Intelligent pet feeding device control system based on Internet of Things

Abstract

The invention discloses an intelligent pet feeding device control system based on the Internet of Things, and relates to the technical field of intelligent pet feeding, and the system comprises a data collection module, an edge calculation module, an execution control module, a cloud intelligent analysis module, a mobile terminal control module and a safety communication module. The data acquisition module is used for acquiring feed weight, pet approaching state, environment temperature and humidity, illumination and image information; the edge calculation module is used for determining the instant feeding demand quantity and controlling the discharging duration according to the weight change, the environmental parameters and the pet behaviors; the cloud analysis module calculates daily energy demand and total feeding amount according to the weight, activity intensity and feed heat of the pet, and optimizes the proportion of each time period through historical feeding data; the system supports encrypted communication and offline emergency control, and the mobile terminal can monitor feeding parameters in real time and remotely set the feeding parameters. The control system of the device can dynamically adjust the feeding amount according to the state of the pet, and safe, accurate and individualized pet feeding management is achieved.

Description

Technical Field

[0001] This invention relates to the field of intelligent pet feeding technology, and more specifically, to a control system for an intelligent pet feeding device based on the Internet of Things. Background Technology

[0002] With the improvement of residents' living standards and the increase in the number of urban single-person households, the number of pets in households continues to grow. Pets have high requirements for regular eating habits, nutritional balance, and health and safety in daily care. However, traditional feeding methods still mainly rely on manual feeding at fixed times and in fixed quantities. Due to factors such as the owner's work and rest schedule, going out, and business trips, problems such as irregular feeding times, excessive or insufficient feed, and spoiled feed often occur. This not only affects the pet's health but may also cause food waste and environmental hygiene hazards.

[0003] Most existing automatic feeding devices on the market use a timed motor drive structure, achieving periodic feeding through mechanical timers or simple microcontroller control. While these devices reduce the burden of manual feeding to some extent, they have the following shortcomings:

[0004] (1) The control logic is simple and can only feed according to the set time. It cannot be personalized according to the pet's real-time behavior, weight, activity level or environmental changes.

[0005] (2) The equipment relies heavily on local timed control and lacks cloud data analysis and remote interaction functions, making it difficult to achieve dynamic optimization;

[0006] (3) There is a lack of identification and feedback mechanisms for pet eating behavior. When pets refuse to eat, spill food, or the equipment is clogged, there is no way to actively alarm or automatically correct the situation.

[0007] (4) Most products have not established a systematic data acquisition and algorithm decision-making closed loop, and cannot achieve multi-source perception, intelligent learning and adaptive adjustment;

[0008] (5) The security of communication and data transmission between devices is insufficient, and they are susceptible to external interference or information leakage risks.

[0009] With the popularization of Internet of Things (IoT) technology, artificial intelligence (AI) algorithms, and mobile internet, pet management models based on multi-sensor information fusion, cloud-based intelligent analysis, and mobile remote control have gradually become a research hotspot. Utilizing sensors to collect information on pets' eating habits, weight, temperature, humidity, and images, analyzing energy needs and behavioral characteristics through cloud-based algorithm models, and then combining this with terminal interaction to achieve dynamic feeding and health management has become an important development direction for the smart home ecosystem.

[0010] Therefore, there is an urgent need for an IoT-based intelligent pet feeding device control system to solve these problems. Summary of the Invention

[0011] The purpose of this invention is to solve the technical problems mentioned in the background section and to provide a control system for a smart pet feeding device based on the Internet of Things.

[0012] The above-mentioned objective of the present invention is achieved through the following technical solution:

[0013] A control system for an IoT-based smart pet feeding device includes a data acquisition module, an edge computing module, an execution control module, a cloud-based intelligent analysis module, a mobile terminal control module, and a secure communication module.

[0014] The data acquisition module is used to obtain the pet's eating status and environmental parameters, including feed weight, initial weight, eating time, pet proximity status, ambient temperature, humidity, light intensity, and camera recognition information.

[0015] The edge computing module is used to analyze the collected data, determine the real-time feeding demand based on changes in feed weight, the pet's proximity status, and environmental correction parameters, and adjust the feeding time according to feed density and discharge flow rate.

[0016] The execution control module is used to drive the discharge motor or the electronically controlled valve, and release the feed according to the control signal of the edge computing module;

[0017] The cloud-based intelligent analysis module is used to calculate the daily energy requirement and total feeding amount based on the pet's weight, activity intensity, and feed calorie parameters, and dynamically adjust the feeding ratio and control parameters for each time period in combination with the pet's historical feeding data.

[0018] The cloud-based intelligent analysis module sends the optimized parameter weights to the edge computing module to update the local decision-making logic;

[0019] The mobile terminal control module is used to display the device status and feeding records, and to receive remote commands from the user;

[0020] The secure communication module is used to establish encrypted communication channels between the device, the cloud, and the mobile terminal, and to verify data integrity.

