time series data to organize the time segment energy consumption data of the equipment on all paths to obtain the path equipment energy consumption data set;

[0059] To obtain the equipment energy consumption data for each path in the storage area at a specified time, first determine all possible paths in the storage area and list the equipment involved in these paths, such as conveyor belts, sorters, automatic guided vehicles (AGVs), and other equipment. Then, install energy consumption monitoring devices on these devices to record the energy consumption data of each device in different time periods in real time. For example, the conveyor belt on a certain path consumes 5kWh of electricity from 9:00 to 10:00 in the morning, the sorter consumes 3kWh of electricity in the same time period, and the AGV consumes 4kWh of electricity. Then, collect the equipment energy consumption data on all paths and organize the data format to form an equipment energy consumption record table for each path in a specified time period.

[0060] As shown below: Table 1 Warehouse path equipment energy consumption record table;

[0061] Path number Device Name Time period (hours) Energy consumption (kWh) P1 conveyor 9:00-10:00 5 P1 Sorting Machine 9:00-10:00 3 P1 AGV 9:00-10:00 4 P2 conveyor 9:00-10:00 6 P2 Sorting Machine 9:00-10:00 2

[0062] As shown in Table 1, the energy consumption data of the equipment on each path in different time periods are stored for the subsequent calculation of the energy consumption gradient of the path. Finally, all the data are sorted out to establish a database containing the path number, equipment type, time period and corresponding energy consumption, and the path equipment energy consumption data set is obtained.

[0063] The energy consumption gradient calculation submodule is based on the path device energy consumption dataset and uses the formula:

[0064]

[0065] Calculate the energy consumption gradient distribution G of the equipment on the path p, get the path energy consumption gradient distribution value;

[0066] Among them, E t represents the energy consumption of the device at time t, Represents the average device energy consumption of the path, D t Represents the running time difference of the devices on the path, and n represents the total number of time slices.

[0067] First, for each path, obtain the equipment energy consumption data in continuous time periods. For example, a path consumes 5kWh in the first hour, 6kWh in the second hour, and 7kWh in the third hour. Then, calculate the energy consumption change rate between adjacent time periods to obtain the energy consumption gradient. For example, the energy consumption gradient between the second hour and the first hour is calculated as follows: That is, the energy consumption increases by 20%. Similarly, the energy consumption gradient between the third hour and the second hour is: Next, calculate the average energy consumption of the path during the entire observation period, assuming that it is calculated as follows: And calculate the square root of the sum of squares of the deviations of energy consumption in each time period from the average energy consumption:

[0068] In addition, considering the difference in running time of devices on the path, assuming that each time period is 1 hour and the total running time is 3 hours, the following formula is used to calculate the path energy consumption gradient distribution value:

[0069] In the calculation process, the threshold of the energy consumption gradient is set to 0.3. The value is set based on the mean square error calculation result of the historical energy consumption data of all paths in the storage center. The mean square error reflects the discrete degree of energy consumption changes. Usually, a value greater than 1.5 times the mean square error is taken as the basis for judging whether the energy consumption fluctuation is significant. Assuming that the mean square error is 0.2, the threshold is set to T = 1.5 × 0.2 = 0.3. The specific adjustment of this value can fluctuate with the type of equipment and task load on the path. For example, if the path is mainly composed of high-energy consumption equipment (such as large sorting machines), the threshold can be appropriately increased to 0.35-0.4 to avoid misjudgment; if the equipment on the path is mostly low-energy consumption equipment (such as conveyor belts), the threshold can be reduced to 0.25 to enhance sensitivity. Finally, the path energy consumption gradient distribution value is obtained by calculation, and the energy consumption change trend of the equipment on the path is judged to obtain the path energy consumption gradient distribution.

[0070] The path energy consumption optimization submodule compares the energy consumption gradient with the preset path energy consumption threshold according to the path energy consumption gradient distribution value, selects the path whose energy consumption gradient exceeds the threshold at consecutive moments, adjusts the path, reallocates the path selection of some storage tasks, optimizes the equipment's operating time window, and obtains the path energy consumption optimization result;

[0071] Call the path energy consumption gradient distribution, compare the energy consumption gradient with the preset path energy consumption threshold, assume that the energy consumption threshold is set to 0.3 (based on the above mean square error calculation result), and screen out the paths whose energy consumption gradients exceed 0.3 at consecutive moments. For example, the energy consumption gradients of a path in three consecutive time periods are 0.35, 0.4 and 0.45 respectively, all exceeding the threshold of 0.3, then the path is judged to have high energy consumption. Then, adjust the task execution mode of the path. The specific adjustment plan is as follows, Table 2 Energy consumption optimization adjustment plan:

[0072] Path number Original mission Adjusted tasks Adjustment method P1 Cargo A transportation Cargo B Transportation Task transfer to low energy path P2 Cargo C Transportation Cargo C Transportation Adjusting the device operating time window

[0073] When optimizing and adjusting the path energy consumption, the principle of task redistribution is to ensure that the task load on the high-energy consumption path is reduced by no more than 30%, so as to avoid overloading other paths due to excessive task diversion. This ratio is calculated through the path load balancing within the storage center. Finally, the energy consumption of the adjusted path is recorded to obtain the path energy consumption optimization result.

[0074] See also Figure 3 ,The task execution time monitoring module includes:

[0075] The task time acquisition submodule obtains the corresponding storage task time information in the path energy consumption optimization result, extracts the start time, execution duration and completion time of each task, and sorts out the task time distribution data in each time period according to the task scheduling on the path to obtain the task time distribution data set;

[0076] To obtain the corresponding storage task time information in the path energy consumption optimization result, first, extract the start time, execution duration and completion time of all tasks. For different storage tasks, including picking, handling, storage and other tasks, during the scheduling process, the time node of each task is recorded in the scheduling system. For example, task A starts at 08:00, lasts for 30 minutes, and is completed at 08:30, while task B starts at 08:20, lasts for 40 minutes, and is completed at 09:00. These data need to be sorted according to the dimensions of task type, execution path, time, etc. to form task time distribution data. For details, please refer to Table 3 Storage Task Time Distribution Table:

[0077]

[0078]

[0079] As shown in Table 3, the time distribution of different tasks can be organized into a data set and used for subsequent calculation of task execution density. Finally, a task time distribution data set is obtained.

[0080] The execution density calculation submodule is based on the task time distribution data set and uses the formula:

[0081]

[0082] Calculate the task execution density H within the time period t′ t′ ;

[0083] Among them, N t′ represents the number of tasks executed within time t′, T i represents the execution duration of task i, W t′ represents the available execution window of the task within time t′, G i Represents the interval time between task i.

[0084] If the execution durations of two tasks during 08:30-09:00 are 40 minutes and 30 minutes respectively, the available execution window is 60 minutes, and the interval between tasks is 10 minutes, then the task execution density is calculated as follows: When calculating task density, the time window W t′ The setting of W is based on the overall throughput of the warehouse. Generally, it can be taken as the time slice length corresponding to the maximum number of parallel tasks in the warehouse. For example, a warehouse can run 10 tasks at the same time, and each task lasts an average of 30 minutes. Then W t′ It is set to 30 minutes, and this value is adjusted with the warehouse throughput capacity. For example, when the warehouse adds new equipment or automation systems, this value can be expanded to 45 minutes. Conversely, when equipment maintenance or manpower is reduced, this value can be reduced to 20 minutes. Finally, the task execution density value is obtained by calculation, and the task distribution in different time periods is determined to obtain the task execution density distribution.

[0085] The congestion risk judgment submodule compares the distribution density with the preset density threshold according to the task execution density value, selects the time period when the task density exceeds the set threshold, and marks the selected time period with congestion risk to obtain the task execution time distribution information;

[0086] Call the task execution density distribution, compare the distribution density with the preset density threshold, assume that the density threshold is set to 0.8, that is, when the task execution density value ≥ 0.8, it is determined that there is a congestion risk in the time period, and the time period where the task density exceeds the set threshold is screened out. For example, the density value during 08:30-09:00 is 1.0, which is higher than 0.8. Therefore, this time period is marked as a congested time period, and the data of this time period is stored in the congestion record table, and the task time distribution trend data is established. Table 4 Task congestion time period record table:

[0087] Time period Task density Congestion risk 08:00-08:30 0.5 no 08:30-09:00 1.0 yes 09:00-09:30 0.8 yes

[0088] The density threshold of 0.8 is set based on the historical task load of the warehouse and the impact of task overload. Usually, 80% of the task load is selected as the upper limit to prevent the high task density from affecting the smoothness of warehouse operation. For example, if the maximum number of parallel tasks in the warehouse is 10, the number of tasks running within a 30-minute window should not exceed 8. The corresponding density threshold can be calculated as: When the warehouse throughput capacity is improved, the density threshold can be appropriately adjusted to 0.85-0.9. For example, if the automation equipment increases the operating capacity by 20%, the new first threshold is: T′ 阈1 =0.8×(1+0.2)=0.96. If the warehouse throughput capacity decreases by 10% due to maintenance or equipment failure, the second threshold is adjusted to: T′ 阈2 =0.8×(1-0.1)=0.72. As shown in Table 4, the time period with a task density higher than 0.8 is marked as a congestion risk. Finally, the task execution time distribution information is obtained.

