Linkage control method and system of space intelligent retail cabin

Abstract

The present application relates to the technical field of intelligent retail, and more particularly to a linkage control method and system for a space intelligent retail cabin. By collecting user position, commodity operation data and commodity information, effective interaction events are obtained through space-time matching and time sequence consistency filtering, behavior sequences are constructed and time sequence and space features are extracted, the target commodity position is predicted by adjusting the features after weighting the environmental influence coefficient calculated in combination with temperature, the control dependence relationship between the lighting area and the shelf is determined, the control instruction is sent and the shelf drive is triggered according to the lighting state, and the interaction recognition accuracy and equipment are improved.

Description

Technical Field

[0001] This invention relates to the field of smart retail technology, and in particular to a linkage control method and system for smart retail cabins. Background Technology

[0002] As a new type of unmanned retail terminal, the core of the intelligent retail cabin lies in achieving unmanned and precise service through the collaboration of sensors and execution units. Current conventional practices mostly employ independent sensing and control architectures: user location is detected via visual cameras or infrared arrays, goods retrieval and placement events are monitored using shelf weighing sensors, lighting systems are independently switched on and off according to preset time periods or human body induction, and shelf drive devices rely on manual user operation or simple mechanical triggers. Some solutions attempt to align user location with shelf load data in time to identify operational behavior, but this usually only involves simple timestamp matching and lacks filtering mechanisms for sensor noise, redundant triggers, or non-operational actions (such as leaning or obstructing). Furthermore, the control logic of lighting and shelf drive equipment is isolated. Lighting often uses a fixed zone strategy or overall on / off within the human body induction range, while shelf drive waits for user confirmation or mechanical button signals. There is no dynamic dependency between the two, resulting in both inefficient energy consumption and delayed service response.

[0003] This conventional approach has significant drawbacks. First, the accuracy of user interaction event recognition is difficult to guarantee. Single or simple time-series matching can easily classify environmental disturbances (such as shelf vibration or accidental touches) as valid operations, thus misleading subsequent behavior prediction and linkage control. Such misidentifications not only increase invalid calculations but may also cause incorrect shelf driving or unnecessary lighting activation, reducing system reliability. Second, the existing control architecture does not incorporate environmental variables (such as cabin temperature) into user behavior analysis, ignoring the potential impact of temperature on user dwell time, touch habits, and product selection preferences. This results in poor generalization ability of behavioral feature spatial features and a significant deviation between product location prediction and actual intent. Isolated control of lighting and shelves further amplifies this deviation; for example, lighting may prematurely illuminate non-target areas, or shelf driving may be delayed, causing a fragmented user experience. Overall, existing methods have significant shortcomings in terms of anti-interference capabilities and adaptive coordination, making it difficult to meet the requirements of high-precision, low-latency smart retail scenarios. Summary of the Invention

[0004] The present invention provides a linkage control method and system for a smart retail cabin, which can solve the problems in the prior art.

[0005] A first aspect of the present invention provides a linkage control method for a smart retail cabin, comprising:

[0006] Collect user interaction data, product operation data, and product information within the smart retail cabin; The user location data in the user interaction data is spatiotemporally matched with the product operation data to obtain the initial interaction event. By comparing the temporal consistency between the time of the initial interaction event and the time of data change in the product information, falsely identified events are filtered out, and valid interaction events are output. Based on effective interaction events, user behavior sequences are constructed and behavioral temporal and spatial features are extracted. The behavioral temporal features are correlated with the current cabin temperature to calculate the environmental impact coefficient. The behavioral spatial features are weighted and adjusted according to the environmental impact coefficient to obtain the adjusted features. The location of the user's target product is then predicted based on the adjusted features. The lighting area and shelf location are determined based on the location of the user's target product. When the distance condition between the lighting area and the shelf location is met, a control dependency relationship is established between the lighting equipment and the shelf drive equipment. Control commands are sent to the lighting equipment based on control dependencies, and the status feedback of the lighting equipment is monitored. The status feedback is used as a trigger condition to send control commands to the rack drive equipment.

[0007] Preliminary interaction events are obtained by spatiotemporally matching user location data from user interaction data with product operation data. Falsely identified events are filtered out by comparing the temporal consistency between the timing of the preliminary interaction events and the timing of data changes in the product information. Valid interaction events are output, including: The user's movement trajectory is constructed based on the user's location data. The speed decay feature of the user's movement trajectory at the corresponding location of the product operation data is extracted, and the matching weight is calculated based on the speed decay feature. Calculate the spatial distance and time difference between user location data and product operation data. Combine the matching weight with the spatial distance and time difference to obtain the matching degree. Based on the matching degree, perform spatiotemporal matching between user location data and product operation data to obtain the initial interaction event. Extract the initial interaction event time and operation intensity parameters from the initial interaction event. Extract the time of change of shelf load-bearing data from the product information as the time of data change, extract the acceleration of the change of shelf load-bearing data, and calculate the time deviation between the time of the initial interaction event and the time of data change. When the time deviation meets the preset time window condition, the ratio of the acceleration of the change in shelf load data to the operation force parameter is calculated. When the ratio falls into the preset response range, it is determined that the time of the initial interaction event and the time of the change in shelf load data have temporal consistency. Based on the temporal consistency, falsely identified events are filtered out, and valid interaction events are output.

[0008] Based on effective interaction events, user behavior sequences are constructed, and temporal and spatial features of the behaviors are extracted. The environmental impact coefficient is calculated by correlating the temporal features of the behaviors with the current cabin temperature. The effective interaction events are arranged in chronological order to form a user behavior sequence. The time interval between adjacent effective interaction events, the duration of effective interaction events, and the frequency of operation are extracted from the user behavior sequence and combined to form the behavior temporal characteristics. Extract the product location coordinates corresponding to valid interaction events from the user behavior sequence, calculate the spatial movement distance between the product location coordinates of adjacent valid interaction events, and count the distribution density of product location coordinates in the user behavior sequence within the smart retail cabin. Combine the spatial movement distance and distribution density to form behavioral spatial features. The current cabin temperature is obtained, and the temperature sensitivity value is obtained by fitting the time interval in the behavior time series feature with the current cabin temperature. The operation delay value is obtained by fitting the duration in the behavior time series feature with the current cabin temperature. The temperature response intensity is calculated as a weighted combination of the temperature sensitivity value and the operation delay value. The temperature correction factor is obtained by multiplying the temperature response intensity with the operation frequency in the behavioral time sequence characteristics. The environmental impact coefficient is obtained by normalizing the temperature correction factor and the distribution density in the behavioral space characteristics.

[0009] The current cabin temperature is obtained. The temperature sensitivity value is obtained by fitting the time interval in the behavioral time series features to the current cabin temperature. The operation delay value is obtained by fitting the duration in the behavioral time series features to the current cabin temperature. Obtain the current cabin temperature and the reference temperature of the smart retail cabin, and calculate the temperature deviation between the current cabin temperature and the reference temperature. Extract time interval sequences from behavioral temporal features, calculate the average time interval of the time interval sequence, calculate the deviation between each time interval value in the time interval sequence and the average time interval, and multiply the temperature deviation by the deviation to obtain the temperature response offset. The temperature response offset and the time interval sequence are fitted with a polynomial to obtain the polynomial coefficients. The polynomial coefficients and the temperature deviation are then summed by a power weight to obtain the temperature sensitivity value. Extract the duration sequence from the behavioral temporal features, calculate the decay rate between adjacent duration values ​​in the duration sequence, and calculate the temperature effect decay coefficient by ratio of temperature deviation to decay rate. The exponential parameter is obtained by performing an exponential fitting operation on the temperature effect attenuation coefficient and the duration sequence. The operating delay value is obtained by exponentially powering the exponential parameter and the temperature deviation.