[0021] As a preferred technical solution of the present invention: the cloud-based intelligent analysis module determines the daily energy requirement of the pet based on the metabolic weight model, and the energy requirement is positively correlated with the pet's weight and activity intensity; the cloud-based intelligent analysis module calculates the total daily feeding amount based on the unit calorie parameter of the feed, providing basic data for subsequent feeding allocation and behavior prediction.

[0022] As a preferred technical solution of the present invention: the cloud-based intelligent analysis module allocates the total daily feeding amount according to the proportion of time periods; the feeding ratio of each time period is adaptively adjusted according to the pet's eating habits, activity time, and environmental conditions to keep the total ratio normalized; the cloud-based intelligent analysis module continuously corrects the weight value of each time period through historical behavior data to form an individualized dynamic feeding plan.

[0023] As a preferred technical solution of the present invention: the cloud-based intelligent analysis module adopts a behavior prediction model that combines temporal convolutional networks and long short-term memory networks to predict and analyze the feeding trend of pets; the model input features include feed weight, pet proximity status, ambient temperature and humidity, light intensity, feeding time and pet weight, and the output is the next feeding time and recommended feeding amount; the model parameters are iteratively updated through the error backpropagation algorithm to continuously improve the prediction accuracy.

[0024] As a preferred technical solution of the present invention: when the edge computing module experiences communication failure with the cloud or network interruption, it automatically activates an offline emergency strategy; the emergency strategy generates an emergency feeding amount based on the historical feeding records cached locally and after trend correction of the average value of the most recent feedings, so that the system can still maintain stable operation in the offline state.

[0025] As a preferred technical solution of the present invention: the camera recognition module identifies the frequency of the target pet's appearance in the feeding area by collecting image frame sequences of the feeding area; when the detected appearance frequency is lower than a set threshold, the system determines that the pet is not eating normally and sends an alarm message to the user through the mobile terminal.

[0026] As a preferred technical solution of the present invention: the secure communication module adopts a message queue telemetry transmission protocol and a transport layer security encryption mechanism to realize encrypted data transmission between the device, the cloud and the mobile terminal; and ensures the authenticity and integrity of the data source through a signature verification method based on hash message authentication code.

[0027] As a preferred technical solution of the present invention: the mobile terminal control module has a feedback adjustment function, which performs gain correction according to the deviation between the target feeding amount set by the user and the actual feedback amount from the device; the feedback adjustment process realizes the smooth adjustment of feeding parameters and prevents pets' eating disorders caused by sudden changes.

[0028] As a preferred technical solution of the present invention: the system has a self-learning function. When there is a deviation between the actual feeding amount detected by the weighing sensor and the predicted feeding amount, the error value is calculated and the model parameters are updated accordingly. The updated parameters are trained and optimized in the cloud and then sent back to the edge computing module to correct the local control strategy and improve the prediction accuracy.

[0029] Compared with the prior art, the present invention has the following beneficial effects:

[0030] 1. This invention, by incorporating a feed weight sensor, an infrared distance sensor, an ambient temperature and humidity sensor, and a camera recognition unit, can simultaneously collect data on the pet's eating status and environmental conditions. The edge computing module determines the real-time feeding demand based on changes in feed weight, pet proximity signals, and environmental correction parameters, and controls the feeding time by combining feed density and discharge flow rate, thus enabling dynamic adjustment of the feeding process based on real-time data. This structure improves feeding response speed and quantitative accuracy, avoiding overfeeding or underfeeding caused by fixed-duration feeding in traditional timed devices.

[0031] 2. The cloud-based intelligent analysis module of this invention calculates the daily energy requirement and total feeding amount based on the pet's weight, activity intensity, and feed calorie parameters, and optimizes the proportion of each time period through historical feeding data, making the daily feeding allocation more in line with individual differences; the system adopts a message queue telemetry transmission protocol and a transmission layer encryption mechanism to establish a secure data channel between the device, the cloud, and the mobile terminal, and has an offline emergency strategy. When communication is interrupted, it can automatically control the feeding based on historical records to ensure the continuity of equipment operation and the security of data transmission.

[0032] 3. This invention introduces a model parameter update mechanism between the cloud and the edge. When there is a deviation between the actual feeding amount and the predicted amount, the algorithm weight can be automatically adjusted according to the error to achieve adaptive optimization of the feeding strategy. The mobile terminal provides parameter feedback and user interaction interface, allowing users to remotely correct the set values ​​according to the pet's status. The system architecture has good scalability and can be further configured with functions such as voice control, automatic food replenishment detection, environmental anomaly prompts, and multi-pet recognition. It is suitable for various family scenarios and different pet types, and has high promotion and application value. Attached Figure Description

[0033] Figure 1 This invention proposes a system block diagram for a pet intelligent feeding device control system based on the Internet of Things.

