A multifunctional control method and system of an AI intelligent storage cabinet

Abstract

The application provides a multifunctional control method and system of an AI intelligent storage cabinet, belongs to the technical field of intelligent control, acquires user behavior data, environment data, article attribute data and real-time cabinet state data, generates a multi-dimensional intelligent storage cabinet data set, carries out feature extraction and fusion on the multi-dimensional intelligent storage cabinet data set, generates an intelligent storage cabinet evaluation index set, provides comprehensive and scientific data support for subsequent data analysis, constructs a storage cabinet intelligent control model based on a storage cabinet intelligent control mechanism, realizes dynamic balance of regional load, accurate matching of storage time limit and accurate implementation of personalized anti-forgetting intervention through the storage cabinet intelligent control model, acquires real-time user demand data, carries out demand analysis based on the intelligent storage cabinet evaluation index set, acquires a demand analysis result, and generates an intelligent storage scheme in combination with the storage cabinet intelligent control model, thereby improving resource utilization rate and user experience of the intelligent storage cabinet.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent control, in particular to a multifunctional control method and system of an AI intelligent storage cabinet. BACKGROUND

[0002] As an important carrier of modern urban logistics and convenient services, intelligent storage cabinets are widely used in various scenarios, but there are still some limitations in user demand and system efficiency in actual operation.

[0003] On the one hand, the regional load balancing capability of the traditional technology is relatively weak, only by dividing the load level through static rules, it often lacks the combination of environmental impact indicators and storage cabinet operation indicators to dynamically adjust the regional load of the intelligent storage cabinet, so that the phenomenon of storage cabinet storage grid saturation in high load area and resource idling in low load area coexist; on the other hand, the storage time limit allocation is mechanical, often without constructing a flexible adjustment mechanism according to user behavior law indicators and article attribute indicators, and the fixed storage time limit often leads to waste of storage resources; on the other hand, the anti-forgetting intervention strategy lacks pertinence, only according to a single risk threshold triggering standard reminder, without considering the response difference of different users to intervention methods, it often easily produces the phenomenon of over-intervention to users with high daily access frequency and insufficient intervention to users with frequent history overtime.

[0004] In summary, the traditional technology often cannot dynamically balance the regional load, accurately match the storage time limit, and implement personalized anti-forgetting intervention, resulting in low resource utilization rate of the intelligent storage cabinet and uneven user experience. SUMMARY

[0005] In order to solve the above technical problems, the present application provides a multifunctional control method and system of an AI intelligent storage cabinet to solve the technical problems that the traditional technology often cannot dynamically balance the regional load, accurately match the storage time limit, and implement personalized anti-forgetting intervention in the prior art.

[0006] The purpose and effect of the multifunctional control method and system of an AI intelligent storage cabinet of the present application are achieved by the following specific technical means:

[0007] A multifunctional control method of an AI intelligent storage cabinet, comprising:

[0008] Obtaining user behavior data, environment data, article attribute data and real-time cabinet state data to generate a multi-dimensional intelligent storage cabinet data set;

[0009] Feature extraction and fusion are performed on the multi-dimensional intelligent storage cabinet data set to generate an intelligent storage cabinet evaluation index set;

[0010] An intelligent storage cabinet control model is constructed based on an intelligent storage cabinet control mechanism;

[0011] Real-time user demand data is acquired, and demand analysis is performed based on the intelligent storage cabinet evaluation index set;

[0012] The demand analysis result is acquired, and an intelligent storage scheme is generated in combination with the intelligent storage cabinet control model.

[0013] As a further scheme of the present application, feature extraction and fusion are performed on the multi-dimensional intelligent storage cabinet data set to generate an intelligent storage cabinet evaluation index set, including:

[0014] Time dimension feature extraction is performed on user behavior data to acquire user behavior regularity indexes, the user behavior regularity indexes at least including daily average access times, peak period access frequency, and pickup timeout duration distribution;

[0015] Statistical dimension feature extraction is performed on environment data to acquire environment influence indexes, the environment influence indexes at least including temperature influence coefficient, humidity influence coefficient, and regional peak passenger flow;

[0016] Classification feature extraction is performed on item attribute data to acquire item attribute indexes, the item attribute indexes at least including perishable item proportion, special storage demand item quantity, and average item volume;

[0017] State feature extraction is performed on real-time cabinet state data to acquire storage cabinet operation indexes, the storage cabinet operation indexes at least including storage compartment turnover rate, fault repair efficiency, and function utilization rate;

[0018] Feature fusion is performed on the user behavior regularity indexes, the environment influence indexes, the item attribute indexes, and the storage cabinet operation indexes to generate an intelligent storage cabinet evaluation index set.

[0019] As a further scheme of the present application, an intelligent storage cabinet control model is constructed based on an intelligent storage cabinet control mechanism, including:

[0020] A regional balanced load module of the intelligent storage cabinet control model is constructed based on a regional load balancing function contained in the intelligent storage cabinet control mechanism;

[0021] A user storage time limit allocation module of the intelligent storage cabinet control model is constructed based on a credit time correlation function contained in the intelligent storage cabinet control mechanism;

[0022] A forgetfulness prevention intervention module of the intelligent storage cabinet control model is constructed based on a forgetfulness risk function contained in the intelligent storage cabinet control mechanism;

[0023] An intelligent storage scheme is generated according to output results of the regional balanced load module, the user storage time limit allocation module, and the forgetfulness prevention intervention module.

[0024] As a further scheme of the present application, a regional load balancing module of the intelligent control model of the storage cabinet is constructed based on a regional load balancing function contained in the intelligent control mechanism of the storage cabinet, comprising:

[0025] A regional load coefficient is obtained based on the regional load balancing function and in combination with an environmental impact index and a storage cabinet operation index contained in the intelligent storage cabinet evaluation index set;

[0026] The method of regionally grading the distribution region of the storage cabinet is defined as a plurality of regionally grading strategies, and the plurality of regionally grading strategies are taken as arms to be selected in the contextual multi-arm bandit algorithm, a contextual feature vector of the arms to be selected is constructed based on an environmental impact index and a storage cabinet operation index contained in the intelligent storage cabinet evaluation index set;

[0027] The optimal arm is screened from the arms to be selected according to the LinUCB algorithm and a preset reward function, a threshold correction factor is generated according to the obtained optimal arm, and the threshold corresponding to the regionally grading is corrected based on the threshold correction factor;

[0028] The distribution region of the storage cabinet is regionally graded based on the regional load coefficient.

[0029] As a further scheme of the present application, a user storage time limit allocation module of the intelligent control model of the storage cabinet is constructed based on a credit time effectiveness correlation function contained in the intelligent control mechanism of the storage cabinet, comprising:

[0030] A user credit score is obtained based on the credit time effectiveness correlation function and in combination with a user behavior regularity index contained in the intelligent storage cabinet evaluation index set;

[0031] The user stored items are divided into three storage grades of short-term storage, standard storage and long-term storage based on an item attribute index contained in the intelligent storage cabinet evaluation index set, and a corresponding basic storage time limit is allocated to each storage grade;

[0032] The basic storage time limit is adjusted according to the user credit score to generate an initial storage time limit;

[0033] A time limit tuning function is constructed, a preset scaling factor is taken as the input of the time limit tuning function, and a user actual overtime rate is taken as the output, the pickup efficiency corresponding to different scaling factors is evaluated based on the user actual overtime rate, the optimal scaling factor is obtained based on the Gaussian process kernel function with the objective of minimizing the actual overtime rate, and the initial storage time limit is nonlinearly calibrated according to the optimal scaling factor;

[0034] The user actual overtime rate is represented as a ratio of the number of times of user overtime pickup to the total number of pickups, and the overtime pickup behavior is defined as a pickup behavior with a pickup overtime time exceeding a value of a product of an initial storage time limit and a scaling factor.