[0010] The behavioral spatial features are weighted and adjusted based on the environmental impact coefficient to obtain the adjusted features. The location of the user's target product is then predicted based on the adjusted features, including: Spatial movement distance and distribution density are extracted from behavioral spatial features. The environmental impact coefficient is multiplied by the spatial movement distance to obtain the adjusted spatial movement distance. The environmental impact coefficient is multiplied by the distribution density to obtain the adjusted distribution density. The adjusted spatial movement distance and the adjusted distribution density are combined to form the adjusted features. The adjusted spatial movement distance in the adjusted features is coupled with the time interval of effective interaction events in the user behavior sequence to generate spatiotemporal coupling weights. The spatiotemporal coupling weights are applied to the adjusted distribution density of the adjusted features to obtain the spatiotemporal weighted density distribution. Extract density peak regions from the spatiotemporal weighted density distribution, and calculate the consistency of the projection between the product location coordinates and the movement direction formed by continuous effective interaction events in the user behavior sequence within the density peak regions. The product location coordinates with the highest projection consistency are selected as the predicted target location. The spatial distance between the predicted target location and the product location coordinates of the last valid interaction event in the user behavior sequence is calculated. The spatial distance is then compared with the adjusted spatial movement distance in the adjusted features. When the ratio is less than a preset distance threshold, the predicted target location is determined as the user's target product location.

[0011] Extracting density peak regions from the spatiotemporal weighted density distribution, and calculating the consistency of the projection between the product location coordinates and the movement directions formed by consecutive valid interaction events in the user behavior sequence within these density peak regions, includes: Extract the product location coordinate sequence corresponding to consecutive valid interaction events from the user behavior sequence, perform curve fitting on the product location coordinate sequence to obtain the movement trajectory curve, and calculate the tangent vector of the movement trajectory curve at the product location coordinate of the last valid interaction event as the movement direction. A connection vector is constructed by linking the coordinates of each product position in the spatiotemporal weighted density distribution with the coordinates of the product position in the last valid interaction event. The cosine of the angle between the connection vector and the direction of movement is calculated as the direction weight. The direction-aware density distribution is obtained by multiplying the density values ​​corresponding to the coordinates of each product position in the spatiotemporal weighted density distribution with the direction weight. Extract spatially connected regions with density values ​​greater than a preset density threshold from the direction-aware density distribution as density peak regions; The maximum value of the directional weight corresponding to the commodity location coordinates within the density peak region is determined as the projection consistency.

[0012] The lighting area and shelf location are determined based on the user's target product location. When the lighting area and shelf location meet the distance condition, the control dependency relationship between the lighting equipment and the shelf drive equipment is established, including: Based on the location of the user's target product, query the corresponding shelf identifier, extract the coordinates of the shelf identifier as the shelf location, extract the physical dimension parameters of the shelf corresponding to the shelf location, and calculate the lighting coverage boundary to determine the lighting area based on the physical dimension parameters of the shelf. Calculate the spatial overlap rate between the lighting area and the shelf location, and calculate the lighting adaptation index by weighting the spatial overlap rate with the merchandise density of the shelf location. When the lighting adaptation index is greater than the average lighting adaptation index of all shelf locations within the lighting area, it is determined that the lighting area and the shelf location meet the distance condition. When the distance condition between the lighting area and the shelf location is met, the lighting equipment identifier corresponding to the lighting area and the shelf drive equipment identifier corresponding to the shelf location are extracted. The control state change sequence of the lighting equipment identifier and the response action sequence of the shelf drive equipment identifier are time-series mapped to generate a state-action dependency chain. In the state-action dependency chain, the brightness of the lighting equipment reaches a stable state as the start-up trigger condition for the shelf drive equipment. The lighting equipment identifier is controlled and bound to the shelf drive equipment identifier to establish a control dependency relationship between the lighting equipment and the shelf drive equipment.

[0013] A second aspect of the present invention provides a linkage control system for a smart retail cabin, comprising: The data acquisition unit is used to collect user location data, product operation data, and shelf load-bearing data within the intelligent retail cabin. The event matching unit is used to perform spatiotemporal matching between user location data in user interaction data and product operation data to obtain preliminary interaction events. By comparing the temporal consistency between the time of the preliminary interaction event and the time of data change in product information, falsely identified events are filtered out, and valid interaction events are output. The behavior analysis unit is used to construct user behavior sequences based on effective interaction events and extract behavioral temporal and spatial features. It then correlates the behavioral temporal features with the current cabin temperature to calculate the environmental impact coefficient. The feature adjustment unit is used to weight and adjust the behavioral space features according to the environmental impact coefficient to obtain the adjusted features, and predict the location of the user's target product based on the adjusted features; The dependency establishment unit is used to determine the lighting area and shelf location based on the location of the user's target product. When the lighting area and shelf location meet the distance condition, the control dependency relationship between the lighting equipment and the shelf drive equipment is established. The instruction control unit is used to send control instructions to the lighting equipment according to the control dependency, monitor the status feedback of the lighting equipment, and send control instructions to the rack drive equipment using the status feedback as a trigger condition.

[0014] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0015] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0016] This invention improves the accuracy of interactive event recognition and reduces the false positive rate through multi-source data fusion and temporal consistency verification. It correlates ambient temperature with the temporal characteristics of user behavior to achieve dynamic compensation for environmental factors in behavior prediction, improving the accuracy of target location prediction. Based on spatiotemporal coupling and projection consistency analysis, it achieves accurate prediction of user intent, making lighting and shelf responses more aligned with actual needs. Through state feedback-driven control dependencies, it achieves precise linkage between lighting equipment and shelf-driven equipment, avoiding control conflicts between devices, reducing energy consumption, and improving the smoothness of the user shopping experience. Multi-level behavioral feature extraction and weighted adjustment mechanisms enhance the system's adaptability to complex shopping scenarios, improving the service quality and operational efficiency of the smart retail cabin. Attached Figure Description

[0017] Figure 1 A flowchart illustrating the linkage control method for a smart retail cabin in space; Figure 2 Flowchart for binding shelf lighting and drive equipment control. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0020] Figure 1 This is a flowchart illustrating the linkage control method of the intelligent retail cabin in an embodiment of the present invention.

[0021] The linkage control methods for the intelligent retail cabin include: Collect user location data, product operation data, and shelf load-bearing data within the intelligent retail cabin; The user location data in the user interaction data is spatiotemporally matched with the product operation data to obtain the initial interaction event. By comparing the temporal consistency between the time of the initial interaction event and the time of data change in the product information, falsely identified events are filtered out, and valid interaction events are output. Based on effective interaction events, user behavior sequences are constructed and behavioral temporal and spatial features are extracted. The behavioral temporal features are correlated with the current cabin temperature to calculate the environmental impact coefficient. The behavioral spatial features are weighted and adjusted according to the environmental impact coefficient to obtain the adjusted features. The location of the user's target product is then predicted based on the adjusted features. The lighting area and shelf location are determined based on the location of the user's target product. When the distance condition between the lighting area and the shelf location is met, a control dependency relationship is established between the lighting equipment and the shelf drive equipment. Control commands are sent to the lighting equipment based on control dependencies, and the status feedback of the lighting equipment is monitored. The status feedback is used as a trigger condition to send control commands to the rack drive equipment.

[0022] The user interaction data includes user location, actions, gestures, facial expressions, and voice. Corresponding activities are initiated based on this data, such as gestures that change content, and voice-controlled ordering or content changes. Furthermore, the system acquires status information of the spatial display module and product information within the smart retail cabin, including product location, type, weight, appearance, size, temperature, RFID tags, and NFC tags. The spatial display module's status information includes its location and content status. When the device senses a user approaching, it simultaneously switches the screen display content and overall interaction status to proactively welcome them. Subsequent actions such as capturing user gaze, hesitation during purchase, and departure are used to push targeted product details, comparison information, and limited-time offers. The system also synchronizes product inventory status in real time, providing interface prompts and alternative product recommendations for out-of-stock items; and automatically completes pickup guidance and delivery operations for online orders. In addition, the device adaptively adjusts display effects and interaction rhythm according to changes in ambient lighting and pedestrian flow. If a hardware malfunction is detected, a fault warning immediately pops up and delivery operations are stopped.

[0023] Preliminary interaction events are obtained by spatiotemporally matching user location data from user interaction data with product operation data. Falsely identified events are filtered out by comparing the temporal consistency between the timing of the preliminary interaction events and the timing of data changes in the product information. Valid interaction events are output, including: The user's movement trajectory is constructed based on the user's location data. The speed decay feature of the user's movement trajectory at the corresponding location of the product operation data is extracted, and the matching weight is calculated based on the speed decay feature. Calculate the spatial distance and time difference between user location data and product operation data. Combine the matching weight with the spatial distance and time difference to obtain the matching degree. Based on the matching degree, perform spatiotemporal matching between user location data and product operation data to obtain the initial interaction event. Extract the initial interaction event time and operation intensity parameters from the initial interaction event. Extract the time of change of shelf load-bearing data from the product information as the time of data change, extract the acceleration of the change of shelf load-bearing data, and calculate the time deviation between the time of the initial interaction event and the time of data change. When the time deviation meets the preset time window condition, the ratio of the acceleration of the change in shelf load data to the operation force parameter is calculated. When the ratio falls into the preset response range, it is determined that the time of the initial interaction event and the time of the change in shelf load data have temporal consistency. Based on the temporal consistency, falsely identified events are filtered out, and valid interaction events are output.

