Intelligent canteen meal supply amount intelligent regulation system and method

Abstract

The application provides a kind of intelligent canteen meal supply intelligent regulation and control system and method.Therein, the system includes: multi-modal sensor unit, computing device and executor, wherein the computing device includes coupling control unit and control module group, and the control module group includes compensator, suppressor and trigger, respectively for: according to dynamic correlation weight matrix, proportional-integral control algorithm is compensated in real time to integral time constant, and the required meal supplement amount is calculated;Based on inventory weight change data and dynamic consumption rate fluctuation data in sliding time window, calculate the feedforward compensation amount in future time window;When real-time consumption rate exceeds the preset consumption rate threshold, generate the pulse width modulation signal with duty cycle positively correlated with consumption rate gradient;The executor executes meal supplement operation according to meal supplement amount, feedforward compensation amount and pulse width modulation signal to form a closed-loop control loop.The application can realize the accuracy of intelligent regulation and control of meal supply.

Description

Technical Field

[0001] This application relates to the field of catering management technology, and in particular to a smart canteen food supply intelligent control system and method. Background Technology

[0002] In a smart canteen setting, precise control of food supply is a core requirement for improving dining efficiency and reducing waste. Specifically, this scenario needs to meet the requirements of dynamic response capabilities, collaborative consumption modeling, and adaptive control.

[0003] To address these needs, existing technologies employ three solutions. The first is a single-sensor threshold triggering system. A weighing sensor is deployed at the bottom of the food container. When the inventory weight falls below a preset threshold, a replenishment signal is triggered, providing a basic inventory alarm and replacing manual inspection. The second solution involves timed replenishment based on historical consumption data, creating a replenishment plan and executing replenishment at fixed time windows (e.g., every 2 hours). The third solution is fixed-parameter proportional-integral-derivative (PID) control.

[0004] However, existing technologies suffer from problems such as untimely supply response or over-supplementation. Summary of the Invention

[0005] This application provides a smart canteen food supply intelligent control system and method to solve the problem of poor accuracy in intelligent control in the prior art.

[0006] In a first aspect, embodiments of this application provide a smart canteen food supply intelligent control system, including:

[0007] The multimodal sensing unit includes a weighing sensor, a plate radio frequency reader and writer, and a visual recognition device installed at the bottom of the food container. It is used to collect real-time data on changes in the inventory weight of the food, food identification data, and dynamic consumption correlation data. The dynamic consumption correlation data represents the time-series relationship of coordinated consumption among the food items.

[0008] The computing device includes a coupling control unit and a control module group. The coupling control unit is used to establish a dynamic correlation weight matrix through the real-time data stream of the multimodal sensing unit, and to calculate the real-time consumption rate based on the gradient feedback signal of the change in inventory weight and the deviation of the replenishment amount of the actuator. The dynamic correlation weight matrix is ​​used to characterize the collaborative consumption characteristics between meals in the closed-loop control cycle.

[0009] The control module group includes a compensator, an inhibitor, and a trigger;

[0010] The compensator is used to compensate the integral time constant of the proportional-integral (PI) control algorithm in real time according to the dynamic correlation weight matrix, and to calculate the required amount of supplementary feed.

[0011] The suppressor is used to calculate the feedforward compensation amount in the future time window based on the inventory weight change data and dynamic consumption rate fluctuation data within the sliding time window;

[0012] The trigger is used to generate a pulse width modulation (PWM) signal whose duty cycle is positively correlated with the consumption rate gradient when the real-time consumption rate exceeds a preset consumption rate threshold.

[0013] The actuator is used to drive the feeding robot arm inside the actuator to perform feeding operations based on the feeding amount, feedforward compensation amount and pulse width modulation signal via industrial bus protocol, and update the corresponding inventory data through the multimodal sensing unit to form a closed-loop control loop.

[0014] Optionally, the coupling control unit includes a weight matrix generation module and a real-time consumption rate prediction module connected in sequence;

[0015] The coupling control unit is used to establish a dynamic correlation weight matrix through the real-time data stream of the multimodal sensing unit, and to calculate the real-time consumption rate based on the gradient feedback signal of inventory weight change and the deviation of the actuator's replenishment amount. The dynamic correlation weight matrix is ​​used to characterize the coordinated consumption characteristics between meals within the closed-loop control cycle, including:

[0016] The weight matrix generation module is used to generate a dynamic weight matrix based on real-time collected food identification data and dynamic consumption correlation data through dynamic coupling relationship analysis within a sliding time window. The update period of the dynamic weight matrix is ​​consistent with the sampling period of the proportional-integral control algorithm.

[0017] The real-time consumption rate prediction module is used to calculate the instantaneous consumption rate within the current time window based on the gradient change of the inventory weight change data, and to perform a convolution operation between the instantaneous consumption rate and the dynamic weight matrix based on the error transfer function of a proportional-integral controller to obtain the real-time consumption rate.

[0018] Optionally, the weight matrix generation module includes a data processing submodule, a timestamp recording submodule, a hardware accelerator, and a matrix operation submodule;

[0019] The weight matrix generation module is used to generate a dynamic weight matrix based on real-time collected food identification data and dynamic consumption correlation data, through dynamic coupling relationship analysis within a sliding time window, including:

[0020] The data processing submodule is used to group and cache the meal identification data within the same meal selection combination based on the length of the time window, and generate dynamic consumption-related event data packets.

[0021] The timestamp recording submodule is used to add a corresponding timestamp to each of the dynamic consumption associated event data packets, and to trigger an incremental update of the coupling coefficient when the timestamp interval is less than a preset control period.

[0022] The hardware accelerator is used to encode the food identification data in real time through a parallel computing architecture, generate food feature vectors, and dynamically update the combined feature vectors based on the sampling period of the multimodal sensing unit.

[0023] The matrix operation submodule is used to perform tensor product operation on the combined feature vector and historical consumption rate data to generate a dynamic weight matrix that conforms to the industrial fieldbus data frame structure.