[0035] As a further scheme of the present application, a forgetfulness prevention intervention module of the intelligent storage cabinet control model is constructed based on a forgetfulness risk function contained in the intelligent storage cabinet control mechanism, comprising:

[0036] Based on the forgetfulness risk function and in combination with a user behavior regularity index contained in the intelligent storage cabinet evaluation index set, a forgetfulness risk coefficient is obtained, and a hierarchical intervention strategy is implemented according to the forgetfulness risk value;

[0037] Based on a user behavior regularity index, an environmental impact index and an item attribute index contained in the intelligent storage cabinet evaluation index set, a user feature vector is generated, an intervention mode contained in the hierarchical intervention strategy is defined as a random forest processing variable, and two cases of user timely pickup and user overtime non-pickup are defined as random forest result variables;

[0038] The user feature vector and the random forest processing variable are taken as inputs of the random forest, and the random forest result variable is taken as an output of the random forest for linear regression fitting, so as to obtain an intervention mode contained in different hierarchical intervention strategies and improve the efficiency, and the hierarchical intervention strategy is corrected according to the intervention mode efficiency.

[0039] As a further scheme of the present application, real-time user demand data is obtained, and demand analysis is performed based on the intelligent storage cabinet evaluation index set, comprising:

[0040] Real-time user demand data is obtained according to submission information of the user end, and the real-time user demand data at least includes access time expectation data, item type data, target cabinet point range data and service demand data;

[0041] The access time expectation data is subjected to access time period demand analysis based on a user behavior regularity index contained in the intelligent storage cabinet evaluation index set, and an access time period demand analysis result is obtained;

[0042] The item type data, the target cabinet point range data and the service demand data are subjected to storage service demand analysis based on an environmental impact index, an item attribute index and a storage cabinet operation index contained in the intelligent storage cabinet evaluation index set, and a storage service demand analysis result is obtained;

[0043] The access time period demand analysis result and the storage service demand analysis result are fused according to a demand matching degree function to generate a demand matching degree, and the demand matching degree is used to quantify the matching degree between the user demand and the storage cabinet deployment point;

[0044] The storage cabinet deployment points with a demand matching degree exceeding 0.7 are divided into high-matching-degree storage points, and a priority recommendation list is generated based on the high-matching-degree storage points;

[0045] The storage cabinet deployment points with a demand matching degree in [0.5, 0.7] are divided into standard-matching-degree storage points, and an alternative recommendation list is generated based on the standard-matching-degree storage points;

[0046] The storage cabinet deployment points with a demand matching degree less than 0.5 are divided into low-matching-degree storage points and are not recommended.

[0047] As a further scheme of the present application, a demand analysis result is obtained, and an intelligent storage scheme is generated in combination with a storage cabinet intelligent control model, including:

[0048] Based on the priority recommendation list and the alternative recommendation list contained in the demand analysis result, a load-optimal storage point is obtained by calling a regional load balancing module of the storage cabinet intelligent control model;

[0049] A user storage time limit is allocated according to a user storage time limit allocation module of the storage cabinet intelligent control model;

[0050] Anti-forgetting information is generated according to an anti-forgetting intervention module of the storage cabinet intelligent control model;

[0051] The load-optimal storage point, the user storage time limit, and the anti-forgetting information are encapsulated as an intelligent storage scheme and sent to a user end.

[0052] As a further scheme of the present application, user behavior data, environment data, item attribute data, and real-time cabinet state data are obtained, and a multi-dimensional intelligent storage cabinet data set is generated, including:

[0053] User behavior data, environment data, item attribute data, and storage cabinet state data are obtained by collecting operation logs of each intelligent storage cabinet;

[0054] The user behavior data at least includes historical access time distribution, operation interval duration, and pickup punctuality rate;

[0055] The environment data at least includes real-time temperature, humidity, weather condition, and regional passenger flow;

[0056] The item attribute data at least includes item category, preservation requirement, and estimated volume;

[0057] The storage cabinet state data at least includes cabinet point storage compartment occupancy rate, fault state, and function type;

[0058] The obtained data is subjected to outlier rejection, missing value filling, and data standardization operations to generate a multi-dimensional intelligent storage cabinet data set.

[0059] A multifunctional control system of an AI intelligent storage cabinet comprises:

[0060] A data acquisition module is configured to acquire user behavior data, environment data, item attribute data and real-time cabinet state data, and generate a multi-dimensional intelligent storage cabinet data set;

[0061] An index generation module is configured to perform feature extraction and fusion on the multi-dimensional intelligent storage cabinet data set, and generate an intelligent storage cabinet evaluation index set;

[0062] A model construction module is configured to construct an intelligent storage cabinet control model according to a storage cabinet intelligent control mechanism;

[0063] A demand analysis module is configured to acquire real-time user demand data, and perform demand analysis on the user demand data based on the intelligent storage cabinet evaluation index set;

[0064] A scheme generation module is configured to generate an intelligent storage scheme according to the demand analysis result and the intelligent storage cabinet control model.

[0065] Based on the above aspects, the embodiments of the present application realize the acquisition of user behavior data, environment data, item attribute data and real-time cabinet state data, the generation of a multi-dimensional intelligent storage cabinet data set, the feature extraction and fusion on the multi-dimensional intelligent storage cabinet data set, and the generation of an intelligent storage cabinet evaluation index set, thereby providing a data basis for subsequent data analysis and scheme generation;

[0066] Based on the storage cabinet intelligent control mechanism, an intelligent storage cabinet control model is constructed, and through the constructed intelligent storage cabinet control model, the dynamic balance of regional load, the accurate matching of storage time limit and the accurate implementation of personalized anti-forgetting intervention are realized by the internal area balanced load module, the user storage time limit allocation module and the anti-forgetting intervention module of the intelligent storage cabinet control model;

[0067] Real-time user demand data is acquired, demand analysis is performed based on the intelligent storage cabinet evaluation index set, the demand analysis result is acquired, and an intelligent storage scheme is generated in combination with the intelligent storage cabinet control model, thereby realizing the accurate analysis of user demand and providing personalized intelligent storage schemes for users, so as to improve the resource utilization rate and user experience of the intelligent storage cabinet. BRIEF DESCRIPTION OF DRAWINGS

[0068] Figure 1 is an execution flow schematic diagram of a multifunctional control method of an AI intelligent storage cabinet provided by the embodiments of the present application;

[0069] Figure 2 is a schematic diagram of a multifunctional control system of an AI intelligent storage cabinet provided by the embodiments of the present application;

[0070] Figure 3 Figure 1 is a schematic diagram of a storage cabinet intelligent control model in a multifunctional control method of an AI intelligent storage cabinet according to an embodiment of the present application. DETAILED DESCRIPTION

[0071] The embodiments of the present application will be further described in detail below with reference to the accompanying drawings and embodiments. The following embodiments are used to illustrate the technical solutions of the present application, but cannot be used to limit the protection scope of the present application.

[0072] Embodiment: as shown in the accompanying drawings and embodiments: Figure 1 、 Figure 2 、 Figure 3 :

[0073] The embodiments of the present application provide a multifunctional control method of an AI intelligent storage cabinet, which is suitable for intelligent control and includes the following steps:

[0074] Step S1, acquiring user behavior data, environment data, item attribute data and real-time cabinet state data, and generating a multi-dimensional intelligent storage cabinet data set.

[0075] Specifically, the operation logs of each intelligent storage cabinet are collected to obtain corresponding user behavior data, environment data, item attribute data and storage cabinet state data.

[0076] It can be understood that the user behavior data at least includes historical access time distribution, operation interval length and pickup punctuality rate; the environment data at least includes real-time temperature, humidity, weather condition and regional passenger flow; the item attribute data at least includes item category, preservation requirement and estimated volume; and the storage cabinet state data at least includes each cabinet point storage compartment occupancy rate, fault state and function type.

[0077] Further, the min-max normalization method is used to map the collected user behavior data, environment data, item attribute data and storage cabinet state data to the [0, 1] interval; the Pandas library included in the programming language Python is used for outlier rejection, format unification and timestamp alignment, wherein the outliers include repeated data and invalid data, the repeated data are represented as completely identical records in the operation logs, and the invalid data are represented as null values and values beyond the reasonable range; and the linear interpolation method is used for missing value filling.