[0089] See also Figure 4 ,The task congestion adjustment module includes:

[0090] The task density trend analysis submodule obtains the storage task execution density trend in the task execution time distribution information, extracts the task density values ​​of multiple time periods, arranges the task density change curve according to the time series, analyzes the growth, stability or decline trend of the task density, and obtains the task execution density trend data;

[0091] Obtain the trend of storage task execution density in the task execution time distribution information, extract the task density values ​​in different periods, and sort out the task density change curve according to the time series, calculate the task density change rate in each time period, and analyze the growth, stability or decline trend of task density. First, set the time window W t′ The time window is set according to the average task execution time and scheduling rhythm of the warehouse. Usually, the 75% quantile of the historical task execution time is taken as the window length to ensure that most of the task execution process is covered. Assuming that the average execution time of the historical tasks of the warehouse is 28 minutes, its 75% quantile is 30 minutes. Therefore, W is set. t′ = 30 minutes, and the task density is counted in the time periods of 08:00-08:30, 08:30-09:00, 09:00-09:30, etc. The task density calculation results are calculated according to the above formula, which are 0.167, 0.267, and 0.333 respectively. By calculating the task density in different time periods, the task density trend data can be obtained, and the task density change range can be further analyzed to establish the task execution density trend data.

[0092] The congestion index calculation submodule is based on the task execution density trend data and uses the formula:

[0093]

[0094] Calculate the congestion index value I c ;

[0095] Among them, H t′ represents the task density at time t′, max(H t′ ) and min(H t′ ) represent the maximum and minimum values ​​of the task density in the time window, N represents the total number of time windows for task statistics, T t′ represents the total execution time of all tasks within time t′, G t′ Represents the average time interval between tasks.

[0096] Set the task interval time threshold G 阈 To control the minimum interval between tasks and avoid excessive accumulation between tasks, G 阈 Take the 25% quantile of the historical task interval. Assuming that the 25% quantile of the historical task interval is 12 minutes, then: G 阈 =12, if the calculated G t′ If the time window is lower than the threshold, there may be a risk of task accumulation in this time period. Assuming that the task execution time in the time window is T t′ If the time intervals are 30, 45, and 50 minutes respectively, the task congestion index is calculated as follows: Task congestion index threshold I 阈 It is used to determine whether the task density exceeds the allowable range. The threshold is usually selected from the 90% quantile of the historical congestion index. Assuming that the 90% quantile of the historical task congestion index is 0.0025, then: I 阈 =0.0025. Finally, the calculated task congestion index distribution can be used for subsequent task diversion and adjustment.

[0097] The task diversion and time adjustment submodule compares the congestion index value with the preset storage task congestion threshold, selects the time period when the congestion index exceeds the threshold, and adjusts the tasks in the selected time period, reallocates some tasks to other time windows, adjusts the task execution time and path, and obtains the task congestion adjustment result;

[0098] Call the task congestion index distribution, compare the congestion index value with the preset storage task congestion threshold, filter out the time periods where the congestion index exceeds the threshold, and adjust the tasks in these time periods. Assuming the congestion threshold is set to 0.0025, then: I c=0.0029>0.0025, it is determined that there is task congestion in this time period, and some tasks are reallocated to the low-density time window, and the task execution time and path are adjusted. For example, some tasks from 08:30-09:00 are adjusted to 09:30-10:00, and task scheduling on high-density paths is reduced. In order to ensure that task adjustment does not cause new congestion problems, the path congestion coefficient P is set. 阈 To measure the path load, the threshold is set based on the path throughput capacity. Usually, the 85% quantile of the historical path load is taken as the upper limit. Assuming that the 85% quantile of a path is 0.7, then: P 阈 =0.7. If the adjusted path load exceeds the threshold, the path cannot receive new task scheduling. As shown in Table 5, the path selection of the adjusted task must meet P 调整 ≤P 阈 Conditions, Table 5 Task Adjustment Plan:

[0099]

[0100] As shown in Table 5, the time period when the task congestion index exceeds the threshold is adjusted, and some tasks are rescheduled to low-density time periods and low-load paths. The adjusted path load P 调整 It must be less than the set path load threshold P 阈 =0.7, ensuring that task adjustment will not cause new path congestion problems. The task time distribution after congestion adjustment is re-evaluated based on the throughput capacity of the adjusted execution window. Finally, the adjusted task time distribution generates the task congestion adjustment result.