[0024] Optionally, the real-time consumption rate prediction module is used to calculate the instantaneous consumption rate within the current time window based on the gradient change of the inventory weight change data, and to perform a convolution operation between the instantaneous consumption rate and the dynamic weight matrix based on the error transfer function of a proportional-integral controller to obtain the real-time consumption rate, including:

[0025] Design a convolution kernel function based on the error transfer function of a proportional-integral controller;

[0026] By using a convolution kernel function, the instantaneous consumption rate is discretely convolved with the dynamic weight matrix to obtain the real-time consumption rate. The real-time consumption rate and the replenishment amount deviation are used to adjust the proportional gain and integral gain of the error transfer function.

[0027] Optionally, the matrix operation submodule is used to perform a tensor product operation on the combined feature vector and historical consumption rate data to generate a dynamic weight matrix conforming to the industrial fieldbus data frame structure, including:

[0028] The matrix operation submodule is used to perform tensor product operation on the combined feature vector and historical consumption rate data to generate an initial weight matrix;

[0029] The matrix operation submodule is also used to normalize the initial weight matrix, and encapsulate the normalized weight matrix according to the data frame format of the industrial fieldbus protocol to obtain the dynamic weight matrix.

[0030] Optionally, the executor further includes a task scheduling module;

[0031] The actuator is used to perform a meal replenishment operation based on the meal replenishment amount, the feedforward compensation amount, and the pulse width modulation signal, including:

[0032] The task scheduling module is used to perform multi-task scheduling based on the supplementary meal amount, feedforward compensation amount and pulse width modulation signal to obtain the task scheduling result.

[0033] The robotic arm for replenishing food is used to perform corresponding food replenishment operations based on the task scheduling results.

[0034] Optionally, the task scheduling module is used to perform multi-task scheduling based on the supplementary meal amount, feedforward compensation amount, and pulse width modulation signal to obtain a task scheduling result, including:

[0035] The task scheduling module is used to assign corresponding priority coefficients to the supplementary meal amount, the feedforward compensation amount, and the pulse width modulation signal, respectively.

[0036] The task scheduling module is also used to sort the tasks according to their priority coefficients in descending order and generate task scheduling results.

[0037] Secondly, embodiments of this application provide a method for intelligent control of food supply in a smart canteen, including:

[0038] The real-time data stream of the multimodal sensing unit is obtained by the coupling control unit, a dynamic correlation weight matrix is ​​established, and the real-time consumption rate is calculated based on the gradient feedback signal of the change in inventory weight and the deviation of the replenishment amount of the actuator. The dynamic correlation weight matrix is ​​used to characterize the collaborative consumption characteristics between meals in the closed-loop control cycle.

[0039] The compensator compensates the integral time constant of the proportional-integral control algorithm in real time based on the dynamic correlation weight matrix, and calculates the required amount of supplementary feed.

[0040] The suppressor calculates the feedforward compensation amount for future time windows based on inventory weight change data and dynamic consumption rate fluctuation data within a sliding time window.

[0041] When the real-time consumption rate exceeds a preset consumption rate threshold, a pulse width modulation signal with a duty cycle positively correlated with the consumption rate gradient is generated by a trigger.

[0042] The replenishment amount, the feedforward compensation amount, and the pulse width modulation signal are sent to the actuator, so that the actuator drives the replenishment robot arm in the actuator to perform the replenishment operation according to the replenishment amount, the feedforward compensation amount, and the pulse width modulation signal through the industrial bus protocol, and updates the corresponding inventory data through the multimodal sensing unit to form a closed-loop control loop.

[0043] Thirdly, embodiments of this application provide a computing device, including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are invoked and executed by the processing component to implement a smart canteen meal supply intelligent control method as described in any of the second aspects.

[0044] Fourthly, embodiments of this application provide a computer storage medium storing a computer program, which, when executed by a computer, implements a smart canteen meal supply intelligent control method as described in any of the second aspects.

[0045] This application embodiment provides a smart canteen meal supply intelligent control system, the system comprising:

[0046] A multimodal sensing unit includes a weighing sensor, a plate RFID reader / writer, and a visual recognition device, all mounted on the bottom of the food container. These are used to collect real-time data on changes in the food's inventory weight, food identification data, and dynamic consumption correlation data. The dynamic consumption correlation data characterizes the temporal relationship of coordinated consumption among the food items. The computing device includes a coupling control unit and a control module group. The coupling control unit establishes a dynamic correlation weight matrix based on the real-time data stream from the multimodal sensing unit and calculates the real-time consumption rate based on the gradient feedback signal of inventory weight changes and the deviation of the actuator's replenishment amount. The dynamic correlation weight matrix characterizes the coordinated consumption characteristics among the food items within the closed-loop control cycle. The control module group includes a compensator, an inhibitor, and a trigger. The system comprises: a compensator for real-time compensation of the integral time constant of the proportional-integral control algorithm based on a dynamic correlation weight matrix, and a suppressor for calculating the feedforward compensation amount in a future time window based on inventory weight change data and dynamic consumption rate fluctuation data within a sliding time window; a trigger for generating a pulse width modulation signal with a duty cycle positively correlated with the consumption rate gradient when the real-time consumption rate exceeds a preset consumption rate threshold; and an actuator for driving the replenishment robot arm within the actuator to perform replenishment operations via an industrial bus protocol based on the replenishment amount, feedforward compensation amount, and pulse width modulation signal, and updating the corresponding inventory data through the multimodal sensing unit to form a closed-loop control circuit.