[0078] The correlation between each data is established through data correlation technology, such as correlating the user ID with the corresponding user behavior data, to generate a multi-dimensional intelligent storage cabinet data set.

[0079] Step S2, feature extraction and fusion are performed on the multi-dimensional intelligent storage cabinet data set to generate an intelligent storage cabinet evaluation index set.

[0080] In this embodiment, step S2 includes:

[0081] Step S21, time dimension feature extraction is performed on the user behavior data in the multi-dimensional intelligent storage cabinet data set, and user behavior regularity indexes are obtained.

[0082] Specifically, the user behavior regularity indexes at least include daily access frequency data, peak period access frequency data, and pickup timeout duration distribution data. The user behavior regularity indexes provide a data basis for subsequent allocation of user storage time limits to users.

[0083] In this embodiment, step S21 includes:

[0084] Step S21-1, daily access frequency data is obtained.

[0085] Specifically, a sliding window method is used to obtain the daily access frequency data of a user. For example, if the total access frequency of a user in 30 days is 30, then the daily access frequency data of the user is 1.

[0086] Step S21-2, peak period access frequency data is obtained.

[0087] In one possible embodiment, the access operations of a user are counted in selected morning peak periods and evening peak periods, the total number of accesses in the morning and evening peak periods is obtained, and the peak period access frequency data is obtained in combination with the length of the selected period. For example, 7:00 to 9:00 is selected as the morning peak period, and 17:00 to 19:00 is selected as the evening peak period. Assuming that the average number of accesses in the morning peak period is 120 times, then the morning peak access frequency is 120 / 2=60 times / hour.

[0088] Step S21-3, pickup timeout duration distribution data is obtained.

[0089] Specifically, the duration of each pickup timeout of a user is obtained, and distribution statistics are performed according to different duration intervals. The frequency and frequency of each interval are counted using a histogram, and pickup timeout duration distribution data is generated.

[0090] In one possible embodiment, four duration intervals of [0, 10] minutes, (10, 20] minutes, (20, 30] minutes, and more than 30 minutes are set. The [0, 10] minute interval is used to represent short delay behavior of a user caused by temporary negligence. The (10, 20] minute interval is used to represent low-level timeout behavior. The (20, 30] minute interval is used to represent medium-level timeout behavior. More than 30 minutes is used to represent high-level timeout behavior. The frequency and frequency of the occurrence of timeout events in each duration interval are counted using a histogram, a timeout distribution histogram is generated, and the timeout distribution histogram is output as the pickup timeout duration distribution data.

[0091] Step S22, statistical dimension feature extraction is performed on the environment data to obtain an environment impact index.

[0092] Specifically, the environment impact index at least includes a temperature impact coefficient, a humidity impact coefficient, and a regional peak human flow.

[0093] In this embodiment, step S22 includes:

[0094] Step S22-1, the temperature impact coefficient is obtained.

[0095] Specifically, a piecewise function is used to convert the real-time temperature into the corresponding temperature impact coefficient, and the influence degree of the temperature of the storage cabinet on the storage of the goods is quantified according to the temperature impact coefficient, so as to provide data support for the temperature control system of the intelligent storage cabinet, so as to automatically adjust the temperature in the cabinet according to the temperature impact coefficient, and provide data support for subsequent generation of an intelligent storage scheme.

[0096] For example, the piecewise function is defined as: when the real-time temperature is lower than 10℃, the temperature impact coefficient = 0.1; when the real-time temperature is in the interval [10, 25]℃, the temperature impact coefficient = 0.1+(temperature-10)*0.013; and when the real-time temperature is higher than 25℃, the temperature impact coefficient = 0.3+(temperature-25)*0.02.

[0097] Step S22-2, the humidity impact coefficient is obtained.

[0098] Specifically, a piecewise function is used to convert the real-time humidity into the corresponding humidity impact coefficient, and the influence degree of the humidity of the storage cabinet on the storage of the goods is quantified according to the humidity impact coefficient, so as to provide data support for subsequent generation of an intelligent storage scheme.

[0099] For example, the piecewise function is defined as: when the real-time humidity is lower than 40%, the humidity impact coefficient = 0.1; when the real-time humidity is in the interval [40, 60]%, the humidity impact coefficient = 0.1+(humidity-40)*0.005; and when the real-time humidity is higher than 60%, the humidity impact coefficient = 0.2+(humidity-60)*0.01.

[0100] Step S22-3, the regional peak human flow is obtained.

[0101] Specifically, the maximum value method of the sliding window is adopted to count the maximum passenger flow of the region where the storage cabinet deployment point is located in a day to generate a regional passenger flow peak value. The regional passenger flow peak value is used to quantify the busy degree of the storage cabinet, to provide a decision basis for resource allocation and energy consumption management of the intelligent storage cabinet, so as to improve the user experience and equipment operation efficiency, such as dynamically adjusting the number of storage compartments of the storage cabinet in a certain period or reducing the equipment operation power in the low peak period. For example, with a window size of 1 hour, the maximum passenger flow of a certain region in a day occurs in the period from 18:30 to 19:30, and the maximum passenger flow is 500. Therefore, the corresponding regional passenger flow peak value is 500.

[0102] In step S23, the attribute data of the articles is classified and feature extraction is performed to obtain article attribute indexes.

[0103] Specifically, the article attribute indexes at least include a perishable article proportion, a special storage demand article quantity, and an article average volume.

[0104] In this embodiment, step S23 includes:

[0105] In step S23-1, the perishable article proportion is obtained.

[0106] Specifically, the perishable article proportion is obtained according to a preset perishable article list and a statistic of the perishable article proportion in all storage articles, such as the proportion of fresh food, medicine and other articles in the total amount of storage articles. The perishable article proportion is used to represent the constituent ratio of perishable articles in the storage cabinet, which can assist the manager to optimize the allocation of temperature control equipment resources, thereby reducing the risk of storage article loss.

[0107] In step S23-2, the special storage demand article quantity is obtained.

[0108] Specifically, the special storage demand article quantity is obtained by counting the quantity of articles that require special storage conditions. The special storage demand article quantity can be used to evaluate the bearing pressure of special storage equipment, and based on the evaluation result, the maintenance of the equipment and the storage layout adjustment of the storage cabinet can be assisted to plan, to ensure the safety of article storage.

[0109] In step S23-3, the article average volume is obtained.

[0110] Specifically, the article average volume is obtained by calculating the average volume of all storage articles. Based on the article average volume, the space utilization rate of the storage cabinet can be evaluated, and data support can be provided for optimizing the storage compartment size design and formulating the article placement strategy, so as to improve the storage space utilization efficiency of the storage cabinet and avoid the situation that the article storage is difficult due to insufficient estimation of the article size.

[0111] In step S24, state feature extraction is performed on the real-time cabinet state data to obtain storage cabinet operation indexes.

[0112] Specifically, the storage cabinet operation index at least includes a storage compartment turnover rate, a fault repair efficiency, and a function utilization rate.

[0113] In this embodiment, step S24 includes:

[0114] Step S24-1, obtaining the storage compartment turnover rate.

[0115] In a possible embodiment, it is assumed that each storage compartment can be used only once per hour, and the quotient of the actual use frequency of the storage compartment in a day and the theoretical use number of the storage compartment is taken as the storage compartment turnover rate. The use efficiency of the intelligent storage cabinet is quantified by the storage compartment turnover rate, and data basis is provided for subsequent regional balanced load calculation. For example, a certain intelligent storage cabinet is provided with 50 storage compartments, and the storage cabinet is used 600 times in 24 consecutive hours. The corresponding storage compartment turnover rate is 600 / (50*24)*100% = 50%.

[0116] Step S24-2, obtaining the fault repair efficiency.

[0117] Specifically, the reciprocal of the average time required from the occurrence of a fault to the completion of fault repair is taken as the value of the fault repair efficiency.

[0118] Step S24-3, obtaining the function utilization rate.