[0024] When constructing user movement trajectories based on user location data, the collected continuous position coordinate sequences are arranged by timestamps to form a trajectory set describing the user's movement path within the cabin. Near the spatial location corresponding to the product operation data, users typically experience a significant decrease in walking speed upon noticing the target product—a phenomenon known as speed decay. The speed decay characteristics of the user's movement trajectory at the location corresponding to the product operation data are extracted. Specifically, the rate of change in speed before and after the user enters the neighborhood of that location is calculated. The greater the speed decay, the stronger the user's intention to linger or decelerate at that location, and the higher the correlation with product operation behavior. Matching weights are then calculated based on the speed decay characteristics. The more significant the speed decay, the higher the matching weight. The larger the value, the higher the likelihood that the user will interact with the product at that location.

[0025] Calculate the spatial distance between user location data and product operation data and time difference ,in This represents the Euclidean distance between the user's location coordinates and the location coordinates where the product action event occurred. This represents the absolute difference between the time the user's location was recorded and the time the product's operation data was recorded. (Matching weights) Spatial distance and time difference We perform weighted combination to obtain the overall matching degree. The calculation method is as follows ,in This is the spatial distance penalty coefficient. , where is the time difference penalty coefficient, and both are pre-calibrated positive parameters used to balance the impact of spatial proximity and temporal synchronization on the matching results. When When the preset matching threshold is exceeded, the user's location data and the corresponding product operation data are considered to form a valid spatiotemporal correlation, and the two are combined into a preliminary interaction event record.

[0026] Extract two key parameters from the initial interaction event: the timing of the initial interaction event. and operating force parameters Initial interaction event timing Taking the weighted average of user location data and product operation data provides a more accurate reflection of the actual time of the interaction. (Operation intensity parameter) The intensity of sensor signals derived from product handling data, such as peak signals recorded by pressure or contact sensors, characterizes the magnitude of the force applied by a user when picking up or touching the product. (Operating force parameter) The larger the value, the clearer the user's actions regarding the product, and the more convincing the subsequent verification of the correlation with changes in shelf load capacity.

[0027] The product information includes the location and weight of the shelf, product type, required storage temperature, dimensions, and RFID tags. The changes in shelf load-bearing capacity are extracted from this information as data change moments, and the acceleration of these changes is also extracted. Specifically, the shelf load-bearing capacity data is continuously collected by weighing sensors deployed on each shelf level, recording the changes in load value at every moment. The moments in the shelf load-bearing capacity data where significant changes occur are extracted, i.e., the moments of change in shelf load-bearing capacity. Simultaneously, the acceleration of the change in shelf load-bearing data during this process is calculated. It represents the rate of change of the load-bearing value per unit time. This reflects the drastic change in weight load when goods are removed from or placed on the shelf, and has a physical correspondence with the characteristics of user actions when handling the goods: the more rapid the operation, the greater the acceleration of the change in shelf weight load. The larger the size, the slower the operation. The smaller.

[0028] Calculate the initial interaction event time Changes in shelf load-bearing data Time deviation between ,in Time deviation This reflects the time delay between user interaction events captured by vision or sensors and the physical response of the shelving. This is due to signal transmission delay, sensor sampling period, and the physical propagation time between user action and shelving response. It cannot be zero, but it should fall within a reasonable physical range. The preset time window condition specifies... upper limit threshold , when , it is determined that the two moments have a basic possibility of correlation in time sequence, and the process proceeds to the next step of fine verification; when , it is directly determined that the moment of the preliminary interaction event and the moment of the shelf load-bearing data change do not have a time sequence correlation, and the preliminary interaction event is marked as a misidentification event for filtering.

[0029] After the preliminary screening by the time window, the acceleration of the shelf load-bearing data change is further calculated and the operation force parameter the ratio of , that is . The physical meaning of this ratio is: if the user indeed exerts an operation force on the goods on the shelf , the load-bearing change acceleration recorded by the shelf load-bearing sensor should be consistent with maintain a certain proportional relationship. This proportional relationship is affected by factors such as shelf structure and goods mass distribution, and has a relatively stable value range in the same cabin environment. The preset response interval is exactly a reasonable ratio range determined based on the calibration of the physical characteristics of the shelf in the cabin, in which is the lower limit of the response interval, is the upper limit of the response interval. When falls into , it indicates that there is a physically reasonable causal correspondence between the operation force recorded in the preliminary interaction event and the shelf load-bearing response, and it is determined that the moment of the preliminary interaction event and the moment of the shelf load-bearing data change have time sequence consistency.

[0030] Based on the time sequence consistency determination result, the preliminary interaction event set is filtered. All preliminary interaction events that fail to pass the time deviation test or the ratio interval test are identified as misidentification events, which may originate from the user passing near the shelf but not actually operating the goods, the false goods operation signal triggered by sensor noise, or the incorrect pairing of position data and goods operation data caused by multiple users moving in the cabin at the same time. After being screened by the dual filtering mechanism, the retained interaction events are verified in both the time dimension and the physical response dimension, and are output as effective interaction events for subsequent user behavior sequence construction and behavior feature extraction. This dual verification mechanism based on time sequence consistency can significantly reduce the misidentification rate, and ensure that subsequent linkage control decisions are based on reliable user behavior data.

[0031] In another optional embodiment of the present invention, other changes in product information can be used instead of shelf load data for time-series consistency verification. Optionally, based on changes in RFID tag signals, the moment of RFID tag signal change is extracted from the product information as the data change moment, and the rate of change of RFID tag signal strength is extracted. When a user interacts with a product, the RFID reader detects abrupt changes in the tag signal, including rapid attenuation or enhancement of signal strength. The time deviation between the initial interaction event moment and the RFID tag signal change moment is calculated. When this time deviation meets a preset time window condition (e.g., within 0.5 seconds), the correlation between the RFID tag signal strength change rate and the operation force parameter is further calculated. Since the force of the user's action when picking up the product affects the product's moving speed and distance, and thus affects the rate of change of RFID signal strength, a mapping relationship between the two can be established. When this correlation falls within a preset response range, it is determined that the initial interaction event moment and the RFID tag signal change moment have time-series consistency. Through this verification method based on radio frequency identification technology, false identification events caused by environmental interference or other users passing by can also be effectively filtered out, ensuring that the output valid interaction events accurately reflect the user's real shopping behavior and improving the interaction identification accuracy of the smart retail cabin.

[0032] Based on effective interaction events, user behavior sequences are constructed, and temporal and spatial features of the behaviors are extracted. The environmental impact coefficient is calculated by correlating the temporal features of the behaviors with the current cabin temperature. The effective interaction events are arranged in chronological order to form a user behavior sequence. The time interval between adjacent effective interaction events, the duration of effective interaction events, and the frequency of operation are extracted from the user behavior sequence and combined to form the behavior temporal characteristics. Extract the product location coordinates corresponding to valid interaction events from the user behavior sequence, calculate the spatial movement distance between the product location coordinates of adjacent valid interaction events, and count the distribution density of product location coordinates in the user behavior sequence within the smart retail cabin. Combine the spatial movement distance and distribution density to form behavioral spatial features. The current cabin temperature is obtained, and the temperature sensitivity value is obtained by fitting the time interval in the behavior time series feature with the current cabin temperature. The operation delay value is obtained by fitting the duration in the behavior time series feature with the current cabin temperature. The temperature response intensity is calculated as a weighted combination of the temperature sensitivity value and the operation delay value. The temperature correction factor is obtained by multiplying the temperature response intensity with the operation frequency in the behavioral time sequence characteristics. The environmental impact coefficient is obtained by normalizing the temperature correction factor and the distribution density in the behavioral space characteristics.