[0101] See also Figure 5 , the task coherence optimization module includes:

[0102] The execution order constraint extraction submodule extracts the execution order constraints between each storage task from the task congestion adjustment result, obtains the adjusted task execution time, the dependency relationship and path between tasks, analyzes the correlation between tasks, and obtains the task execution order list;

[0103] Extract the execution order constraints between each storage task. First, obtain the adjusted task execution time information, including the task start time. Task completion time As well as the time interval between tasks, the dependencies between tasks are collected to form a task execution order relationship table, as shown in Table 6. The before-after correlation between tasks is analyzed and a task execution order constraint matrix is ​​constructed, where the execution order constraint of task i1 and task j1 can be expressed as: In the specific execution process, it is assumed that there are three tasks A, B and C, and their initial task execution time is as follows, Table 6 Task execution schedule:

[0104]

[0105] The task execution order constraint matrix calculated according to the above formula is as follows: This matrix represents the dependencies between tasks, where O 1,2 =0 means that the start time of task B is earlier than the end time of task A, there is a dependency conflict, and finally the task execution order constraint information is generated.

[0106] The dependency priority adjustment submodule determines whether the execution order of dependent warehousing tasks conflicts due to the adjustment of warehousing task time based on the task execution order list. If there is a conflict, the earliest executable time of the downstream warehousing task is adjusted based on the dependency priority of the warehousing task to obtain the task dependency optimization result.

[0107] Determine whether the adjustment of the storage task time causes a conflict in the execution order of dependent storage tasks, and calculate the start time of the adjusted task Completion time of its predecessor task If there is a relationship between If a conflict occurs, all conflicting tasks are screened, the task priority list is called, and the earliest executable time of the conflicting tasks is adjusted according to the task urgency, path availability, and resource occupancy. The calculation formula is: in, represents the minimum adjustment time of task i1, which depends on the storage equipment response time and path conversion time. If the equipment response time is 5 minutes and the path conversion time is 3 minutes, then Minutes, the specific adjusted task time is as follows

[0108] As shown in Table 7, the execution time of the task after adjustment is:

[0109]

[0110] As shown in Table 7, since the original start time of tasks B and C is earlier than the completion time of tasks A and B, they need to be adjusted, and finally the task dependency optimization result is obtained.

[0111] See also Figure 6 ,The dynamic storage task allocation module includes:

[0112] The task feature extraction submodule extracts the feature data of resource requirements, execution time, and path energy consumption of the warehousing tasks based on the execution order of the warehousing tasks in the task dependency optimization results, calls the task data records and real-time task flows, filters out invalid feature items, and uses a deep Q network for simulation optimization to obtain the optimized feature data of the warehousing tasks.

[0113] First, obtain the characteristic data of resource requirements, execution time and path energy consumption of warehousing tasks. Suppose there are five tasks, whose resource requirements are 2, 3, 1, 4 and 2 units respectively, the execution time is 5, 6, 3, 7 and 4 hours respectively, and the path energy consumption is 10, 15, 8, 20 and 12 kWh respectively. Next, the deep Q network (DQN) is used to simulate and optimize in the task data records and real-time task streams. DQN is an algorithm that combines deep learning with reinforcement learning and can process high-dimensional input data. First, define the state space, including the resource requirements, execution time and path energy consumption of the task. The action space is the execution order of the task. The reward function is set as the efficiency of task completion and resource utilization. Suppose in a certain state, a task execution order is selected and its reward value is calculated to be 10. Use experience replay and target network and other technologies to train the DQN model. After multiple iterations, the model converges and obtains the optimal task execution strategy. Finally, generate the optimized characteristic data of warehousing tasks.