[0119] In a possible embodiment, it is assumed that each storage compartment can be used only once per hour, and the quotient of the actual use frequency of the specific function storage compartment in a day and the theoretical use number of the specific function storage compartment is taken as the function utilization rate. The use efficiency of the specific function storage compartment in the intelligent storage cabinet is quantified by the function utilization rate, and data basis is provided for subsequent regional balanced load calculation. For example, a certain intelligent storage cabinet is equipped with 20 refrigerated storage compartments. It is assumed that the refrigerated storage compartments are used 24 times in 24 consecutive hours. The refrigeration function utilization rate is 24 / (20*24)*100% = 5% (here, only the function utilization rate is illustrated, and the specific function utilization rate needs to be determined according to the actual situation).

[0120] Step S25, performing feature fusion on the user behavior regularity index, the environmental influence index, the article attribute index, and the storage cabinet operation index to generate an intelligent storage cabinet evaluation index set.

[0121] Specifically, the principal component analysis algorithm is used for feature fusion. The user behavior regularity index, the environmental influence index, the article attribute index, and the storage cabinet operation index are standardized. The covariance matrix of the sample data is constructed, and the corresponding principal component vector group is obtained by eigenvalue decomposition. The principal component vector with a cumulative contribution rate not less than 95% in the principal component vector group is taken as the intelligent storage cabinet evaluation index set.

[0122] Step S3, constructing a storage cabinet intelligent control model based on the storage cabinet intelligent control mechanism.

[0123] In this embodiment, step S3 includes:

[0124] Step S31, constructing a regional load balancing module of the storage cabinet intelligent control model based on a regional load balancing function contained in the storage cabinet intelligent control mechanism.

[0125] In this embodiment, step S31 includes:

[0126] Step S31-1, obtaining a regional load coefficient.

[0127] Specifically, the regional load balancing function is used to obtain the regional load coefficient in combination with the environmental impact indicators and the storage cabinet operation indicators contained in the intelligent storage cabinet evaluation indicator set, wherein the regional load balancing function is used to comprehensively evaluate and quantify the operation load state of the intelligent storage cabinet according to the storage compartment utilization efficiency, equipment maintenance status, functional application degree, environmental impact factors, and regional passenger flow; and the regional load coefficient can be used to reflect the overall pressure level of the storage cabinet in actual operation.

[0128] In one possible embodiment, the regional load balancing function can be expressed as "regional load coefficient = (storage compartment turnover rate / average turnover rate reference value)*0.4 + (1-fault repair efficiency)*0.3 + (functional utilization rate)*0.2 + (temperature influence coefficient)*0.05 + (humidity influence coefficient)*0.03 + (current passenger flow / peak regional passenger flow)*0.02".

[0129] For example, the storage compartment turnover rate of the storage cabinet deployment point A is 60%, the average turnover rate reference value is 80%, the fault repair efficiency is 100%, the functional utilization rate is 70%, the temperature influence coefficient is 0.9, the humidity influence coefficient is 0.8, the current passenger flow is 80, and the peak regional passenger flow is 500. According to the regional load balancing function, the regional load coefficient corresponding to the storage cabinet deployment point A is (0.6 / 0.8)*0.4 + (1-1)*0.3 + 0.7*0.2 + 0.9*0.05 + 0.8*0.03 + (80 / 500)*0.02 = 0.5122≈0.51; the storage compartment turnover rate of the storage cabinet deployment point B is 75%, the fault repair efficiency is 80%, the functional utilization rate is 100%, the temperature influence coefficient is 0.8, the humidity influence coefficient is 0.7, and the current passenger flow is 150. The regional load coefficient of the storage cabinet deployment point B is (0.75 / 0.8)*0.4 + (1-0.8)*0.3 + 1*0.2 + 0.8*0.05 + 0.7*0.03 + (150 / 500)*0.02 = 0.702≈0.70.

[0130] Step S31-2, threshold correction is performed on the threshold corresponding to the regional level division based on the threshold correction factor obtained by the reinforcement learning-based contextual multi-arm bandit algorithm.

[0131] Specifically, the aggressive balancing strategy, the conservative stability strategy and the predictive strategy are defined as arms to be selected in the contextual multi-arm bandit algorithm, a context feature vector is constructed based on the environmental impact index and the storage cabinet operation index, and the context feature vector can be represented as {current period, temperature influence coefficient, humidity influence coefficient, storage compartment turnover rate, fault repair efficiency, load change rate in the past 7 days}; the optimal arm selection is performed on the arms corresponding to the aggressive balancing strategy, the conservative stability strategy and the predictive strategy according to the LinUCB algorithm and the preset reward function, the threshold correction factor is generated according to the obtained optimal arm; and the threshold corresponding to the regional level division is corrected based on the threshold correction factor.

[0132] In a possible embodiment, the aggressive balancing strategy, the conservative stability strategy and the predictive strategy are defined as arm A, arm B and arm C respectively, wherein the aggressive balancing strategy represents reducing the division threshold of the high-load region in the regional level division, thereby maximizing the load balancing; for example, the preset division threshold of the high-load region is 0.7, and the aggressive balancing strategy reduces the division threshold of the high-load region to 0.6, so as to ensure that a large number of storage cabinets are distributed in the high-load region;

[0133] The conservative stability strategy represents maintaining the steady-state operation of the intelligent storage cabinet and reducing the fluctuation risk caused by strategy adjustment; for example, the division threshold of the high-load region is increased to 0.8.

[0134] The predictive strategy represents that a complete period of 24 hours is taken as a complete period, the load change rate is obtained according to the calculation form of “[(regional load coefficient of the current period-regional load coefficient of the last period) / regional load coefficient of the last period]”, and the threshold corresponding to the regional level division is adjusted based on the load change rate; for example, when the load change rate exceeds 0.2, the threshold corresponding to the regional level division is simultaneously reduced by 0.05, and when the load change rate is lower than -0.2, the threshold corresponding to the regional level division is simultaneously increased by 0.05.

[0135] It is assumed that the reward function is (average turnover rate 1 hour after implementing the strategy-average turnover rate 1 hour before implementing the strategy)-load variance 1 hour after implementing the strategy, wherein the average turnover rate represents the average value of the storage compartment turnover rates of each storage cabinet deployment point, and the load variance represents the sum of squares of the difference between the regional load coefficient of each storage cabinet deployment point and the average regional load coefficient / the number of storage cabinet deployment points.

[0136] According to the reward function, the optimal arm is selected, and according to the calculation form of "(the load threshold of the optimal arm - the initial load threshold) / the initial load threshold", the threshold correction factor is obtained, wherein the threshold correction factor only retains the last two digits after the decimal point, for example, the optimal arm of a certain threshold correction is arm C, the division threshold of the high load area of arm C is 0.65, and the initial division threshold of the high load area is 0.7, so the threshold correction factor corresponding to the division threshold of the high load area is -0.07, and the division threshold of the high load area after correction based on the threshold correction factor is 0.63.

[0137] Step S31-3, the distribution area of the storage cabinet is divided into area levels.

[0138] In a possible embodiment, the distribution area of the storage cabinet is divided into area levels based on the area load coefficient, assuming that after threshold correction, the threshold corresponding to the area level division is that the distribution area with an area load coefficient exceeding 0.7 is divided into a high load area, the distribution area with an area load coefficient in the interval [0.3, 0.7] is divided into a medium load area, and the distribution area with an area load coefficient lower than 0.3 is divided into a low load area.

[0139] The area balanced load module performs user drainage of the high load area storage cabinet deployment point according to the area level, obtains the moving cost of the user to the medium and low load area storage cabinet deployment points according to a preset moving cost function, such as moving cost = moving distance * 0.5 + moving time * 0.5, preferentially recommends the low load area storage cabinet deployment point with the lowest moving cost to the user, and obtains the dynamic discount compensation according to a preset discount compensation function, such as dynamic discount compensation = (area load coefficient of high load area - area load coefficient of drainage area) * 0.25; for the medium load area storage cabinet deployment point, the basic pricing is maintained.