[0033] After acquiring valid interaction events, they are arranged sequentially from earliest to latest according to their timestamps, forming a continuous sequence of user behavior on a timeline. This sequence fully records all valid user actions performed within the smart retail cabin from entry to exit. Each sequence node includes the event time, corresponding product number, and shelf coordinates. Based on this sequence, features are extracted along both the temporal and spatial dimensions to provide input for subsequent calculations of environmental impact coefficients.

[0034] The extraction of behavioral temporal features revolves around three quantities. First, the time interval between two adjacent valid interaction events in the user behavior sequence is calculated, i.e., the difference between the time of the subsequent event and the time of the preceding event; this difference reflects the user's pause rhythm between consecutive operations. Second, the duration of each valid interaction event from its trigger to its end is counted; this duration reflects the user's dwell time during a single product interaction. Third, the number of valid interaction events occurring within a unit time window is counted to obtain the operation frequency, which reflects the user's overall operational activity level. These three quantities—time interval, duration, and operation frequency—are combined to form a behavioral temporal feature vector, which is then used for subsequent correlation calculations with temperature data.

[0035] The extraction of behavioral spatial features is centered on product location coordinates. The product location coordinates corresponding to each effective interaction event are read from each node of the user behavior sequence. These coordinates are recorded using a three-dimensional coordinate system within the intelligent retail cabin. The Euclidean distance between the product location coordinates of two adjacent effective interaction events is calculated to obtain the spatial movement distance. This distance describes the user's physical displacement between two consecutive operations; a larger spatial movement distance indicates that the user has moved extensively between shelves, while a smaller distance indicates that the user's attention is concentrated in a local area. Simultaneously, the product location coordinates of all effective interaction events in the user behavior sequence are projected onto a planar grid within the cabin. The number of coordinate points per unit grid area is counted to obtain the distribution density. Areas with high distribution density indicate a high concentration of user actions in that area. The spatial movement distance and distribution density are combined to construct the behavioral spatial features, which are then used for subsequent normalization and fusion with environmental influence coefficients.

[0036] After obtaining the current cabin temperature, the time interval sequence in the behavioral time series is fitted with the current cabin temperature. Specifically, using the current cabin temperature as the independent variable and the mean or weighted mean of the time intervals as the dependent variable, a mapping relationship between the two is established through linear regression or multinomial fitting, outputting a temperature sensitivity value. . The physical significance lies in quantifying the degree to which temperature affects the user's operating rhythm: when the cabin temperature is too high or too low, the user's operating intervals tend to change systematically. The higher the temperature, the more significant the disturbance to user behavior rhythm. Similarly, the duration sequence of behavioral time-series features is fitted with the current cabin temperature, and a fitting model is established with the duration statistic (such as the mean) as the dependent variable, outputting the operation delay value. . This reflects the delay effect of temperature on the duration of a user's single interaction; when temperatures are abnormal, users tend to stay in front of the product for a shorter or longer time. This regular offset was captured.

[0037] Obtain temperature sensitivity values and operation delay value Then, the two are weighted and combined to calculate the temperature response intensity. .set up The weight is , The weight is ,and The temperature response intensity is calculated as follows: .in and The value can be pre-calibrated based on historical data statistics, or it can be dynamically adjusted according to the real-time operation scenario inside the cabin. This comprehensively reflects the overall intensity of temperature's interference with the temporal aspect of user behavior. It also includes the intensity of the temperature response. Operation frequency in behavioral temporal features Perform a product operation to obtain the temperature correction factor. ,Right now Operation frequency This plays a moderating role in amplifying or reducing the influence of temperature: when the user operates frequently, the cumulative effect of temperature on overall behavior is more prominent, and the temperature correction factor increases accordingly; when the operation frequency is low, the interference effect of temperature is moderately compressed.

[0038] Temperature correction factor Distribution density in behavioral space features After normalization, the final environmental impact coefficient is obtained. The purpose of normalization is to eliminate... and The differences in units and numerical ranges allow for fusion at a unified scale. The specific normalization method involves... and Dividing the product by the maximum possible range of each after normalization ensures The output value falls within Within the range. The closer it is to 1, the stronger the overall impact of the current cabin temperature on user behavior, and the more concentrated the user's actions are in space; The closer the value is to 0, the weaker the influence of temperature or the more spatially dispersed the user behavior. Environmental Influence Coefficient This will serve as a core parameter for subsequent weighted adjustments to the behavioral spatial characteristics, directly determining the accuracy of the linkage control between the lighting area and the shelf drive equipment. By introducing temperature as an environmental variable to correct the behavioral characteristics, it is possible to effectively avoid deviations in the prediction of the user's target product location due to changes in environmental conditions, thereby improving the overall response accuracy of the linkage control in the intelligent retail cabin.

[0039] The current cabin temperature is obtained. The temperature sensitivity value is obtained by fitting the time interval in the behavioral time series features to the current cabin temperature. The operation delay value is obtained by fitting the duration in the behavioral time series features to the current cabin temperature. Obtain the current cabin temperature and the reference temperature of the smart retail cabin, and calculate the temperature deviation between the current cabin temperature and the reference temperature. Extract time interval sequences from behavioral temporal features, calculate the average time interval of the time interval sequence, calculate the deviation between each time interval value in the time interval sequence and the average time interval, and multiply the temperature deviation by the deviation to obtain the temperature response offset. The temperature response offset and the time interval sequence are fitted with a polynomial to obtain the polynomial coefficients. The polynomial coefficients and the temperature deviation are then summed by a power weight to obtain the temperature sensitivity value. Extract the duration sequence from the behavioral temporal features, calculate the decay rate between adjacent duration values ​​in the duration sequence, and calculate the temperature effect decay coefficient by ratio of temperature deviation to decay rate. The exponential parameter is obtained by performing an exponential fitting operation on the temperature effect attenuation coefficient and the duration sequence. The operating delay value is obtained by exponentially powering the exponential parameter and the temperature deviation.

[0040] Get the current cabin temperature The reference temperature of the smart retail cabin Calculate the temperature difference between the two. Reference temperature This is a pre-calibrated reference temperature value inside the cabin under standard operating conditions, typically set by the operator during the equipment initialization phase based on the geographical area and seasonal characteristics of the cabin. Temperature deviation This reflects the degree of deviation of the current cabin thermal environment from the standard state; a positive value indicates that the current temperature is higher than the baseline, while a negative value indicates that it is lower than the baseline. This deviation will serve as the core driving force for subsequent fitting calculations, directly participating in the calculation of temperature sensitivity and operational delay values.

[0041] Extracting time interval sequences from behavioral temporal features ,in Represents the first in the user behavior sequence The time interval between adjacent valid interaction events This represents the total length of the time interval sequence. Calculate the average time interval of the time interval sequence. Then calculate the deviation between each time interval value and the average time interval. Deviation range It depicts the degree of fluctuation in user behavior rhythm at different interaction moments relative to the overall mean. Temperature deviation... With each deviation magnitude Perform the product operation to obtain the corresponding temperature response offset. Temperature response offset The physical meaning is that when the cabin temperature is high, the user's behavior rhythm tends to speed up, and the degree of deviation of the time interval from the mean, together with the temperature deviation, determines the quantitative magnitude of the influence of the thermal environment on the behavior rhythm; when the temperature is low, the user's operation tends to be slow, and the product of the deviation magnitude and the temperature deviation can also reflect the inhibitory effect of the cold environment on the behavior rhythm.

[0042] Temperature response offset sequence With time interval sequence Perform polynomial fitting operations to... For independent variable, As the dependent variable, the fitting yields an order of [order missing]. set of polynomial coefficients The purpose of polynomial fitting is to capture the nonlinear influence of the temperature response offset on the time interval distribution, and the order... The order can be adaptively selected based on the length of the behavioral sequence and the fitting residuals, typically between 2 and 4 orders to balance fitting accuracy and computational stability. After obtaining the polynomial coefficients, the coefficients of each order are... ( ) and temperature deviation Perform a power-weighted summation to obtain the temperature sensitivity value. Temperature sensitivity value It comprehensively reflects the intensity of the response of user behavior rhythm to temperature changes under the current thermal environment conditions. The larger the absolute value, the more significant the influence of temperature on the timing of user behavior.