[0047] This application embodiment achieves dynamic capture of food inventory, identification information, and consumption time sequence through the fusion of three-dimensional data from weighing sensors, RFID readers, and visual recognition. This overcomes the limitations of a single sensing mode, establishing a spatiotemporal correlation mapping between weight change gradients and consumption behavior. A dynamic correlation weight matrix is ​​used to model the collaborative consumption characteristics of food items, providing a multi-dimensional feature space for intelligent replenishment. A time-varying parameter compensation mechanism based on proportional-integral control enables dynamic matching of control algorithm parameters with consumption rate. A sliding time window mechanism combined with feedforward compensation calculation effectively addresses non-steady-state fluctuations in consumption demand. Pulse width modulation signal generation achieves gradient response of replenishment intensity, forming multi-timescale collaborative control. A three-in-one control architecture of "compensation-suppression-trigger" is constructed, simultaneously meeting the dual requirements of steady-state regulation and transient response. An actuator driven by an industrial bus protocol achieves a millisecond-level closed-loop response from replenishment decision to robotic arm action. Continuous iterative updates of dynamic consumption correlation data form an intelligent inventory management system with self-learning capabilities.

[0048] Furthermore, a dynamic correlation weight matrix is ​​constructed in real time through multimodal sensing data streams. The dynamic coupling relationship between food item identification data and consumption-related events is analyzed based on a sliding time window. Hardware accelerators are used to encode and update feature vectors, and tensor product operations are used to generate a dynamic weight matrix conforming to the industrial bus protocol. Simultaneously, combining the gradient feedback of inventory weight changes with the error transfer function of the proportional-integral controller, convolution operations are employed to fuse instantaneous consumption rates and the weight matrix, achieving modeling of the collaborative consumption characteristics of food items within a closed-loop cycle and real-time consumption rate prediction. The matrix update cycle is strictly synchronized with the control system sampling. The dynamic weight matrix accurately represents the temporal relationship of collaborative consumption among food items. The hardware-accelerated parallel computing architecture compresses feature processing latency to the millisecond level, enabling the system to dynamically perceive the correlation impact of combined food selection behavior on inventory. Combined with a convolutional prediction algorithm based on controller error transfer, consumption rate prediction errors are reduced. Furthermore, the standardized matrix data frames of the industrial bus protocol enable highly reliable real-time interaction between control commands and actuators, ultimately forming an adaptive closed-loop response capability to nonlinear demand fluctuations in complex catering consumption scenarios, improving the accuracy of replenishment timing matching compared to traditional methods.

[0049] These or other aspects of this application will become more apparent in the following description of the embodiments. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 A schematic diagram of the structure of an intelligent control system for the food supply of a smart canteen, provided in an embodiment of this application;

[0052] Figure 2 A flowchart illustrating a method for intelligent control of food supply in a smart canteen, as provided in this application embodiment;

[0053] Figure 3 This is a schematic diagram of the structure of a computing device provided in an embodiment of this application. Detailed Implementation

[0054] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0055] In some of the processes described in the specification, claims, and accompanying drawings of this application, multiple operations appearing in a specific order are included. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or may be executed in parallel. The operation numbers, such as 11, 12, etc., are merely used to distinguish different operations and do not themselves represent any execution order. Furthermore, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should be noted that the descriptions such as "first," "second," etc., in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to different types.

[0056] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0057] Figure 1 This is a schematic diagram of the structure of a smart canteen food supply intelligent control system provided in an embodiment of this application. Figure 1 As shown, the system includes:

[0058] The multimodal sensing unit 10 includes a weighing sensor, a plate radio frequency reader and writer and a visual recognition device installed at the bottom of the food container, for real-time collection of food inventory weight change data, food identification data and dynamic consumption correlation data, and the dynamic consumption correlation data characterizes the time sequence relationship of collaborative consumption between food items.

[0059] The weighing sensor converts weight changes into electrical signals, which are then converted into weight data using existing algorithms. The consumption is calculated based on the weight difference; for example, an initial weight of 1kg decreasing to 0.8kg indicates a consumption of 200g. The RFID reader for the plates reads / writes label information on the plates using radio frequency signals, such as food identification codes and categories. The visual recognition device distinguishes different food items based on features like shape and color. Furthermore, the visual recognition device can be a recognition model incorporating a camera and YOLOv5 to acquire food identification data and dynamic consumption correlation data.

[0060] The computing device 11 includes a coupled control unit and a control module group. The coupled control unit is connected to the multimodal sensing unit via an industrial fieldbus. It is used to establish a dynamic correlation weight matrix through the real-time data stream of the multimodal sensing unit, and to calculate the real-time consumption rate based on the gradient feedback signal of the inventory weight change and the deviation of the actuator's replenishment amount. The dynamic correlation weight matrix is ​​used to characterize the collaborative consumption characteristics of the food items in the closed-loop control cycle.

[0061] The control module group includes a compensator, a suppressor, and a trigger. The compensator is used to compensate the integral time constant of the proportional-integral control algorithm in real time based on the dynamic correlation weight matrix, and to calculate the required replenishment amount. The suppressor is used to calculate the feedforward compensation amount in the future time window based on inventory weight change data and dynamic consumption rate fluctuation data within the sliding time window. The trigger is used to generate a pulse width modulation signal with a duty cycle positively correlated with the consumption rate gradient when the real-time consumption rate exceeds a preset consumption rate threshold. The compensator adjusts the control algorithm parameters through the dynamic weight matrix to cope with current fluctuations, while the suppressor predicts future demand based on historical data to compensate in advance. The two work together to optimize real-time error and hysteresis error, respectively. "Duty cycle positively correlated with consumption rate gradient" can be understood as: the duty cycle of the pulse width modulation signal is equal to a preset multiple of the real-time consumption rate gradient. The preset multiple can be determined according to the actual situation, such as 2, 3, etc.

[0062] The dynamic correlation weight matrix quantifies the intensity of synergistic consumption among different food items; larger element values ​​indicate stronger correlation. The sliding time window is a fixed duration, such as a 10-minute data analysis interval, that updates over time. The feedforward compensation is a correction value based on predicted or historical data, used to adjust the replenishment amount in advance, reducing errors caused by system response delays.