[0140] For example, assuming that the moving cost of a user to storage cabinet deployment point A and storage cabinet deployment point B is consistent, the area load coefficients of storage cabinet deployment point A and storage cabinet deployment point B are 0.51 and 0.70 respectively, and storage cabinet deployment point A and storage cabinet deployment point B are determined as medium load area and high load area respectively, then storage cabinet deployment point A is preferentially recommended to the user, and the corresponding dynamic discount compensation is calculated as (0.702-0.5122) * 0.25 ≈ 5%.

[0141] Step S32, a user storage time limit allocation module of the storage cabinet intelligent control model is constructed based on a credit time limit association function contained in the storage cabinet intelligent control mechanism.

[0142] In this embodiment, step S32 includes:

[0143] Step S32-1, a user credit score is obtained.

[0144] Specifically, based on the credit time correlation function and combined with the user behavior regularity index contained in the intelligent storage cabinet evaluation index set, the user credit score is obtained.

[0145] In a possible embodiment, the on-time pickup rate is obtained according to the calculation method of "on-time pickup rate = number of times of non-time-out pickup in 30 days / total number of pickup times in 30 days", wherein the judgment rule of time-out pickup is that the pickup time-out duration is greater than a threshold in the time-out duration distribution, such as more than 30 minutes being regarded as time-out; and the payment on-time rate is obtained according to the calculation method of "payment on-time rate = number of times of on-time payment in 30 days / total number of payment times in 30 days", wherein the judgment rule of on-time payment is that the payment operation is completed within a specified time, such as a preset payment time limit being 15 minutes, and completion of payment within 15 minutes is recorded as one on-time payment.

[0146] The weight is adjusted in combination with the daily access frequency data in the user behavior regularity index, for example, for a user with a daily access frequency data of no less than 2 times, the weight of the on-time pickup rate corresponding thereto is increased from 0.4 to 0.5, and the weight of the payment on-time rate is reduced from 0.3 to 0.2.

[0147] The number of times of irregular use of the storage cabinet in 30 days is defined as the number of irregular times, and it is stipulated that 10 points of user credit score are deducted for each irregular use, such as the behaviors of time-out pickup and non-renewal of storage and malicious damage to the storage cabinet.

[0148] The user credit score is obtained according to the calculation method of "user credit score = on-time pickup rate*0.4*100+payment on-time rate*0.3*100-irregular times*10", wherein the on-time pickup rate and the payment on-time rate of the user reflect the user's performance of reliability and transaction integrity, and the on-time pickup rate and the payment on-time rate are regarded as positive incentives, so as to promote the user to complete the pickup and payment on time; the number of irregular times reflects the frequency of the user's violation of the use rules, and 10 points of negative punishment are given for each irregular time, so as to evaluate the user's credit performance in the use of the intelligent storage cabinet.

[0149] Step S32-2, grade division and basic storage time limit allocation are performed on the user stored articles.

[0150] Specifically, based on the article attribute index contained in the intelligent storage cabinet evaluation index set, the user stored articles are divided into three storage grades of short-term storage, standard storage and long-term storage, and the corresponding basic storage time limit is allocated to each storage grade.

[0151] In a possible embodiment, based on the item attribute index and in combination with the K-means clustering algorithm, the storage time of the stored items of the user is clustered into three storage levels, i.e., short-term storage, standard storage, and long-term storage, and a corresponding basic storage time limit is assigned to each storage level, for example, the basic storage time limit of the short-term storage is 12 hours, the basic storage time limit of the standard storage is 48 hours, and the basic storage time limit of the long-term storage is 72 hours.

[0152] In step S32-3, the basic storage time limit is adjusted according to the user credit score to obtain an initial storage time limit.

[0153] In a possible embodiment, for a user whose credit score is higher than 80, the basic storage time limit of the corresponding storage level is extended by 20%, for a user whose credit score is lower than 60, the basic storage time limit is shortened by 10%, and for a user whose credit score is in the interval [60, 80], the basic storage time limit remains unchanged.

[0154] In step S32-4, a Bayesian optimization is introduced to nonlinearly calibrate the initial storage time limit.

[0155] In a possible embodiment, a time limit tuning function is constructed, a preset scaling factor is taken as an input of the time limit tuning function, for example, the scaling factor is only selected in the interval [0.7, 1.3], a user actual overtime rate is taken as an output, the pickup efficiency corresponding to different scaling factors is evaluated based on the user actual overtime rate, the optimal scaling factor is obtained by taking the minimization of the actual overtime rate as a target, the initial storage time limit is nonlinearly calibrated according to the optimal scaling factor, and the overtime pickup behavior in the Bayesian optimization process is defined as a pickup behavior whose pickup overtime time exceeds a value obtained by multiplying the initial storage time limit by the scaling factor.

[0156] The user credit score is divided into a plurality of user credit score intervals, for example, into three user credit score intervals [0, 60), [60, 80), and [80, 100], the stored items are divided into a plurality of item attribute combinations based on the item attribute index, for example, into four item attribute combinations {perishable and special storage requirement}, {perishable}, {special storage requirement}, and {ordinary item}, the user is divided into a plurality of user types according to the plurality of user credit score intervals and the item attribute combinations, for example, into 12 user types according to three user credit score intervals and four item attribute combinations, and the plurality of divided user types are optimized separately.

[0157] The proxy model fitting is performed based on a Gaussian process kernel function to reduce the actual timeout rate of the user as an optimization target. For example, the Gaussian process kernel function selects a square exponential kernel for proxy simulation fitting. The specific proxy simulation process can be represented as follows: 5 values of the scaling factor are randomly selected for each user type as initial sampling points, such as 0.7, 0.9, 1.0, 1.1, and 1.3, and the user type data in which the perishable goods account for 100% is preferentially sampled; the proxy simulation fitting is iterated through an expected improvement function to obtain the optimal scaling factor, wherein the iteration convergence condition can be represented as follows: the variation amplitude of the optimal scaling factor obtained by continuous 3 iterations is lower than a preset variation amplitude threshold, and the proxy simulation fitting is re-performed every 15 accumulated new data points, and the new data points represent a real-time collected multi-dimensional intelligent storage cabinet data set.

[0158] The initial storage time limit is nonlinearly calibrated according to the optimal scaling factor. For example, the user credit score of a certain user is 62, and the stored goods are fresh pork, which is a perishable good. The corresponding goods attribute combination is that the perishable goods account for 100% and have cold storage needs, and it is determined as short-term storage level. The corresponding basic storage time limit is 12 hours. Since the user credit score of the user is located in the [60, 80] interval, the basic storage time limit is not adjusted. Assuming that the optimal scaling factor obtained by the user after Bayesian optimization is 0.95, the calibrated user storage time limit is 12*0.95=11.4 hours.

[0159] In step S33, a forgetfulness prevention intervention module of the intelligent storage cabinet control model is constructed based on a forgetfulness risk function included in the intelligent storage cabinet control mechanism.

[0160] In this embodiment, step S33 includes:

[0161] In step S33-1, a forgetfulness risk coefficient is obtained.

[0162] Specifically, the forgetfulness risk coefficient is obtained based on the forgetfulness risk function and in combination with a user behavior regularity index included in the intelligent storage cabinet evaluation index set.

[0163] In one possible embodiment, the forgetfulness risk function can be represented as “forgetfulness risk coefficient=(historical forgetfulness rate*0.5)+(cabinet leaving speed deviation rate*0.3)+(intervention non-response proportion*0.2)”.