[0043] Extracting duration sequences from behavioral temporal features ,in Indicates user number Duration of each valid interaction event Given the total length of the duration sequence. Calculate the decay rate between adjacent duration values ​​in the duration sequence. ( ), decay rate This describes the decreasing trend of the duration of a single user action during continuous interaction; positive values ​​indicate a decreasing trend in duration, while negative values ​​indicate a prolonged duration. Temperature deviation... With each decay rate The temperature effect attenuation coefficient is obtained by performing a ratio calculation. ,in A minimal positive smoothing term is introduced to prevent the denominator from being zero. Temperature affects the attenuation coefficient. The degree to which temperature deviation modulates the duration decay trend was quantified: when the decay rate is small, the effect of temperature deviation on duration is relatively amplified; when the decay rate is large, the natural decay of duration dominates the behavioral change, and the additional effect of temperature is relatively reduced.

[0044] Temperature-affected attenuation coefficient sequence Duration sequence Perform exponential fitting calculations to... For independent variable, Using the dependent variable, the magnitude parameters of the exponential function are obtained through fitting. and attenuation parameters That is, the fitting form is Exponential fitting can effectively characterize the nonlinear growth or decay of duration as the decay coefficient changes with temperature, and compared to linear fitting, it better reflects the actual changes in user behavior under extreme temperature conditions. This is achieved by obtaining the exponential parameters. Then, compare it with the temperature deviation. Perform exponentiation to obtain the operation delay value. Operational latency value This reflects the degree of delay or acceleration in the duration of a single user interaction operation relative to the baseline state under the current temperature deviation conditions: when and hour, This indicates that high-temperature environments lead to longer user operation durations; when hour, This indicates that low-temperature environments make user operations tend to be shorter.

[0045] The two calculation paths described above comprehensively characterize the impact of the current cabin thermal environment on the temporal characteristics of user behavior from two dimensions: the interval fluctuations of behavioral rhythm and the duration of a single operation. Temperature sensitivity values... It focuses on reflecting the user's response to temperature changes at the level of overall behavioral rhythm, while the operation latency value... This focuses on the degree to which a single interaction is modulated by temperature over its duration. Both factors contribute to the subsequent environmental impact coefficient. The calculation provides a refined temperature-related input, enabling the final weighted adjustment of behavioral features to more accurately reflect the potential interference of the thermal environment on the user's target product positioning in real shopping scenarios.

[0046] The behavioral spatial features are weighted and adjusted based on the environmental impact coefficient to obtain the adjusted features. The location of the user's target product is then predicted based on the adjusted features, including: Spatial movement distance and distribution density are extracted from behavioral spatial features. The environmental impact coefficient is multiplied by the spatial movement distance to obtain the adjusted spatial movement distance. The environmental impact coefficient is multiplied by the distribution density to obtain the adjusted distribution density. The adjusted spatial movement distance and the adjusted distribution density are combined to form the adjusted features. The adjusted spatial movement distance in the adjusted features is coupled with the time interval of effective interaction events in the user behavior sequence to generate spatiotemporal coupling weights. The spatiotemporal coupling weights are applied to the adjusted distribution density of the adjusted features to obtain the spatiotemporal weighted density distribution. Extract density peak regions from the spatiotemporal weighted density distribution, and calculate the consistency of the projection between the product location coordinates and the movement direction formed by continuous effective interaction events in the user behavior sequence within the density peak regions. The product location coordinates with the highest projection consistency are selected as the predicted target location. The spatial distance between the predicted target location and the product location coordinates of the last valid interaction event in the user behavior sequence is calculated. The spatial distance is then compared with the adjusted spatial movement distance in the adjusted features. When the ratio is less than a preset distance threshold, the predicted target location is determined as the user's target product location.

[0047] Extracting spatial movement distance and distribution density from behavioral spatial features is a fundamental step in quantitatively describing user activity trajectories within the cabin. Spatial movement distance reflects the displacement amplitude of a user between consecutive effective interaction events, while distribution density describes the degree of clustering of user dwell and operational behaviors in different areas of the cabin. The adjusted spatial movement distance is obtained by multiplying the environmental influence coefficient by the spatial movement distance. ,Right now ,in This represents the original spatial distance traveled. This is the environmental impact coefficient. Similarly, the adjusted distribution density is obtained by multiplying the environmental impact coefficient by the distribution density. ,Right now ,in This represents the original distribution density. Through this product adjustment, when higher cabin temperatures cause user behavior to slow down, the environmental impact coefficient will decrease accordingly, thus lowering the estimated spatial movement distance and distribution density, making subsequent predictions closer to actual user behavior. The adjusted spatial movement distance... and adjusted distribution density The combined features form an adjusted feature vector for use in subsequent coupling operations.

[0048] After obtaining the adjusted features, the adjusted spatial movement distance will be... The spatiotemporal coupling weights are generated by coupling calculations with the time intervals of valid interaction events in the user behavior sequence. Specifically, the time interval reflects the time span between adjacent valid interaction events. A shorter time interval indicates more intensive user interaction in that area, and the corresponding spatiotemporal coupling weight should be higher; a longer time interval results in a relatively lower weight. The coupling operation uses the ratio of the adjusted spatial movement distance to the time interval, i.e. ,in The time interval between adjacent valid interaction events. To prevent the introduction of a very small positive smoothing term when the denominator is zero. The larger this ratio, the more it indicates that the user completed a large spatial movement in a short period, demonstrating high activity levels, and the prediction priority for the corresponding area should be increased accordingly. Spatiotemporal coupling weights. This comprehensively reflects the intensity of user behavior in both time and space dimensions, providing a reasonable basis for subsequent weighting processing.

[0049] Spatiotemporal coupling weights Adjusted distribution density After weighting, the spatiotemporal weighted density distribution is obtained. ,Right now The spatiotemporal weighted density distribution, based on the original distribution density, superimposes time interval information to correct for spatial density. This results in higher density values ​​for areas frequently visited in a short period, while lower density values ​​for areas that have been visited historically but have sparse recent activity. This approach effectively highlights the core areas of user interest, suppresses historical interference signals, and improves the timeliness and accuracy of predictions. The spatiotemporal weighted density distribution uses spatial coordinates as the horizontal axis and weighted density values ​​as the vertical axis, forming a density field covering the entire cargo hold shelf area. The peak areas in this density field represent the target areas most likely to be visited by the user.

[0050] From the spatiotemporal weighted density distribution Extracting density peak regions is achieved by setting a density threshold. To achieve this, Medium density value exceeds The continuous spatial region is labeled as a density peak region. Multiple density peak regions may exist, each containing several candidate product location coordinates. To further filter the multiple candidate locations to find the target location that best matches the user behavior trend, projection consistency calculation is introduced. Projection consistency measures the degree of directional agreement between a product location coordinate and the movement directions formed by consecutive effective interaction events in the user behavior sequence. Specifically, the location coordinates of consecutive effective interaction events in the user behavior sequence are extracted, the displacement vectors between adjacent events are calculated, and the average of all displacement vectors is obtained to obtain the comprehensive movement direction vector. For the location coordinates of each candidate commodity within the density peak region. Calculate from the user's current location point to Direction vector , then calculate exist Projection values ​​on ,Right now The larger the projection value, the more closely the candidate location matches the user's current movement trend, and the higher the prediction accuracy.

[0051] Select the product location coordinates with the highest projection consistency, i.e. The largest corresponding As the predicted target location To further verify the rationality of the predicted target location and avoid misjudging locations that are too far away or beyond the user's actual behavioral range as targets, the predicted target location was compared with the product location coordinates of the last valid interaction event in the user behavior sequence. Spatial distance between Perform calculations, that is .Will Adjusted spatial movement distance in the adjusted features Perform a ratio calculation to obtain the distance ratio. ,Right now .when Less than the preset distance threshold When the predicted target location is determined to be within the range of mobility demonstrated by the user's behavior sequence, confirmation is made. The location of the user's target product. If... Greater than or equal to If the predicted target location deviates from the user's actual behavioral capabilities, it is necessary to reselect the candidate location with the second highest projection consistency from the density peak area and repeat the above verification process until a predicted target location that meets the distance condition is found.