[0063] For example, the dynamic association weight matrix is ​​an n-order square matrix, where the elements w ij =α*(C ij / C max )+β*(T ij / T total ), α and β are the coefficients of the corresponding terms, α + β = 1, Cij C represents the number of times that dishes i and j are selected together within the sliding time window. max To maximize the number of common selections, T ij T is the time interval for selecting two meals. total This represents the total window duration. The update cycle of the dynamic weight matrix is ​​strictly synchronized with the sampling cycle of the proportional-integral control algorithm to ensure that the latest weight matrix is ​​loaded at the beginning of each control cycle.

[0064] Actuator 12 is used to drive the replenishment robot arm in the actuator to perform replenishment operation according to the replenishment amount, feedforward compensation amount and pulse width modulation signal through industrial bus protocol, and update the corresponding inventory data through multimodal sensing unit to form a closed loop control loop.

[0065] Among them, pulse width modulation signal is an electrical signal that controls the motor speed or actuator force by adjusting the duty cycle.

[0066] Here is a specific example:

[0067] During the peak lunch hour in the cafeteria, weighing sensors detected a 150g decrease in the weight of braised pork inventory per minute. RFID readers identified that 80% of plates were used simultaneously for both braised pork and potatoes, and visual recognition showed that these two items were frequently used consecutively. The coupling control unit generated a correlation weight of 0.7 between braised pork and potatoes, predicting a real-time consumption rate of 180g / min. The compensator shortened the integral time constant by 20%, calculating a replenishment amount of 200g; the suppressor predicted an additional 50g of feedforward compensation for the next 5 minutes of demand; and the trigger generated a pulse width modulation signal with a 60% duty cycle due to the rate exceeding a threshold. The actuator prioritized replenishing 200g, then added another 50g, completing the replenishment. After updating the inventory data, the next control cycle began.

[0068] This application embodiment uses a multimodal sensing unit to collect real-time food consumption data, combined with a dynamic weight matrix and closed-loop control algorithm, to accurately predict consumption rates and dynamically adjust replenishment strategies. A compensator optimizes the integral time constant to handle fluctuations in coordinated consumption, a suppressor responds to future demands in advance through feedforward compensation, a trigger quickly replenishes food during sudden consumption spikes, and an actuator ensures replenishment efficiency through priority scheduling. Ultimately, this reduces food waste, avoids supply shortages, and improves canteen operational efficiency.

[0069] In one possible embodiment, the coupling control unit includes a weight matrix generation module and a real-time consumption rate prediction module connected in sequence. The coupling control unit is used to establish a dynamic correlation weight matrix based on the real-time data stream from the multimodal sensing unit, and to calculate the real-time consumption rate based on the gradient feedback signal of inventory weight change and the deviation of the actuator's replenishment amount. The dynamic correlation weight matrix is ​​used to characterize the coordinated consumption characteristics among food items within the closed-loop control cycle, including:

[0070] The weight matrix generation module is used to generate a dynamic weight matrix based on real-time collected food identification data and dynamic consumption correlation data through dynamic coupling relationship analysis within a sliding time window. The update period of the dynamic weight matrix is ​​consistent with the sampling period of the proportional-integral control algorithm.

[0071] The sliding time window is a fixed-duration data analysis interval (e.g., 10 minutes) that updates as the system runs. Dynamic coupling relationship analysis quantifies the statistical correlation between different food items being consumed collaboratively within the same time window. The dynamic weight matrix is ​​a normalized and encapsulated weight matrix that can be adjusted in real time to adapt to changes in equipment operating status, used to optimize control strategies.

[0072] The real-time consumption rate prediction module is used to calculate the instantaneous consumption rate within the current time window based on the gradient change of inventory weight data. The instantaneous consumption rate is then convolved with the dynamic weight matrix using an error transfer function based on a proportional-integral controller to obtain the real-time consumption rate.

[0073] The error transfer function is the mathematical model of the proportional-integral controller, taking the error signal as input and the control quantity as output. Convolution is a weighted sum of a matrix and a scalar using a kernel function, used for data fusion and correction.

[0074] Here is a specific example:

[0075] During the lunch rush, weighing sensors detected a 200g decrease in braised pork inventory per minute, RFID readers recorded that 80% of plates simultaneously used both braised pork and rice, and visual recognition devices captured the frequent consecutive use of both. The weight matrix generation module generates a dynamic weight matrix through sliding window analysis, and the real-time consumption rate prediction module convolves the instantaneous rate of 200g / min with the weight matrix, outputting a corrected rate of 220g / min. The proportional-integral controller adjusts the replenishment amount based on this rate, triggering the actuator to replenish 250g of braised pork, including a base amount of 200g and a related compensation of 50g, while the rice replenishment amount increases by 40% simultaneously. The system continuously optimizes the weight and rate prediction through closed-loop updates.

[0076] In this embodiment, the system uses a dynamic weight matrix generation module to accurately quantify the collaborative consumption relationship between dishes. The real-time consumption rate prediction module, combined with the error transfer function of the PI controller, corrects the instantaneous consumption rate to a real-time rate that reflects the influence of related dishes.

[0077] In one possible embodiment, the weight matrix generation module includes a data processing submodule, a timestamp recording submodule, a hardware accelerator, and a matrix operation submodule. The weight matrix generation module is used to generate a dynamic weight matrix based on real-time collected food identification data and dynamic consumption correlation data, through dynamic coupling relationship analysis within a sliding time window, including:

[0078] The data processing submodule is used to group and cache the meal identification data within the same meal selection combination based on the length of the time window, and generate dynamic consumption-related event data packets.

[0079] Among them, the dynamic consumption associated event data packet is a data structure used to record associated dish combinations within the same time window, which includes fields such as dish name pairs and combination frequency.