[0164] Based on the pickup timeout duration distribution data in the user behavior regularity index, the operation of being overtime for more than 30 minutes and not being continued is defined as the forgetting behavior, the historical forgetting rate is obtained according to the calculation form of "historical forgetting rate = number of forgotten pickups in 60 days / total number of pickups in 60 days", when the total number of pickups in 60 days is less than 5 times, the total number of pickups in 60 days is calculated as 5 times, and the insufficient number is counted as not forgetting, for example, if the total number of pickups of a user in 60 days is 3 times and the number of forgotten pickups in 60 days is 1 time, the historical forgetting rate of the user is 1 / 5 = 0.2;

[0165] The off-cabinet speed deviation rate is obtained according to the calculation form of "off-cabinet speed deviation rate = |current off-cabinet speed- average off-cabinet speed| / average off-cabinet speed";

[0166] The intervention non-response proportion is obtained according to the calculation form of "intervention non-response proportion = number of non-response reminders in 30 days / total number of reminders";

[0167] According to the storage cabinet operation index, the turnover rate of the storage compartment is adjusted, for example, for the storage cabinet deployment point with a storage compartment turnover rate greater than 0.7, the historical forgetting rate weight is increased from 0.5 to 0.6, the off-cabinet speed deviation rate weight is reduced from 0.3 to 0.2, and the intervention non-response proportion is reduced from 0.2 to 0.1.

[0168] Step S33-2, implementing a hierarchical intervention strategy based on the forgetting risk coefficient.

[0169] When the forgetting risk coefficient is less than 0.2, only the data corresponding to the current user is recorded;

[0170] When the forgetting risk coefficient is in [0.2, 0.5], a pickup reminder information is sent to the user before the user storage time limit is reached;

[0171] When the forgetting risk coefficient is in (0.5, 0.8] and the pickup is not completed before the user storage time limit is reached, the goods are automatically continued to be stored and the continued storage information is pushed to the user terminal at the same time;

[0172] When the forgetting risk coefficient is more than 0.8 and the pickup is not completed before the user storage time limit is reached, a delivery scheme is generated and sent to the user terminal.

[0173] Step S33-3, correcting the hierarchical intervention strategy based on machine learning.

[0174] In a possible embodiment, a user feature vector X is generated based on a user behavior regularity index, an environmental impact index, and an item attribute index, such as {daily average access frequency data, peak access frequency data, pickup timeout duration distribution data, regional peak passenger flow, and average item volume}; assuming that the hierarchical intervention strategy only includes three intervention modes of APP reminder, SMS reminder, and telephone reminder, the intervention modes are defined as processing variables T of the random forest, and {APP reminder}, {APP reminder and SMS reminder}, and {telephone reminder} correspond to T=0, 1, and 2, respectively; and {user timely pickup} and {user timeout without pickup} are defined as result variables Y of the random forest, and correspond to Y=1 and 0, respectively.

[0175] The fitting of the three conditional expectations E[Y|X,T=0], E[Y|X,T=1], and E[Y|X,T=2] is performed by taking the user feature vector X and the processing variable T as inputs and the result variable Y as output, to learn the complex nonlinear relationship in the multi-dimensional intelligent storage cabinet dataset by using the random forest;

[0176] The residual values of the actual values and the predicted values of the result variable Y, the residual value when T=1, and the residual value when T=2 are obtained, and linear regression is used to fit the values to obtain the promotion effect L1 of telephone reminder compared with only APP reminder and the promotion effect L2 of SMS reminder compared with only APP reminder, for example, the residual value when T=1 can be calculated by I(T=1)-P(T=1|X), where I(T=1) is an indicator function, that is, only when T=1, the value is 1, otherwise 0;

[0177] The hierarchical intervention strategy is corrected according to the values of L1 and L2 corresponding to the user, for example, if L1>L2+0.100, telephone reminder is directly used regardless of the initial level; if L1 and L2 are both lower than 0.03, only APP reminder is used.

[0178] Step S4, real-time user demand data is obtained, and demand analysis is performed based on the intelligent storage cabinet evaluation index set.

[0179] In this embodiment, step S4 includes:

[0180] Step S41, real-time user demand data is obtained.

[0181] Specifically, real-time user demand data is obtained according to the submission information of the user end, and the real-time user demand data at least includes access time expectation data, item type data, target cabinet point range data, and service demand data.

[0182] It can be understood that the access time expected data represents the start and end time of the user storing the item; the item type data represents the item type stored by the user, such as fresh food, medicine; the target cabinet point range data represents the straight-line distance of the user from the target cabinet point; and the service demand data represents the special storage demand required by the user to store the item, such as frozen storage and insurance storage.

[0183] In step S42, demand analysis is performed based on the intelligent storage cabinet evaluation index set.

[0184] In this embodiment, step S42 includes:

[0185] In step S42-1, access time period demand analysis is performed to obtain an access time period demand analysis result.

[0186] Specifically, the access time expected data is subjected to access time period demand analysis based on the user behavior rule index included in the intelligent storage cabinet evaluation index set, and an access time period demand analysis result is obtained.

[0187] In one possible embodiment, the peak period in the access time expected data is predicted using a long short-term memory network, and seasonal time series analysis is combined to query the historical storage compartment turnover rate of the target storage cabinet deployment point corresponding to the period of the access time expected data, such as using the weighted average value of the last 30 days as the historical storage compartment turnover rate.

[0188] According to the "period matching degree = 1-(m*user storage target cabinet point corresponding period historical storage compartment turnover rate-n*user storage target cabinet point area average turnover rate) / (n*user storage target cabinet point area average turnover rate)", wherein m and n are dynamically adjusted according to the environmental impact index through reinforcement learning; if the result is negative, take 0, and at the same time set a high turnover rate threshold, exceeding which is considered as a serious period conflict, for example, a user stores an item in cabinet A at 2 pm on the weekend, the area average turnover rate of cabinet A is 0.3, the historical storage compartment turnover rate of cabinet A at 2 pm on the weekend is 0.8, and assuming that m and n of the area corresponding to cabinet A are 0.6 and 0.4 respectively, then the corresponding period matching degree is 1-(0.6*0.8-0.4*0.3) / (0.4*0.3)=-2, since the period matching degree is negative, it indicates that a serious period conflict occurs.

[0189] In step S42-2, storage service demand analysis is performed to obtain a storage service demand analysis result.

[0190] Specifically, the item type data, the target cabinet point range data, and the service demand data are subjected to storage service demand analysis based on the environmental impact index, the item attribute index, and the storage cabinet operation index included in the intelligent storage cabinet evaluation index set, and a storage service demand analysis result is obtained.

[0191] In a possible embodiment, the service comprehensive score is obtained according to a calculation manner of "service comprehensive score = function matching degree score * 2 * 0.5 * environment adaptation degree score * 0.3 + reliability score * 0.2";

[0192] The function matching degree is represented as whether the user storage target cabinet point meets the service demand data of the stored article, for example, whether there is an idle refrigeration compartment, if yes, the function matching degree score is set to 1, otherwise, the function matching degree score is set to 0, and if the function utilization rate of the target cabinet point is less than 0.5, the function matching degree score is increased by 0.2.

[0193] The environment adaptation degree is obtained according to a calculation manner of "temperature influence coefficient * 0.6 + humidity influence coefficient * 0.4", the higher the environment adaptation degree value is, the greater the influence of the environment is, if the environment adaptation degree is less than 0.3, the environment adaptation degree score is set to 1, if the environment adaptation degree is located in [0.3, 0.5], the environment adaptation degree score is set to 0.5, and if the environment adaptation degree exceeds 0.5, the environment adaptation degree score is set to 0.

[0194] The reliability score of the cabinet point with the fault repair efficiency greater than a preset threshold is set to 1, otherwise, it is set to 0.

[0195] The space distance score of the user is obtained according to the target cabinet point range data, for example, if the straight line distance from the user to the target cabinet point is 500 meters, it is converted into 0.2 after normalization, assuming that there is a linear mapping relationship between the normalized value of the straight line distance from the user to the target cabinet point and the space distance score, and the mapping ratio is 1:1, then the corresponding space distance score is 0.2.

[0196] In step S42-3, the demand matching degree is obtained based on the access time period demand analysis result and the storage service demand analysis result.

[0197] Specifically, the access time period demand analysis result and the storage service demand analysis result are fused according to a demand matching degree function to generate the demand matching degree, which is used to quantify the matching degree between the user demand and the storage cabinet deployment point.