[0052] Preset distance threshold The design is based on statistical analysis of historical user behavior data. By statistically analyzing the spatial movement distances of a large number of historical valid interaction events, the mean and standard deviation are calculated. The mean plus a certain multiple of the standard deviation is then used as the basis for the designation. The reference values ​​ensure that the movement range of the vast majority of normal users is covered, while effectively eliminating interference from abnormally erratic locations. In actual deployment, It can be dynamically adjusted according to different cabin layouts and user group characteristics to adapt to the prediction needs of different scenarios. The entire prediction process forms a multi-level, progressively converging user target product location prediction mechanism through weighted adjustment of environmental impact coefficients, generation of spatiotemporal coupling weights, calculation of spatiotemporal weighted density distribution, screening of projection consistency, and verification of distance ratios. It can achieve high-precision location prediction in complex cabin environments, providing accurate spatial basis for the subsequent linkage control of lighting equipment and shelf drive equipment.

[0053] Extracting density peak regions from the spatiotemporal weighted density distribution, and calculating the consistency of the projection between the product location coordinates and the movement directions formed by consecutive valid interaction events in the user behavior sequence within these density peak regions, includes: Extract the product location coordinate sequence corresponding to consecutive valid interaction events from the user behavior sequence, perform curve fitting on the product location coordinate sequence to obtain the movement trajectory curve, and calculate the tangent vector of the movement trajectory curve at the product location coordinate of the last valid interaction event as the movement direction. A connection vector is constructed by linking the coordinates of each product position in the spatiotemporal weighted density distribution with the coordinates of the product position in the last valid interaction event. The cosine of the angle between the connection vector and the direction of movement is calculated as the direction weight. The direction-aware density distribution is obtained by multiplying the density values ​​corresponding to the coordinates of each product position in the spatiotemporal weighted density distribution with the direction weight. Extract spatially connected regions with density values ​​greater than a preset density threshold from the direction-aware density distribution as density peak regions; The maximum value of the directional weight corresponding to the commodity location coordinates within the density peak region is determined as the projection consistency.

[0054] After constructing the spatiotemporal weighted density distribution, it is necessary to further extract the density peak region that can reflect the user behavior trend, and evaluate the consistency between the location coordinates of each product and the direction of user behavior movement within this region, so as to provide a more accurate directional basis for the final prediction of the target product location.

[0055] A sequence of product location coordinates corresponding to consecutive valid interaction events is extracted from the user behavior sequence. This coordinate sequence is arranged chronologically according to the events, forming a set of discrete coordinate points with a temporal relationship within the cabin space. Curve fitting is performed on this coordinate sequence using parametric spline interpolation or least-squares polynomial fitting. By smoothing the discrete coordinate points, a continuous movement trajectory curve is obtained. This curve reflects the spatial movement pattern of the user during valid interactions, eliminating the jitter caused by positional noise in a single interaction event. After obtaining the movement trajectory curve, the tangent vector at the product location coordinates of the last valid interaction event is calculated. This tangent vector is taken as the movement direction of the user's behavior at the current moment. The tangent vector is calculated by taking the first derivative of the trajectory curve at that coordinate point; the resulting derivative vector is the tangent direction. After normalization, a normalized movement direction vector is obtained, denoted as . This vector has a clear directional component in the two-dimensional cabin planar coordinate system, which can accurately represent the spatial extension trend of user behavior.

[0056] The coordinates of all product locations in the spatiotemporal weighted density distribution are compared with the coordinates of the product location in the last valid interaction event. Construct a connection vector. For any commodity location coordinate in the spatiotemporal weighted density distribution... , and Connection vectors between Defined from point to The direction vector, for After normalization, calculate its relationship with the direction of movement vector. The cosine of the included angle is used as the direction weight of that coordinate point. ,Right now The range of values ​​for the cosine of the included angle is: ,when near When the coordinate point is directly in front of the user's movement direction, the directional consistency is high; when near When the coordinate point is in the opposite direction of movement, it indicates low directional consistency. The spatiotemporal weighted density distribution... Coordinates of each product's location Corresponding density value with direction weight Perform a product operation to obtain the orientation-aware density distribution. ,Right now This product operation fuses density and direction information, resulting in regions with higher density in the user's movement direction receiving a higher overall score, while regions on the opposite side of the movement direction, even if they have high density, are suppressed due to their negative direction weights, thus effectively reducing the interference of reverse regions on the prediction results.

[0057] Density distribution perceived by direction Extracting density values ​​greater than a preset density threshold The spatial connectivity region that satisfies this condition is identified as the density peak region. The method for determining the spatial connectivity region is as follows: In the intra-cabin planar coordinate system, all regions that satisfy this condition are identified as density peak regions. The coordinates of the points are considered as candidate points. A connectivity analysis is performed on the candidate point set. If the Euclidean distance between two candidate points is less than a preset connectivity distance threshold, the candidate point is considered a candidate point. If the two points are not clearly defined, they are considered to belong to the same connected region. This connectivity analysis can aggregate dispersed high-density points into several spatially connected density peak regions. Each density peak region corresponds to a local area within the cabin space where user behavior is concentrated and directional consistency is high. When multiple density peak regions exist, they can be sorted according to the sum of the directional perception density values ​​within each region, and the region with the highest comprehensive score is selected as the primary density peak region for subsequent projection consistency calculations.

[0058] After determining the density peak region, the orientation weights corresponding to the location coordinates of all commodities within the density peak region are compared, and the maximum value is taken as the projection consistency metric for that density peak region, denoted as . ,Right now Projection consistency This reflects the directional alignment of the product location coordinates within the density peak region with the user's movement direction. When the value is large, it indicates that there are candidate locations within the density peak region that are highly consistent with the direction of movement, and the reliability of the prediction results is high; when When the value is small or even negative, it indicates that the density peak region deviates significantly from the direction of movement. In this case, the prediction results can be corrected or reduced in weight by combining other auxiliary criteria.

[0059] In practical applications, when users browse products within the cabin, the locations of products corresponding to their effective interaction events often exhibit certain directional patterns in space. By calculating the tangent to the movement trajectory curve, we can accurately capture the user's behavioral trend at the end of the most recent interaction, rather than relying solely on the average direction of historical trajectories. By fusing this tangent direction with a spatiotemporally weighted density distribution to construct a direction-aware density distribution, and then extracting density peak regions and calculating projection consistency, the entire prediction process retains density statistics while fully utilizing directional cues from user behavior, thereby improving the accuracy and robustness of target product location prediction.

[0060] like Figure 2 The diagram shown illustrates the process of binding the shelf lighting and drive equipment control in this embodiment.

[0061] The lighting area and shelf location are determined based on the user's target product location. When the lighting area and shelf location meet the distance condition, the control dependency relationship between the lighting equipment and the shelf drive equipment is established, including: Based on the location of the user's target product, query the corresponding shelf identifier, extract the coordinates of the shelf identifier as the shelf location, extract the physical dimension parameters of the shelf corresponding to the shelf location, and calculate the lighting coverage boundary to determine the lighting area based on the physical dimension parameters of the shelf. Calculate the spatial overlap rate between the lighting area and the shelf location, and calculate the lighting adaptation index by weighting the spatial overlap rate with the merchandise density of the shelf location. When the lighting adaptation index is greater than the average lighting adaptation index of all shelf locations within the lighting area, it is determined that the lighting area and the shelf location meet the distance condition. When the distance condition between the lighting area and the shelf location is met, the lighting equipment identifier corresponding to the lighting area and the shelf drive equipment identifier corresponding to the shelf location are extracted. The control state change sequence of the lighting equipment identifier and the response action sequence of the shelf drive equipment identifier are time-series mapped to generate a state-action dependency chain. In the state-action dependency chain, the brightness of the lighting equipment reaches a stable state as the start-up trigger condition for the shelf drive equipment. The lighting equipment identifier is controlled and bound to the shelf drive equipment identifier to establish a control dependency relationship between the lighting equipment and the shelf drive equipment.

[0062] After obtaining the position of the user's target commodity, it is necessary to associate the position information with the physical layout information of the shelves in the compartment to determine the range of devices that actually need to be illuminated and driven. Specifically, a coordinate query is executed in the shelf database according to the position of the user's target commodity, the shelf record closest to the position is matched, and the corresponding shelf identifier is extracted. The shelf identifier is a unique code for each shelf unit in the compartment, and is stored in the shelf information table together with the three-dimensional coordinate information, physical dimension parameters and device binding information of the shelf. After extracting the shelf identifier, the coordinates corresponding to the shelf identifier are read from the shelf information table, the coordinates are taken as the shelf position, and the physical dimension parameters of the shelf corresponding to the shelf position are read at the same time, including the width, height, depth of the shelf and the installation orientation angle of the shelf in the compartment.