[0080] The timestamp recording submodule is used to add a corresponding timestamp to each dynamically consumed associated event data packet, and to trigger an incremental update of the coupling coefficient when the timestamp interval is less than a preset control period. The incremental update of the coupling coefficient dynamically adjusts the coefficient of the association strength between dishes based on the time density.

[0081] The hardware accelerator is used to encode food identification data in real time through a parallel computing architecture, generating food feature vectors, and dynamically updating the combined feature vectors based on the sampling period of the multimodal sensing units. The parallel computing architecture utilizes the parallel computing units of field-programmable gate arrays or graphics processing units to achieve high-speed feature encoding. The combined feature vector is a multidimensional vector composed of multiple device operating features, such as temperature, pressure, and rotational speed, used to characterize the overall state of the equipment.

[0082] For example, the hardware accelerator can be implemented using existing chips and may contain 64 parallel processing units, each configured with 128KB of dedicated cache, interacting with the main processor via an existing bus, and the feature vector encoding latency can be less than 2ms.

[0083] The matrix operation submodule is used to perform tensor product operations on the combined feature vector and historical consumption rate data to generate a dynamic weight matrix that conforms to the industrial fieldbus data frame structure.

[0084] Tensor product operation, a vector outer product operation, is used to map features to historical data as a matrix relationship. Industrial fieldbus data frames conform to standardized data encapsulation formats that comply with industrial communication protocols. Historical consumption rate data records time-series data of the equipment's resource consumption rates, such as electrical energy and raw material consumption rates, over a specific period.

[0085] Here is a specific example:

[0086] During the lunch rush, the RFID reader detected three consecutive plates simultaneously using braised pork and rice, prompting the data processing submodule to generate an event packet. The timestamp recording submodule found that the interval between the two events was only 0.8 seconds, less than the 1-second threshold, and increased the coupling coefficient between the braised pork and rice from 0.7 to 0.75. The hardware accelerator encoded the braised pork [1,0] and rice [0,1] into a combined vector [1,0,0,1]. The matrix operation submodule performed a tensor product with the historical rate [200,150] to generate an initial matrix [[200,0,0,200],[150,0,0,150]]. After normalization, this matrix was encapsulated into a data frame and input to the actuator to control a 20% synchronous increase in the amount of food replenished.

[0087] This application utilizes hardware accelerators and parallel computing to meet the real-time processing requirements of high-frequency data (once per second). Timestamp-triggered incremental updates ensure that the weight matrix adjusts rapidly according to actual consumption behavior. Data frame encapsulation allows the weight matrix to directly drive industrial actuators.

[0088] In one possible embodiment, the real-time consumption rate prediction module is used to calculate the instantaneous consumption rate within the current time window based on the gradient change of inventory weight change data, and to perform a convolution operation between the instantaneous consumption rate and the dynamic weight matrix based on the error transfer function of a proportional-integral controller to obtain the real-time consumption rate, including:

[0089] A convolution kernel function is designed based on the error transfer function of a proportional-integral controller. The convolution kernel function is a weight vector designed based on the PI controller parameters, used for data fusion and correction.

[0090] By using a convolution kernel function, the instantaneous consumption rate is discretely convolved with the dynamic weight matrix to obtain the real-time consumption rate. The real-time consumption rate, along with the replenishment deviation, jointly adjusts the proportional gain and integral gain of the error transfer function. The discrete convolution operation involves weighted summation of matrix elements according to the kernel function, achieving data fusion. The replenishment deviation refers to the difference between the target replenishment amount and the real-time consumption rate, driving the adjustment of control parameters.

[0091] Here is a specific example:

[0092] During the dinner peak, the weighing sensor detected an instantaneous consumption rate of 200g / min for braised pork, and the dynamic weight matrix showed a correlation weight of 0.8 between the braised pork and rice. Based on the current PI parameters, with a proportional gain of 1.2 and an integral gain of 0.5, the system generates a convolution kernel [1.2, 0.05] and performs a convolution operation on the instantaneous rate and the weight matrix, outputting a real-time rate of 220g / min. A deviation of 30g from the target replenishment amount of 250g triggers an increase in the proportional gain to 1.5 and the integral gain to 0.6. The updated convolution kernel [1.5, 0.06] is used for the next cycle calculation, making the replenishment amount more accurately match the actual consumption demand.

[0093] This application's embodiments incorporate the correlation between dishes into consumption rate prediction through convolution operations, improving accuracy. Real-time rate and deviation dynamically adjust the PI parameters to optimize replenishment response speed and stability. The adjusted PI parameters are fed back to the convolution kernel design, forming a continuously improving control loop.

[0094] In one possible embodiment, the matrix operation submodule is used to perform a tensor product operation on the combined feature vector and historical consumption rate data to generate a dynamic weight matrix conforming to the industrial fieldbus data frame structure, including:

[0095] The matrix operation submodule is used to perform tensor product operation on the combined feature vector and the historical consumption rate data to generate an initial weight matrix. The initial weight matrix is ​​an unnormalized weight matrix generated by the tensor product operation, which reflects the preliminary weight distribution between the combined feature vector and the historical consumption rate data.

[0096] The matrix operation submodule is also used to normalize the initial weight matrix, and then encapsulates the normalized weight matrix according to the data frame format of the Industrial Fieldbus protocol to obtain the dynamic weight matrix. The Industrial Fieldbus protocol is a standardized protocol for communication between industrial devices, defining data frame structures, transmission rules, etc. The data frame format is a data structure organized according to a specific protocol, typically containing fields such as a frame header, data payload, and checksum to ensure the integrity and correctness of data transmission.

[0097] Here is a specific example:

[0098] A company cafeteria successfully optimized the food supply during peak lunch hours by deploying a dynamic weight matrix system. The system first collects current environmental characteristics: weekday lunchtime coefficient 0.8, rainstorm weather coefficient 0.4, and vegetable demand surge coefficient 0.7, generating a combined feature vector F=[0.8, 0.4, 0.7]; at the same time, it retrieves historical data: braised pork ribs consumed an average of 85 servings, garlic broccoli 120 servings, and tomato scrambled eggs 95 servings, forming a historical consumption vector H=[85, 120, 95].