[0114] The task allocation optimization submodule is based on the warehouse task optimization feature data, calls the task priority, equipment availability and operation window constraints, calculates the adaptability score of the task execution sequence, dynamically sorts the tasks according to their priorities, adjusts the execution time and execution objects of conflicting tasks, and obtains the task allocation optimization results;

[0115] First, we obtain the task priority, device availability, and job window constraints. Assume that there are three tasks with priorities 1, 2, and 3, respectively. The device availability is that device A is available from 8:00 to 12:00, and device B is available from 13:00 to 17:00. The job window is that task 1 needs to be completed between 8:00 and 10:00, task 2 between 10:00 and 14:00, and task 3 between 14:00 and 16:00. Next, we calculate the adaptability score of the task execution order. The adaptability score is set as: S = αP+βA+γW, where S is the adaptability score of the task execution order; P represents the priority score of the task, which is set to P=100-10×(priority-1); A represents the equipment availability score, which is set to A=100-20×(whether it complies with the equipment availability constraint, 0 if it complies, and 1 if it does not comply); W represents the job window constraint score, which is set to W=100-30×(whether it complies with the job window constraint, 0 if it complies, and 1 if it does not comply); α, β, and γ are weight parameters, which are set to α=0.4, β=0.3, and γ=0.3.

[0116] If Task 1, Task 2, and Task 3 are executed in the order of (Task 1 → Task 2 → Task 3), calculate their adaptability scores: Calculate the adaptability score of Task 1: Priority P1 = 100 - 10 × (1 - 1) = 100;

[0117] Task 1 is executed on device A, which meets the device availability constraint, so

[0118] A1=100-20×0=100; Task 1 is executed from 8:00 to 10:00, which meets the job window, so W1=100-30×0=100; S1=0.4×100+0.3×100+0.3×100=100.

[0119] Calculate the fitness score for Task 2:

[0120] Priority P2 = 100-10×(2-1) = 90; Task 2 is executed on device B, which meets the device availability constraint, so A2 = 100-20×0 = 100; Task 2 is executed from 10:00 to 14:00, which meets the job window, so W2 = 100-30×0 = 100; S2 = 0.4×90+0.3×100+0.3×

[0121] 100=96.

[0122] Calculate the adaptability score of task 3: priority P3 = 100 - 10 × (3 - 1) = 80;

[0123] Task 3 is executed on device A, which meets the device availability constraint, so

[0124] A3=100-20×0=100; Task 3 is executed from 14:00 to 16:00, which meets the job window, so W3=100-30×0=100; S3=0.4×80+0.3×100+0.3×100=92.

[0125] Finally, the overall fitness score of the task execution sequence (task 1 → task 2 → task 3) is:

[0126]

[0127] Dynamically sort tasks according to their priorities, adjust the execution time and objects of conflicting tasks, ensure job windows and equipment availability constraints, and finally obtain task allocation optimization results.

[0128] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. AI-based logistics warehousing task allocation system, characterized by: The system comprises: The energy consumption gradient path avoidance module obtains the equipment energy consumption data of each path in the storage area, calculates the gradient distribution of equipment operation energy consumption on the path, adjusts the path according to the gradient distribution, and obtains the path energy consumption optimization result; The task execution time monitoring module obtains the corresponding storage task time in the path energy consumption optimization result, calculates the execution density of the storage task in the time dimension, marks the congestion risk according to the execution density, and obtains the task execution time distribution information; The task congestion adjustment module obtains the execution density trend of the storage task in the task execution time distribution information, calculates the congestion index of the storage task with congestion risk, adjusts the execution time window of some storage tasks, and obtains the task congestion adjustment result; The task consistency optimization module extracts the execution order constraints between each storage task from the task congestion adjustment result, adjusts the earliest executable time of the downstream storage task, and obtains the task dependency optimization result; The dynamic warehousing task allocation module uses a deep Q network to perform simulation optimization based on the warehousing task execution order in the task dependency optimization result, dynamically allocates task execution objects and execution time, and obtains task allocation optimization results.

2. The AI-based logistics warehousing task allocation system according to claim 1 is characterized in that: The path energy consumption optimization result includes path equipment energy consumption gradient distribution, energy consumption threshold comparison result, and path energy consumption adjustment; the task execution time distribution information includes warehouse task execution density, congestion risk time period mark, and time dimension task distribution trend; the task congestion adjustment result includes congestion index calculation result, warehouse task diversion, and task execution time window adjustment; the task dependency optimization result includes task execution order constraint, dependent task execution order conflict detection, and downstream task earliest executable time adjustment; the task allocation optimization result includes warehouse task resource demand, execution time schedule, path energy consumption characteristic data, and task execution object allocation.