[0063] When calculating the illumination coverage boundary based on the physical dimension parameters of the shelf, taking the shelf coordinates as the center, and expanding a certain margin outward in combination with the width and depth of the shelf, a rectangular or polygonal illumination coverage area is formed, which is the illumination area. The size of the margin is usually pre-calibrated according to the light cone angle and installation height of the lighting equipment, so as to ensure that the illumination area can completely cover the commodity display surface of the shelf. The calculation of the illumination coverage boundary comprehensively considers the actual projected area of the shelf and the effective irradiation range of the lighting equipment, so as to avoid insufficient illumination or over-illumination caused by the difference in shelf dimensions.

[0064] After the illumination area is determined, the spatial overlap rate between the illumination area and the shelf position is calculated. The spatial overlap rate reflects the spatial overlap degree between the illumination area and the actual occupied area of the shelf, which is obtained by calculating the ratio of the intersection area of the two to the floor area of the shelf, with a value ranging from 0 to 1. A larger value indicates a more sufficient coverage of the shelf by the illumination area. On this basis, the commodity density at the shelf position is introduced as an auxiliary judgment index. The commodity density represents the number of commodities stored per unit area at the current shelf position. A higher commodity density indicates a more urgent demand for illumination quality at the shelf position. A weighting operation is performed on the spatial overlap rate and the commodity density to obtain an illumination adaptation index , the calculation method is , where is the spatial overlap rate, is the commodity density, and are respectively the weighting coefficients of the spatial overlap rate and the commodity density, the sum of the two is 1, and the specific values are calibrated according to the actual operation scenario in the compartment.

[0065] The illumination adaptation index is introduced so that the judgment of the distance condition no longer only depends on the absolute value of the spatial distance, but comprehensively considers the illumination coverage quality and the distribution state of commodities on the shelf, thereby avoiding unnecessary linkage control for shelves with sparse commodities. When the illumination adaptation index The lighting adaptability index is greater than the average of all shelf locations within the lighting area. When the distance between the lighting area and the shelf location is satisfied, it is determined that the distance condition is met. This relative threshold mechanism, obtained by averaging the lighting adaptation index of all candidate shelf locations within the lighting area, can dynamically adapt to different interior layouts and merchandise display states without requiring manual preset of fixed thresholds, and has strong adaptability.

[0066] After determining that the distance between the lighting area and the shelf location meets the distance requirement, the lighting equipment identifiers corresponding to the lighting area and the shelf drive equipment identifiers corresponding to the shelf location are extracted. There is a pre-configured mapping relationship between the lighting equipment identifiers and the lighting areas; one lighting area may correspond to one or more lighting equipment identifiers, and all relevant identifiers are extracted. Similarly, there is a correspondence between the shelf drive equipment identifiers and the shelf location. The shelf drive equipment is responsible for performing physical actions such as rotating, pushing, or adjusting the display angle of the shelf to assist users in retrieving goods.

[0067] After extracting the device identifier, the control state change sequence of the lighting device identifier is time-series mapped to the response action sequence of the shelf drive device identifier, generating a state-action dependency chain. The control state change sequence describes the complete state transition process of the lighting device from receiving a control command to achieving brightness stabilization, including state nodes such as "command reception," "starting brightness," and "brightness stabilization," as well as the expected timestamps of each node. The response action sequence describes the action flow of the shelf drive device after receiving the start command, including action nodes such as "waiting for trigger," "starting drive," "action execution," and "action completion." The core of the time-series mapping is to establish a predecessor-successor relationship between the state nodes of the lighting device and the action nodes of the shelf drive device according to time sequence, ensuring that the start time of the shelf drive device is precisely aligned with the state completion time of the lighting device, avoiding premature shelf action before the lighting is stable, which would lead to a degraded user experience.

[0068] In the state-action dependency chain, the stable brightness state of the lighting equipment is configured as the trigger condition for the rack drive equipment to start. The determination of a stable brightness state is based on the real-time brightness feedback value reported by the lighting equipment. When the fluctuation range of the feedback brightness value within a certain number of consecutive sampling periods is lower than a preset stability threshold, the lighting equipment is considered to have entered a stable state, and the trigger condition is met. This configuration decouples the physical state of the lighting equipment from the control logic of the rack drive equipment. The lighting equipment does not need to be aware of the existence of the rack drive equipment, while the rack drive equipment determines its own start-up timing by listening to the state feedback from the lighting equipment, forming a unidirectional dependency control chain.

[0069] After configuring the trigger conditions, the lighting equipment identifier and the shelf drive device identifier are bound together to establish a control dependency relationship between them. The control binding record includes fields such as lighting equipment identifier, shelf drive device identifier, state-action dependency chain description, trigger condition parameters, and binding activation timestamp, and is stored in the linkage control configuration table. Subsequently, when sending control commands to the lighting equipment, the linkage control logic automatically retrieves the shelf drive device identifier bound to that lighting equipment identifier. When the lighting equipment's state feedback meets the trigger conditions, it issues a control command to the corresponding shelf drive device based on the timing mapping relationship in the dependency chain, achieving coordinated linkage control between lighting and shelf drive. This state feedback-driven linkage mechanism offers higher reliability compared to fixed-delay triggering methods, adapting to changes in response time caused by aging or load differences in different lighting equipment, ensuring the accuracy and consistency of linkage control.

[0070] A second aspect of the present invention provides a linkage control system for a smart retail cabin, comprising: The data acquisition unit is used to collect user location data, product operation data, and shelf load-bearing data within the intelligent retail cabin. The event matching unit is used to perform spatiotemporal matching between user location data in user interaction data and product operation data to obtain preliminary interaction events. By comparing the temporal consistency between the time of the preliminary interaction event and the time of data change in product information, falsely identified events are filtered out, and valid interaction events are output. The behavior analysis unit is used to construct user behavior sequences based on effective interaction events and extract behavioral temporal and spatial features. It then correlates the behavioral temporal features with the current cabin temperature to calculate the environmental impact coefficient. The feature adjustment unit is used to weight and adjust the behavioral space features according to the environmental impact coefficient to obtain the adjusted features, and predict the location of the user's target product based on the adjusted features; The dependency establishment unit is used to determine the lighting area and shelf location based on the location of the user's target product. When the lighting area and shelf location meet the distance condition, the control dependency relationship between the lighting equipment and the shelf drive equipment is established. The instruction control unit is used to send control instructions to the lighting equipment according to the control dependency, monitor the status feedback of the lighting equipment, and send control instructions to the rack drive equipment using the status feedback as a trigger condition.

[0071] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0072] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0073] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0074] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A linkage control method for a spatial intelligent retail cabin, characterized in that, include: Collect user interaction data, product operation data, and product information within the smart retail cabin; The user location data in the user interaction data is spatiotemporally matched with the product operation data to obtain the initial interaction event. By comparing the temporal consistency between the time of the initial interaction event and the time of data change in the product information, falsely identified events are filtered out, and valid interaction events are output. Based on effective interaction events, user behavior sequences are constructed and behavioral temporal and spatial features are extracted. The behavioral temporal features are correlated with the current cabin temperature to calculate the environmental impact coefficient. The behavioral spatial features are weighted and adjusted according to the environmental impact coefficient to obtain the adjusted features. The location of the user's target product is then predicted based on the adjusted features. The lighting area and shelf location are determined based on the location of the user's target product. When the distance condition between the lighting area and the shelf location is met, a control dependency relationship is established between the lighting equipment and the shelf drive equipment. Control commands are sent to the lighting equipment based on control dependencies, and the status feedback of the lighting equipment is monitored. The status feedback is used as a trigger condition to send control commands to the rack drive equipment.

2. The method according to claim 1, characterized in that, Preliminary interaction events are obtained by spatiotemporally matching user location data from user interaction data with product operation data. Falsely identified events are filtered out by comparing the temporal consistency between the timing of the preliminary interaction events and the timing of data changes in the product information. Valid interaction events are output, including: The user's movement trajectory is constructed based on the user's location data. The speed decay feature of the user's movement trajectory at the corresponding location of the product operation data is extracted, and the matching weight is calculated based on the speed decay feature. Calculate the spatial distance and time difference between user location data and product operation data. Combine the matching weight with the spatial distance and time difference to obtain the matching degree. Based on the matching degree, perform spatiotemporal matching between user location data and product operation data to obtain the initial interaction event. Extract the initial interaction event time and operation intensity parameters from the initial interaction event. Extract the time of change of shelf load-bearing data from the product information as the time of data change, extract the acceleration of the change of shelf load-bearing data, and calculate the time deviation between the time of the initial interaction event and the time of data change. When the time deviation meets the preset time window condition, the ratio of the acceleration of the change in shelf load data to the operation force parameter is calculated. When the ratio falls into the preset response range, it is determined that the time of the initial interaction event and the time of the change in shelf load data have temporal consistency. Based on the temporal consistency, falsely identified events are filtered out, and valid interaction events are output.