[0099] A 3×3 initial weight matrix is ​​generated through tensor product operations. After row normalization, the matrix is ​​obtained as [[0.52,0.28,0.42],[0.41,0.34,0.38],[0.39,0.31,0.36]], which is then encapsulated into an industrial data frame conforming to the protocol. When the vision system detects that a 40% drop in the number of diners due to heavy rain, the central controller immediately integrates the dynamic weight matrix with the real-time dining queue data and outputs precise instructions to reduce the supply of braised pork ribs by 30%, maintain the standard for garlic broccoli, and increase the supply of scrambled eggs with tomatoes by 20%. Ultimately, this reduces the daily food waste rate from 19% to 7.2%, while ensuring that the last employee to arrive still receives a complete meal.

[0100] This application embodiment uses a matrix operation submodule to perform tensor product operations on the combined feature vector and historical consumption rate data to generate an initial weight matrix. This initial weight matrix is ​​then normalized and encapsulated using an industrial fieldbus protocol to obtain a dynamic weight matrix. This solution can dynamically adjust equipment control parameters, improve resource utilization, and ensure reliable data transmission in industrial networks, making it suitable for scenarios such as intelligent manufacturing and energy management.

[0101] In one possible embodiment, the actuator further includes a task scheduling module; the actuator is configured to perform a meal replenishment operation based on the meal replenishment amount, the feedforward compensation amount, and the pulse width modulation signal, including:

[0102] The task scheduling module is used to perform multi-task scheduling based on the amount of supplementary food, the amount of feedforward compensation, and the pulse width modulation signal, and obtain the task scheduling result.

[0103] Among them, multi-task scheduling is the process by which the robotic arm allocates execution order and resources among multiple meal replenishment tasks to improve efficiency and response speed.

[0104] The robotic arm for food replenishment is used to perform corresponding food replenishment operations based on task scheduling results.

[0105] The robotic arm that automates meal replenishment operations typically consists of a servo motor, an end effector, and a control system. The servo motor is a motor that precisely controls position and speed, driving the robotic arm's movement.

[0106] Here is a specific example:

[0107] During peak lunch hours, a smart canteen implemented precise meal replenishment through a task scheduling module. When the central prediction system detected that the real-time consumption rate of braised pork exceeded the expected 15%, a dynamic scheduling mechanism was triggered. The task scheduling module simultaneously received three sets of parameters: 1. Replenishment quantity matrix [Braised pork +20 servings / 5min, stir-fried vegetables ±0, Mapo tofu -5 servings / 5min], 2. Feedforward compensation vector [+8,0,-3], and 3. Pulse width modulation signal PWM=15kHz. After calculation by a multi-objective optimization algorithm, a scheduling instruction was generated: prioritize the replenishment of braised pork, simultaneously adjust the pulse interval of the conveyor belt stepper motor to 50ms, and drive the six-axis replenishment robot arm to complete 8 rounds of precise feeding at a constant speed of 0.18m / s. During this period, the torque feedback dynamically compensated for the offset of the end effector caused by the stacking of trays. Finally, the emergency replenishment was completed within 7 minutes and 32 seconds, ensuring that there was no shortage of dishes during the window period, while avoiding the waste of 3.2kg of food caused by excessive replenishment. Compared with the traditional timed replenishment mode, the continuity of meal supply was improved by 38%. This application embodiment dynamically optimizes the meal replenishment strategy through a task scheduling module, reduces hysteresis errors by combining it with feedforward compensation, and precisely controls the robotic arm's movements using PWM signals to achieve efficient and accurate automated meal replenishment. The meal replenishment robotic arm executes operations according to the scheduling results, and a PID algorithm is used to ensure stability, making it suitable for meal replenishment scenarios during peak lunch hours in smart canteens.

[0108] In one possible embodiment, the task scheduling module is used to perform multi-task scheduling based on the supplementary meal amount, the feedforward compensation amount, and the pulse width modulation signal, to obtain the task scheduling result, including:

[0109] The task scheduling module is used to assign corresponding priority coefficients to the replenishment amount, feedforward compensation amount, and pulse width modulation signal, respectively. The priority coefficient is a numerical value that quantifies the urgency of the task; the higher the value, the higher the execution priority. Task urgency is categorized as very urgent, relatively urgent, moderately urgent, and not urgent. The calculation of the priority coefficient is a conventional technique in this field and will not be elaborated upon here.

[0110] The task scheduling module is also used to sort tasks according to their priority coefficients in descending order, generating task scheduling results. The generated scheduling results output an ordered task queue for the executor to execute sequentially. Descending sorting arranges tasks according to their priority coefficients from largest to smallest, for example, PWM signal coefficient 4 > feedforward compensation coefficient 3 > feedforward compensation coefficient 2.4.

[0111] Here is a specific example:

[0112] At the end of the lunch rush, the system detected a 30% shortage of braised pork in stock and predicted increased demand in the next 5 minutes. Simultaneously, a sudden queue triggered a PWM signal. The task scheduling module prioritized the PWM signal, followed by the amount of pork to be replenished, and finally the amount of feedforward compensation. The replenishing robotic arm first replenished 200g of braised pork at high speed with an 80% duty cycle, then replenished 50g at the standard speed, and finally completed the feedforward compensation. If a new PWM signal arrived during this process, the current task response was immediately interrupted.

[0113] This application's embodiments ensure that high-frequency tasks are prioritized by calculating coefficients in real time. Orderly scheduling avoids conflicts in robotic arm movements, improving meal replenishment efficiency. High-priority tasks can interrupt low-priority tasks to meet urgent needs.