3. The AI-based logistics warehousing task allocation system according to claim 2 is characterized in that: The energy consumption gradient path avoidance module includes: The path energy consumption data acquisition submodule acquires the equipment energy consumption data of each path in the storage area at a specified time, calls the equipment energy consumption record value, extracts the equipment operation time and power consumption value, and combines the time series data to organize the time segment energy consumption data of the equipment on all paths to obtain the path equipment energy consumption data set; The energy consumption gradient calculation submodule is based on the path device energy consumption data set and adopts the formula: Calculate the energy consumption gradient distribution G of the equipment on the path p , get the path energy consumption gradient distribution value; Among them, E t represents the energy consumption of the device at time t, Represents the average device energy consumption of the path, D t Represents the running time difference of the devices on the path, and n represents the total number of time slices; The path energy consumption optimization submodule compares the energy consumption gradient with the preset path energy consumption threshold according to the path energy consumption gradient distribution value, selects the path whose energy consumption gradient exceeds the threshold at consecutive moments, adjusts the path, reallocates the path selection of some storage tasks, optimizes the equipment's operating time window, and obtains the path energy consumption optimization result.

4. The AI-based logistics warehousing task allocation system according to claim 3 is characterized in that: The task execution time monitoring module includes: The task time acquisition submodule obtains the storage task time information corresponding to the path energy consumption optimization result, extracts the start time, execution duration and completion time of each task, and sorts out the task time distribution data in each time period according to the task scheduling situation on the path to obtain the task time distribution data set; The execution density calculation submodule is based on the task time distribution data set and uses the formula: Calculate the task execution density H within the time period t′ t′ ; Among them, N t′ represents the number of tasks executed within time t′, T i represents the execution duration of task i, W t′ represents the available execution window of the task within time t′, G i represents the interval time between tasks i; The congestion risk judgment submodule compares the distribution density with a preset density threshold according to the task execution density value, filters the time period when the task density exceeds the set threshold, and marks the filtered time period with congestion risk to obtain task execution time distribution information.

5. The AI-based logistics warehousing task allocation system according to claim 4 is characterized in that: The task congestion adjustment module includes: The task density trend analysis submodule obtains the storage task execution density trend in the task execution time distribution information, extracts the task density values ​​of multiple time periods, arranges the task density change curve according to the time series, analyzes the growth, stability or decline trend of the task density, and obtains the task execution density trend data; The congestion index calculation submodule is based on the task execution density trend data and adopts the formula: Calculate the congestion index value I c ; Among them, H t′ represents the task density at time t′, max(H t′ ) and min(H t′ ) represent the maximum and minimum values ​​of the task density in the time window, N represents the total number of time windows for task statistics, T t′ represents the total execution time of all tasks within time t′, G t′ represents the average time interval between tasks; The task diversion and time adjustment submodule compares the congestion index value with the preset storage task congestion threshold, filters the time period when the congestion index exceeds the threshold, adjusts the tasks in the filtered time period, reallocates some tasks to other time windows, adjusts the task execution time and path, and obtains the task congestion adjustment result.

6. The AI-based logistics warehousing task allocation system according to claim 5 is characterized in that: The task consistency optimization module includes: The execution order constraint extraction submodule extracts the execution order constraints between each storage task from the task congestion adjustment result, obtains the adjusted task execution time, the dependency relationship and path between tasks, analyzes the before-after correlation between tasks, and obtains a task execution order list; The dependency priority adjustment submodule determines, based on the task execution order list, whether the storage task time adjustment causes an execution order conflict between dependent storage tasks. If there is a conflict, the earliest executable time of the downstream storage task is adjusted based on the storage task dependency priority to obtain the task dependency optimization result.

7. The AI-based logistics warehousing task allocation system according to claim 6 is characterized in that: The dynamic storage task allocation module includes: The task feature extraction submodule extracts the feature data of resource requirements, execution time, and path energy consumption of the warehousing tasks based on the execution order of the warehousing tasks in the task dependency optimization results, calls the task data records and real-time task flows, filters invalid feature items, and uses a deep Q network for simulation optimization to obtain optimized feature data of the warehousing tasks; The task allocation optimization submodule is based on the warehouse task optimization feature data, calls task priority, equipment availability and operation window constraints, calculates the adaptability score of the task execution order, dynamically sorts the tasks according to their priorities, adjusts the execution time and execution objects of conflicting tasks, and obtains the task allocation optimization results.

Images