3. The method according to claim 1, characterized in that, Based on effective interaction events, user behavior sequences are constructed, and temporal and spatial features of the behaviors are extracted. The environmental impact coefficient is calculated by correlating the temporal features of the behaviors with the current cabin temperature. The effective interaction events are arranged in chronological order to form a user behavior sequence. The time interval between adjacent effective interaction events, the duration of effective interaction events, and the frequency of operation are extracted from the user behavior sequence and combined to form the behavior temporal characteristics. Extract the product location coordinates corresponding to valid interaction events from the user behavior sequence, calculate the spatial movement distance between the product location coordinates of adjacent valid interaction events, and count the distribution density of product location coordinates in the user behavior sequence within the smart retail cabin. Combine the spatial movement distance and distribution density to form behavioral spatial features. The current cabin temperature is obtained, and the temperature sensitivity value is obtained by fitting the time interval in the behavior time series feature with the current cabin temperature. The operation delay value is obtained by fitting the duration in the behavior time series feature with the current cabin temperature. The temperature response intensity is calculated as a weighted combination of the temperature sensitivity value and the operation delay value. The temperature correction factor is obtained by multiplying the temperature response intensity with the operation frequency in the behavioral time sequence characteristics. The environmental impact coefficient is obtained by normalizing the temperature correction factor and the distribution density in the behavioral space characteristics.

4. The method according to claim 3, characterized in that, The current cabin temperature is obtained. The temperature sensitivity value is obtained by fitting the time interval in the behavioral time series features to the current cabin temperature. The operation delay value is obtained by fitting the duration in the behavioral time series features to the current cabin temperature. Obtain the current cabin temperature and the reference temperature of the smart retail cabin, and calculate the temperature deviation between the current cabin temperature and the reference temperature. Extract time interval sequences from behavioral temporal features, calculate the average time interval of the time interval sequence, calculate the deviation between each time interval value in the time interval sequence and the average time interval, and multiply the temperature deviation by the deviation to obtain the temperature response offset. The temperature response offset and the time interval sequence are fitted with a polynomial to obtain the polynomial coefficients. The polynomial coefficients and the temperature deviation are then summed by a power weight to obtain the temperature sensitivity value. Extract the duration sequence from the behavioral temporal features, calculate the decay rate between adjacent duration values ​​in the duration sequence, and calculate the temperature effect decay coefficient by ratio of temperature deviation to decay rate. The exponential parameter is obtained by performing an exponential fitting operation on the temperature effect attenuation coefficient and the duration sequence. The operating delay value is obtained by exponentially powering the exponential parameter and the temperature deviation.

5. The method according to claim 1, characterized in that, The behavioral spatial features are weighted and adjusted based on the environmental impact coefficient to obtain the adjusted features. The location of the user's target product is then predicted based on the adjusted features, including: Spatial movement distance and distribution density are extracted from behavioral spatial features. The environmental impact coefficient is multiplied by the spatial movement distance to obtain the adjusted spatial movement distance. The environmental impact coefficient is multiplied by the distribution density to obtain the adjusted distribution density. The adjusted spatial movement distance and the adjusted distribution density are combined to form the adjusted features. The adjusted spatial movement distance in the adjusted features is coupled with the time interval of effective interaction events in the user behavior sequence to generate spatiotemporal coupling weights. The spatiotemporal coupling weights are applied to the adjusted distribution density of the adjusted features to obtain the spatiotemporal weighted density distribution. Extract density peak regions from the spatiotemporal weighted density distribution, and calculate the consistency of the projection between the product location coordinates and the movement direction formed by continuous effective interaction events in the user behavior sequence within the density peak regions. The product location coordinates with the highest projection consistency are selected as the predicted target location. The spatial distance between the predicted target location and the product location coordinates of the last valid interaction event in the user behavior sequence is calculated. The spatial distance is then compared with the adjusted spatial movement distance in the adjusted features. When the ratio is less than a preset distance threshold, the predicted target location is determined as the user's target product location.

6. The method according to claim 5, characterized in that, Extracting density peak regions from the spatiotemporal weighted density distribution, and calculating the consistency of the projection between the product location coordinates and the movement directions formed by consecutive valid interaction events in the user behavior sequence within these density peak regions, includes: Extract the product location coordinate sequence corresponding to consecutive valid interaction events from the user behavior sequence, perform curve fitting on the product location coordinate sequence to obtain the movement trajectory curve, and calculate the tangent vector of the movement trajectory curve at the product location coordinate of the last valid interaction event as the movement direction. A connection vector is constructed by linking the coordinates of each product position in the spatiotemporal weighted density distribution with the coordinates of the product position in the last valid interaction event. The cosine of the angle between the connection vector and the direction of movement is calculated as the direction weight. The direction-aware density distribution is obtained by multiplying the density values ​​corresponding to the coordinates of each product position in the spatiotemporal weighted density distribution with the direction weight. Extract spatially connected regions with density values ​​greater than a preset density threshold from the direction-aware density distribution as density peak regions; The maximum value of the directional weight corresponding to the commodity location coordinates within the density peak region is determined as the projection consistency.

7. The method according to claim 1, characterized in that, The lighting area and shelf location are determined based on the user's target product location. When the lighting area and shelf location meet the distance condition, the control dependency relationship between the lighting equipment and the shelf drive equipment is established, including: Based on the location of the user's target product, query the corresponding shelf identifier, extract the coordinates of the shelf identifier as the shelf location, extract the physical dimension parameters of the shelf corresponding to the shelf location, and calculate the lighting coverage boundary to determine the lighting area based on the physical dimension parameters of the shelf. Calculate the spatial overlap rate between the lighting area and the shelf location, and calculate the lighting adaptation index by weighting the spatial overlap rate with the merchandise density of the shelf location. When the lighting adaptation index is greater than the average lighting adaptation index of all shelf locations within the lighting area, it is determined that the lighting area and the shelf location meet the distance condition. When the distance condition between the lighting area and the shelf location is met, the lighting equipment identifier corresponding to the lighting area and the shelf drive equipment identifier corresponding to the shelf location are extracted. The control state change sequence of the lighting equipment identifier and the response action sequence of the shelf drive equipment identifier are time-series mapped to generate a state-action dependency chain. In the state-action dependency chain, the brightness of the lighting equipment reaches a stable state as the start-up trigger condition for the shelf drive equipment. The lighting equipment identifier is controlled and bound to the shelf drive equipment identifier to establish a control dependency relationship between the lighting equipment and the shelf drive equipment.

8. A linkage control system for a spatial intelligent retail cabin, used to implement the method as described in any one of claims 1-7, characterized in that, include: The data acquisition unit is used to collect user location data, product operation data, and shelf load-bearing data within the intelligent retail cabin. The event matching unit is used to perform spatiotemporal matching between user location data in user interaction data and product operation data to obtain preliminary interaction events. By comparing the temporal consistency between the time of the preliminary interaction event and the time of data change in product information, falsely identified events are filtered out, and valid interaction events are output. The behavior analysis unit is used to construct user behavior sequences based on effective interaction events and extract behavioral temporal and spatial features. It then correlates the behavioral temporal features with the current cabin temperature to calculate the environmental impact coefficient. The feature adjustment unit is used to weight and adjust the behavioral space features according to the environmental impact coefficient to obtain the adjusted features, and predict the location of the user's target product based on the adjusted features; The dependency establishment unit is used to determine the lighting area and shelf location based on the location of the user's target product. When the lighting area and shelf location meet the distance condition, the control dependency relationship between the lighting equipment and the shelf drive equipment is established. The instruction control unit is used to send control instructions to the lighting equipment according to the control dependency, monitor the status feedback of the lighting equipment, and send control instructions to the rack drive equipment using the status feedback as a trigger condition.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images