[0114] Figure 2 A flowchart illustrating an intelligent control method for the food supply in a smart canteen, provided as an embodiment of this application, is applied to a computing device within an intelligent control system for the food supply in a smart canteen. Figure 2 As shown, the method includes:

[0115] S21. The real-time data stream of the multimodal sensing unit is obtained through the coupling control unit, a dynamic correlation weight matrix is ​​established, and the real-time consumption rate is calculated based on the gradient feedback signal of the change in inventory weight and the deviation of the actuator's replenishment amount. The dynamic correlation weight matrix is ​​used to characterize the collaborative consumption characteristics of the meals in the closed-loop control cycle.

[0116] S22. The compensator compensates the integral time constant of the proportional-integral control algorithm in real time according to the dynamic correlation weight matrix, and calculates the required amount of supplementary feed.

[0117] S23. Calculate the feedforward compensation amount for the future time window based on the inventory weight change data and dynamic consumption rate fluctuation data within the sliding time window using the suppressor.

[0118] S24. When the real-time consumption rate exceeds the preset consumption rate threshold, a pulse width modulation signal with a duty cycle positively correlated with the consumption rate gradient is generated by the trigger.

[0119] S25. Send the replenishment amount, feedforward compensation amount, and pulse width modulation signal to the actuator, so that the actuator can drive the replenishment robot arm in the actuator to perform the replenishment operation according to the replenishment amount, feedforward compensation amount, and pulse width modulation signal through the industrial bus protocol, and update the corresponding inventory data through the multi-modal sensing unit to form a closed-loop control loop.

[0120] Figure 2 The intelligent control method for the supply of food in a smart canteen, as shown in the embodiment, is similar in principle and technical effect to the intelligent control system for the supply of food in a smart canteen, and will not be described in detail here.

[0121] In one possible design, Figure 1 The computing device in the intelligent control system for food supply in a smart canteen, as shown in the embodiment, is... Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32. The processing component 32 includes a coupled control unit and a control module group.

[0122] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are invoked and executed by the processing component 32.

[0123] The processing component 32 is used to: acquire the real-time data stream of the multimodal sensing unit through the coupling control unit, establish a dynamic correlation weight matrix, and calculate the real-time consumption rate based on the gradient feedback signal of inventory weight change and the deviation of the actuator's replenishment amount. The dynamic correlation weight matrix is ​​used to characterize the collaborative consumption characteristics between items within the closed-loop control cycle. The compensator performs real-time compensation on the integral time constant of the proportional-integral control algorithm according to the dynamic correlation weight matrix and calculates the required replenishment amount. The suppressor calculates the feedforward compensation amount in the future time window based on the inventory weight change data and dynamic consumption rate fluctuation data within the sliding time window. When the real-time consumption rate exceeds the preset consumption rate threshold, the trigger generates a pulse width modulation signal with a duty cycle positively correlated with the consumption rate gradient. The replenishment amount, feedforward compensation amount, and pulse width modulation signal are sent to the actuator so that the actuator drives the replenishment robot arm in the actuator to perform the replenishment operation according to the replenishment amount, feedforward compensation amount, and pulse width modulation signal through the industrial bus protocol, and updates the corresponding inventory data through feedback from the multimodal sensing unit to form a closed-loop control loop.

[0124] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above-described method. Alternatively, the processing component may be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above-described method.

[0125] Storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented from any type of volatile or non-volatile storage device or a combination thereof, such as Random Access Memory (RAM), Static Random-Access Memory (SRAM), Electrically Erasable Programmable Read Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Read Only Memory (PROM), Read Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0126] Of course, computing devices may also include other components, such as input / output interfaces, display components, communication components, etc.

[0127] Input / output interfaces provide interfaces between processing components and peripheral interface modules, which can be output devices, input devices, etc.

[0128] The communication components are configured to facilitate wired or wireless communication between computing devices and other devices.

[0129] The computing device can be a physical device or an elastic computing host provided by a cloud computing platform. In this case, the computing device can refer to a cloud server, and the aforementioned processing components, storage components, etc., can be basic server resources rented or purchased from the cloud computing platform.

[0130] This application also provides a computer storage medium storing a computer program, which, when executed by a computer, can perform the above-described functions. Figure 1 The embodiment shown illustrates a method for intelligently controlling the supply of food in a smart canteen.

[0131] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0132] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0133] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0134] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application 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 of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A smart canteen meal serving quantity intelligent regulation system, characterized in that, The method comprises the following steps: A multi-modal sensing unit comprising a weighing sensor arranged at the bottom of a food container, a meal tray radio frequency reader and a visual recognition device, is used to collect real-time inventory weight change data, meal identification data and dynamic consumption correlation data of the food, and the dynamic consumption correlation data represents the time sequence relationship of the collaborative consumption between the food; The computing device comprises a coupling control unit and a control module group, the coupling control unit is used to establish a dynamic correlation weight matrix through the real-time data stream of the multi-modal sensing unit, and calculate a real-time consumption rate based on the gradient feedback signal of the inventory weight change and the deviation of the replenishment amount of the actuator, and the dynamic correlation weight matrix is used to represent the collaborative consumption characteristics between the food within a closed-loop control cycle; The control module group comprises a compensator, an inhibitor and a trigger; The compensator is used to compensate the integral time constant of a proportional-integral control algorithm in real time according to the dynamic correlation weight matrix, and calculate the required replenishment amount; The inhibitor is used to calculate the feedforward compensation amount in the future time window based on the inventory weight change data and the dynamic consumption rate fluctuation data within the sliding time window; The trigger is used to generate a pulse width modulation signal with a duty cycle positively related to the consumption rate gradient when the real-time consumption rate exceeds a preset consumption rate threshold; The actuator is used to drive the replenishment mechanical arm in the actuator to perform a replenishment operation according to the replenishment amount, the feedforward compensation amount and the pulse width modulation signal, and update the corresponding inventory data through the multi-modal sensing unit to form a closed-loop control loop.