[0198] For example, the time period matching degree is 0.9, the service comprehensive score is 1, and the space distance score is 0.67, assuming that the demand matching degree is calculated as "time period matching degree * 0.4 + service comprehensive score * 0.4 + space distance score * 0.2", then the corresponding demand matching degree is 0.89 (here, it is only a simple example of demand matching degree calculation, and the calculation result is only kept to two decimal places, and the specific determination needs to be made according to the actual situation).

[0199] The storage cabinet deployment points with the demand matching degree exceeding 0.7 are divided into high matching degree storage points, and a priority recommendation list is generated based on the high matching degree storage points.

[0200] the storage cabinet deployment points with the demand matching degree in [0.5, 0.7] are divided into standard matching degree storage points, and an alternative recommendation list is generated based on the standard matching degree storage points;

[0201] the storage cabinet deployment points with the demand matching degree lower than 0.5 are divided into low matching degree storage points, and no recommendation is made.

[0202] In step S5, the demand analysis result is obtained, and an intelligent storage scheme is generated in combination with the intelligent storage cabinet control model.

[0203] Specifically, based on the priority recommendation list and the alternative recommendation list contained in the demand analysis result, the load optimal storage point is obtained by calling the regional balanced load module of the intelligent storage cabinet control model.

[0204] The user storage time limit is allocated according to the user storage time limit allocation module of the intelligent storage cabinet control model.

[0205] The anti-forgetting information is generated according to the anti-forgetting intervention module of the intelligent storage cabinet control model.

[0206] The load optimal storage point, the user storage time limit and the anti-forgetting information are encapsulated as an intelligent storage scheme, and are sent to the user end.

[0207] For example, part of the intelligent storage scheme pushed to a user is: recommended cabinet point: {storage cabinet A, ID: A007, address: XX Road XX No.}; recommended storage compartment: {ID: A-05, storage service: refrigeration, size: 25cm*20cm*15cm}; storage time: {start time: 18:00, end time: next day 8:24}, anti-forgetting information: {next day 7:24 APP reminder}.

[0208] The embodiment of the application provides a multifunctional control system of an AI intelligent storage cabinet, which is suitable for intelligent control and comprises:

[0209] A data acquisition module is configured to acquire user behavior data, environment data, article attribute data and real-time cabinet state data, and generate a multi-dimensional intelligent storage cabinet data set.

[0210] An index generation module is configured to perform feature extraction and fusion on the multi-dimensional intelligent storage cabinet data set, and generate an intelligent storage cabinet evaluation index set.

[0211] A model construction module is configured to construct an intelligent storage cabinet control model according to an intelligent storage cabinet control mechanism.

[0212] The demand analysis module is used for acquiring real-time user demand data and performing demand analysis on the user demand data based on a smart storage cabinet evaluation index set;

[0213] The scheme generation module is used for generating a smart storage scheme according to the demand analysis result and a storage cabinet intelligent control model.

[0214] The specific use mode and role of the embodiment are as follows:

[0215] Firstly, by acquiring user behavior data, environment data, article attribute data and real-time cabinet state data, a multi-dimensional smart storage cabinet data set is generated, feature extraction and fusion are performed on the multi-dimensional smart storage cabinet data set, a smart storage cabinet evaluation index set is generated, thereby a comprehensive and scientific smart storage cabinet evaluation mechanism is generated, and data support is provided for subsequent data analysis and scheme generation;

[0216] Then, based on the storage cabinet intelligent control mechanism, a storage cabinet intelligent control model is constructed, by the in-area balanced load module, the user storage time limit allocation module and the anti-forgetting intervention module in the constructed storage cabinet intelligent control model, dynamic balance of regional load, accurate matching of storage time limit and accurate implementation of personalized anti-forgetting intervention are realized.

[0217] Finally, real-time user demand data is acquired, demand analysis is performed based on the smart storage cabinet evaluation index set, a demand analysis result is acquired, and a smart storage scheme is generated in combination with the storage cabinet intelligent control model, thereby accurate analysis of user demand is realized, a smart storage scheme meeting the demand of the user is provided, and the resource utilization rate and user experience of the smart storage cabinet are improved.

[0218] In addition, the embodiment of the present application further provides an electronic device, which comprises at least one processor and a memory in communication connection with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method in the above embodiment.

[0219] The various constituent components of the electronic device will be specifically introduced as follows:

[0220] The processor is the control center of the electronic device, and can be one processor or a combination of multiple processing elements. For example, the processor is one or more central processing units (CPUs), application specific integrated circuits (ASICs), or one or more integrated circuits configured to implement one or more embodiments of the present application, such as one or more digital signal processors (DSPs), or one or more field programmable gate arrays (FPGAs).

[0221] The processor can execute various functions of the electronic device by running or executing software programs stored in the memory and calling data stored in the memory.

[0222] The memory is used to store software programs for implementing the embodiments of the present application, and is controlled by the processor to execute. The specific implementation manner can refer to the above method embodiments, and will not be described here.

[0223] The memory can be a real-only memory (ROM) or other type of static storage device that can store static information and instructions, a random access memory (RAM) or other type of dynamic storage device that can store information and instructions, an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only (CD-ROM), or other optical disk storage, optical disk storage (including compact disks, laser disks, optical disks, digital versatile disks, Blu-ray disks, etc.), magnetic disk storage medium, or other magnetic storage device, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and capable of being accessed by a computer, but is not limited thereto. The memory can be integrated with the processor or exist independently and coupled to the processor through the interface circuit of the electronic device, and the embodiments of the present application do not make a specific limitation in this regard.

[0224] The above-described embodiments can be implemented in whole or in part by software, hardware (such as a circuit), firmware, or any combination thereof. When implemented in software, the above-described embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are wholly or partially generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transferred from one website, computer, server or data center to another website, computer, server or data center by limited (for example, infrared, wireless, microwave, etc.) mode. The computer-readable storage medium can be any available medium accessible by a computer or a data storage device such as a server, data center, etc. containing one or more available medium collections. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid state disk.

[0225] It should be understood that the term "and / or" herein merely describes the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B, which can represent the existence of A alone, the existence of A and B, and the existence of B alone, where A and B can be singular or plural. In addition, the character " / " herein generally represents that the associated objects before and after are an "or" relationship, but can also represent an "and / or" relationship, which can be understood according to the context before and after.

[0226] It should be understood that in the embodiments of the present application, the size of the sequence number of the above-mentioned processes does not mean the order of execution, and the execution order of the processes should be determined according to their functions and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0227] The above-described embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A multi-functional control method for an AI intelligent storage cabinet, characterized in that, The method includes: Acquire user behavior data, environmental data, item attribute data, and real-time cabinet status data to generate a multi-dimensional intelligent storage cabinet dataset; Feature extraction and fusion are performed on the multidimensional intelligent storage cabinet dataset to generate an intelligent storage cabinet evaluation index set; Based on the intelligent control mechanism of the storage cabinet, an intelligent control model for the storage cabinet is constructed. The intelligent control model for the storage cabinet includes a regional load balancing module, a user storage time limit allocation module, and an anti-forgetting intervention module. The regional load balancing module of the intelligent control model of the storage cabinet is constructed based on the regional load balancing function contained in the intelligent control mechanism of the storage cabinet. The user storage time limit allocation module of the intelligent control model of the storage cabinet is constructed based on the credit timeliness correlation function contained in the intelligent control mechanism of the storage cabinet. An anti-forgetting intervention module is constructed based on the forgetting risk function contained in the intelligent control mechanism of the storage cabinet to build an intelligent control model for the storage cabinet; Based on the output results of the regional load balancing module, the user storage time limit allocation module, and the anti-forgetting intervention module, an intelligent storage solution is generated. Acquire real-time user demand data and perform demand analysis based on the intelligent storage cabinet evaluation index set; Obtain the results of the requirements analysis and generate an intelligent storage solution by combining them with the intelligent control model of the storage cabinet.