2. The system of claim 1, wherein, The coupling control unit comprises a weight matrix generation module and a real-time consumption rate prediction module connected in sequence; The coupling control unit is used to establish a dynamic correlation weight matrix through the real-time data stream of the multi-modal sensing unit, and calculate a real-time consumption rate based on the gradient feedback signal of the inventory weight change and the deviation of the replenishment amount of the actuator, and the dynamic correlation weight matrix is used to represent the collaborative consumption characteristics between the food within a closed-loop control cycle, comprising: The weight matrix generation module is used to generate a dynamic weight matrix based on the real-time collected meal identification data and dynamic consumption correlation data through dynamic coupling relationship analysis within a sliding time window, wherein the update period of the dynamic weight matrix is consistent with the sampling period of the proportional-integral control algorithm; The real-time consumption rate prediction module is used to calculate an instantaneous consumption rate within the current time window according to the gradient change of the inventory weight change data, and perform convolution operation on the instantaneous consumption rate and the dynamic weight matrix based on the error transfer function of the proportional-integral controller to obtain the real-time consumption rate.

3. The system of claim 2, wherein, The weight matrix generation module comprises a data processing submodule, a timestamp recording submodule, a hardware accelerator and a matrix operation submodule; The weight matrix generation module is used to generate a dynamic weight matrix based on the real-time collected meal identification data and dynamic consumption correlation data through dynamic coupling relationship analysis within a sliding time window, comprising: The data processing sub-module is configured to group and cache the meal product identification data in the same meal combination based on the length of the time window, and generate a dynamic consumption correlation event data packet. The timestamp recording sub-module is configured to add a corresponding timestamp to each dynamic consumption correlation event data packet, and trigger an incremental update of the coupling coefficient when the timestamp interval is less than a preset control period. The hardware accelerator is configured to encode the meal product identification data in real time through a parallel computing architecture, generate a meal product feature vector, and dynamically update the combination feature vector based on a sampling period of the multi-modal sensing unit. The matrix operation sub-module is configured to perform a tensor product operation on the combination feature vector and historical consumption rate data, and generate a dynamic weight matrix conforming to an industrial field bus data frame structure.

4. The system of claim 2, wherein, The real-time consumption rate prediction module is configured to calculate an instantaneous consumption rate in the current time window according to the gradient change of the inventory weight change data, perform a convolution operation on the instantaneous consumption rate and the dynamic weight matrix based on an error transfer function of a proportional-integral controller, and obtain a real-time consumption rate, including: designing a convolution kernel function based on the error transfer function of the proportional-integral controller; performing a discrete convolution operation on the instantaneous consumption rate and the dynamic weight matrix using the convolution kernel function to obtain the real-time consumption rate, and adjusting the proportional gain and the integral gain of the error transfer function together with the meal supplement amount deviation.

5. The system of claim 3, wherein, The matrix operation sub-module is configured to perform a tensor product operation on the combination feature vector and historical consumption rate data, and generate a dynamic weight matrix conforming to an industrial field bus data frame structure, including: The matrix operation sub-module is configured to perform a tensor product operation on the combination feature vector and historical consumption rate data, and generate an initial weight matrix. The matrix operation sub-module is further configured to perform normalization processing on the initial weight matrix, encapsulate the normalized weight matrix in a data frame format according to an industrial field bus protocol, and obtain a dynamic weight matrix.

6. The system of claim 1, wherein, The executor further includes a task scheduling module; The executor is configured to perform a meal supplement operation according to the meal supplement amount, the feedforward compensation amount, and the pulse width modulation signal, including: The task scheduling module is configured to perform multi-task scheduling according to the meal supplement amount, the feedforward compensation amount, and the pulse width modulation signal, and obtain a task scheduling result. The meal supplement mechanical arm is configured to perform a corresponding meal supplement operation according to the task scheduling result.

7. The system of claim 6, wherein, The task scheduling module is configured to perform multi-task scheduling according to the meal supplement amount, the feedforward compensation amount, and the pulse width modulation signal, and obtain a task scheduling result, including: The task scheduling module is configured to assign corresponding priority coefficients to the meal supplement amount, the feedforward compensation amount, and the pulse width modulation signal, respectively. The task scheduling module is further configured to sort the priority coefficients in descending order to generate a task scheduling result.

8. A method for intelligent regulation of meal serving quantity in a smart canteen, characterized in that, A computing device applied to a smart canteen meal product supply intelligent regulation and control system, including: The real-time data stream of the multi-modal sensing unit is acquired by the coupling control unit, a dynamic correlation weight matrix is established, and a real-time consumption rate is calculated based on the gradient feedback signal of the inventory weight change and the deviation of the compensating amount of the actuator, wherein the dynamic correlation weight matrix is used to represent the cooperative consumption characteristics between dishes within a closed-loop control cycle; The integral time constant of the proportional-integral control algorithm is compensated in real time by the compensator based on the dynamic correlation weight matrix, and the required compensating amount is calculated; The feedforward compensation amount in the future time window is calculated by the suppressor based on the inventory weight change data and the dynamic consumption rate fluctuation data in the sliding time window; The pulse width modulation signal with the duty cycle positively correlated with the consumption rate gradient is generated by the trigger when the real-time consumption rate exceeds the preset consumption rate threshold; The compensating amount, the feedforward compensation amount, and the pulse width modulation signal are sent to the actuator, so that the actuator drives the compensating mechanical arm in the actuator to perform the compensating operation through the industrial bus protocol according to the compensating amount, the feedforward compensation amount, and the pulse width modulation signal, and the corresponding inventory data is updated through the multi-modal sensing unit to form a closed-loop control loop.

9. A computing device, comprising: The method comprises a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to realize the intelligent regulation and control method of the meal supply amount in the smart cafeteria according to claim 8.

10. A computer storage medium, characterized in that, The computer program is stored in the computer, and when the computer program is executed by the computer, the intelligent regulation and control method of the meal supply amount in the smart cafeteria according to claim 8 is realized.

Images