2. The multi-functional control method for an AI intelligent storage cabinet according to claim 1, characterized in that, Feature extraction and fusion are performed on the multidimensional intelligent storage cabinet dataset to generate an intelligent storage cabinet evaluation index set, including: Extract time-dimensional features from user behavior data to obtain user behavior pattern indicators, which include at least the average daily number of accesses, access frequency during peak hours, and the distribution of pickup timeout duration. Statistical dimensional feature extraction is performed on environmental data to obtain environmental impact indicators, which include at least temperature impact coefficient, humidity impact coefficient, and peak regional population flow. The item attribute data is classified and feature extracted to obtain item attribute indicators, which include at least the proportion of perishable items, the number of items with special storage requirements, and the average volume of items. Real-time cabinet status data is used to extract status features and obtain storage cabinet operation indicators, which include at least storage compartment turnover rate, fault repair efficiency and functional utilization rate. By integrating user behavior patterns, environmental impact indicators, item attribute indicators, and storage cabinet operation indicators, a set of evaluation indicators for intelligent storage cabinets is generated.

3. The multi-functional control method for an AI intelligent storage cabinet according to claim 1, characterized in that, The regional load balancing module of the intelligent control model for storage cabinets, based on the regional load balancing function included in the intelligent control mechanism of the storage cabinet, includes: Based on the regional load balancing function and combined with the environmental impact indicators and storage cabinet operation indicators included in the intelligent storage cabinet evaluation index set, the regional load coefficient is obtained. The method of dividing the distribution area of ​​storage cabinets into regional levels is defined as multiple regional level division strategies, and the multiple regional level division strategies are used as candidates in the context multi-armed gambling machine algorithm. The context feature vector of the candidate arm is constructed based on the environmental impact indicators and storage cabinet operation indicators contained in the intelligent storage cabinet evaluation index set. The candidate arms are selected for optimal arm screening based on the LinUCB algorithm and a preset reward function. A threshold correction factor is generated based on the obtained optimal arm. The threshold corresponding to the regional level division is corrected based on the threshold correction factor. The distribution areas of the storage cabinets are classified into different regional levels based on the regional load coefficient.

4. The multi-functional control method for an AI intelligent storage cabinet according to claim 1, characterized in that, The user storage time limit allocation module, which is built based on the credit timeliness correlation function contained in the intelligent control mechanism of the storage cabinet, includes: Based on the credit timeliness correlation function and combined with the user behavior pattern indicators contained in the smart storage cabinet evaluation indicator set, a user credit score is obtained. Based on the item attribute indicators included in the evaluation index set of smart storage cabinets, the items stored by users are divided into three storage levels: short-term storage, standard storage, and long-term storage, and a corresponding basic storage time limit is assigned to each storage level. The basic storage time limit is adjusted based on the user's credit score to generate the initial storage time limit; Construct a time limit tuning function, take a preset scaling factor as the input of the time limit tuning function, take the actual timeout rate of the user as the output, evaluate the retrieval efficiency corresponding to different scaling factors based on the actual timeout rate of the user, obtain the optimal scaling factor based on the Gaussian process kernel function with the goal of minimizing the actual timeout rate, and perform nonlinear calibration on the initial storage time limit according to the optimal scaling factor. The actual user timeout rate is expressed as the ratio of the number of times a user timed out to the total number of pickups. Timeout pickup behavior is defined as pickup behavior where the pickup timeout exceeds the product of the initial storage time limit and the scaling factor.

5. The multi-functional control method for an AI intelligent storage cabinet according to claim 1, characterized in that, The anti-forgetting intervention module, which constructs the intelligent control model of the storage cabinet based on the forgetting risk function contained in the intelligent control mechanism of the storage cabinet, includes: Based on the forgetting risk function and combined with the user behavior pattern indicators contained in the evaluation index set of the smart storage cabinet, the forgetting risk coefficient is obtained, and a graded intervention strategy is implemented according to the forgetting risk value. User feature vectors are generated based on user behavior patterns, environmental impact indicators, and item attribute indicators included in the evaluation index set of smart storage cabinets. The intervention methods included in the hierarchical intervention strategy are defined as random forest processing variables, and the two cases of users picking up their items on time and users not picking them up after the time limit are defined as random forest outcome variables. Using user feature vectors and random forest processing variables as inputs to the random forest, and random forest outcome variables as outputs, linear regression fitting is performed using the random forest algorithm to obtain the intervention improvement efficacy of the random forest processing variables, and the hierarchical intervention strategy is modified based on the intervention improvement efficacy.

6. The multi-functional control method for an AI intelligent storage cabinet according to claim 1, characterized in that, Acquire real-time user demand data and perform demand analysis based on the intelligent storage cabinet evaluation index set, including: Real-time user demand data is obtained based on the information submitted by the user. The real-time user demand data includes at least the expected access time data, item type data, target counter range data, and service demand data. Based on the user behavior patterns included in the evaluation index set of smart storage cabinets, we conduct a demand analysis of access time periods based on the expected access time data and obtain the results of the access time period demand analysis. Based on the environmental impact indicators, item attribute indicators and storage cabinet operation indicators included in the intelligent storage cabinet evaluation index set, we conduct storage service demand analysis on item type data, target cabinet location range data and service demand data to obtain storage service demand analysis results. The demand matching degree function is used to merge the demand analysis results of the access period with the demand analysis results of the storage service to generate the demand matching degree. The demand matching degree is used to quantify the matching degree between user demand and storage cabinet deployment point. Storage cabinet deployment points with a demand matching degree exceeding 0.7 are classified as high-matching storage points, and a priority recommendation list is generated based on the high-matching storage points; Storage cabinet deployment points with a demand matching degree of [0.5, 0.7] are divided into standard matching degree storage points, and a candidate recommendation list is generated based on the standard matching degree storage points; Storage cabinet deployment sites with a demand matching degree of less than 0.5 are classified as low-matching storage sites and are not recommended.

7. The multi-functional control method for an AI intelligent storage cabinet according to claim 1, characterized in that, Obtain the requirements analysis results and combine them with the intelligent control model of the storage cabinet to generate an intelligent storage solution, including: Based on the priority recommendation list and alternative recommendation list included in the demand analysis results, and by calling the regional load balancing module of the storage cabinet intelligent control model, the optimal storage point for the load is obtained. The user storage time limit is allocated according to the user storage time limit allocation module of the storage cabinet intelligent control model; Anti-forgetting intervention module generates anti-forgetting information based on the intelligent control model of the storage cabinet; The optimal storage point for the load, the user storage time limit, and the information to prevent forgetting are encapsulated into an intelligent storage solution and sent to the user terminal.

8. The multi-functional control method for an AI intelligent storage cabinet according to claim 1, characterized in that, Acquire user behavior data, environmental data, item attribute data, and real-time cabinet status data to generate a multi-dimensional intelligent storage cabinet dataset, including: Data is collected from the operation logs of each smart storage cabinet to obtain corresponding user behavior data, environmental data, item attribute data, and storage cabinet status data; The user behavior data includes at least the historical access time distribution, operation interval duration, and on-time pickup rate; The environmental data includes at least real-time temperature, humidity, weather conditions, and population flow in the area; The item attribute data shall include at least the item category, preservation requirements, and estimated volume; The storage cabinet status data includes at least the occupancy rate of each cabinet's storage compartments, fault status, and function type. The acquired data is then subjected to outlier removal, missing value imputation, and data standardization to generate a multidimensional intelligent storage cabinet dataset.

9. A multi-functional control system for an AI intelligent storage cabinet, used to implement the method described in any one of claims 1 to 8, characterized in that, include: The data acquisition module is used to acquire user behavior data, environmental data, item attribute data and real-time cabinet status data to generate a multi-dimensional intelligent storage cabinet dataset. The indicator generation module is used to extract and fuse features from the multidimensional intelligent storage cabinet dataset to generate an evaluation indicator set for the intelligent storage cabinet. A model building module, which is used to build an intelligent control model for the storage cabinet based on the intelligent control mechanism of the storage cabinet; The demand analysis module is used to acquire real-time user demand data and perform demand analysis on the user demand data based on the intelligent storage cabinet evaluation index set. The solution generation module is used to generate intelligent storage solutions based on the requirements analysis results and the intelligent control model of the storage cabinet.

Images