[0034] Figure 2 This is a flowchart of a method for controlling a smart pet feeding device based on the Internet of Things, as proposed in this invention. Detailed Implementation

[0035] To make the objectives, technical solutions, and advantages of this invention clearer, the following description is provided in conjunction with embodiments and appendices. Figure 1-2 The present invention will be further described in detail below. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0036] The technical solution of this invention will be further explained below with reference to a practical and implementable intelligent pet feeding device control system. This embodiment is implemented in an environment with an existing Internet of Things (IoT) home network. The home router provides 2.4GHz Wi-Fi access, the feeding device has a built-in wireless module and can actively connect to the cloud platform, allowing users to view and control it through a mobile application. The purpose of this embodiment is to automate, visualize, adjust, and learn pet feeding, so that the feeding amount is not a fixed time, but can automatically change according to the pet's weight, activity, environment, and even eating behavior.

[0037] The overall structure of the device includes several sensing units installed inside the feeder body, an edge control board for local computing, a discharging actuator linked to the feed hopper, a cloud processing platform connected to the Internet, a mobile application for human-machine interaction, and a communication encryption module to ensure the security of the entire link. The outer shell structure of the device can adopt a common form of upper storage hopper + lower discharging bowl. An electric discharging port or electric screw is set at the bottom of the storage hopper, and a weighing bowl is set below the discharging port. A weight sensor is installed below the weighing bowl, a camera is set above or to the side of the bowl, and an infrared rangefinder or infrared photodiode is set at the pet entry area to detect whether a pet is approaching.

[0038] During operation, the device first periodically collects various operational data through its data acquisition module. The main data types include: current feed weight. Initial feed weight The actual duration of this meal Is the pet present in the feeding area? Ambient temperature Ambient humidity Light intensity

[0039] And pet image features or recognition results identified by the camera. .in, The readings are provided in real time by a weighing sensor installed below the weighing bowl, and the unit is grams (g). This is the weight of the feed in the bowl at the start of this feeding, also in grams (g), which can be recorded by the control panel at the start of a feeding action; The duration of this feeding behavior is measured in seconds (s) and can be calculated from the moment the pet first approaches to the moment it leaves. The value represents the pet's proximity status; it is 1 when the pet is in the feeding area and 0 when it is not. It is dimensionless. The ambient temperature is expressed in degrees Celsius (°C). The ambient humidity is expressed as a percentage (%). Light intensity, measured in lux (Ix); The image recognition result can be a Boolean value indicating whether the target pet was recognized, or a specific pet ID calculated by the camera algorithm. Here, it's sufficient to know that it's the output of the image recognition module. The edge computing module is located inside the feeding device and uses an embedded processor with Wi-Fi capability, such as an ESP32 or a combination of STM32 and ESP8266. This module performs a quick calculation on the above data locally, primarily to achieve real-time control of "feeding when it arrives," "supplementing when it's low," and "not overfeeding." The edge computing module first calculates the current feed consumption rate based on the weighing data. The calculation formula used in this embodiment is:

[0040] ;

[0041] in, Feed consumption rate, expressed in grams per second (g / s); This refers to the feed weight at the start of this feeding session, in units of... ; The current weight of the feed in the bowl, in grams; This refers to the duration of the feeding behavior, measured in seconds (s). The formula means that if a pet eats quickly, it will lose more weight in the same amount of time. If the size is large, it indicates a strong current need for food; if it merely sniffs the bowl and its weight remains essentially unchanged, then... It's very small, so we don't need to add more feed.

[0042] Then, the edge computing module needs to take into account whether the pet is actually there and whether the environment is suitable for feeding, so it calculates an instant feeding demand index:

[0043]

[0044] in, This is a dimensionless index representing the immediate feeding demand. and Weighting coefficients, either pre-set locally or distributed from the cloud, are dimensionless and used to balance the arrival of a pet with the need to adjust environmental conditions. This indicates the pet's proximity status; a value of 1 is assigned when the pet is present, and 0 is assigned when the pet is absent. This is an environmental correction function that reduces the amount of food given when the temperature is too high, the humidity is too high, or the light is too dim, or appropriately increases the amount of food given during cold winter months. The output of this function is dimensionless. Thus, when the pet is not present... Even if the weight in the bowl decreases slightly, no additional food will be added due to error; when the pet arrives and environmental conditions are met, It will take a relatively high value.

[0045] After obtaining the feed consumption rate and demand index, this implementation method calculates the immediate feeding amount that actually needs to be issued to the implementing agency according to the following formula:

[0046] ;

[0047] in, The amount to be fed immediately is expressed in grams (g). The above formula represents the feed consumption rate, expressed in g / s. This is a dimensionless index representing the immediate feeding demand. The meaning of this formula is that multiplying "eats quickly" by "actually needs to be fed now" gives "how much supplementation is needed." If the pet suddenly leaves... ,thereby decline, They will also drop, to avoid airdrops.

[0048] The execution control module is connected to the mechanical discharge mechanism of the hopper. This can be either a DC geared motor driving a screw to extrude feed, or a solenoid valve controlling the descent of the pelleted feed. In this embodiment, the execution control module calculates the feed output based on the above parameters. To deduce the time when the material discharge was activated. The formula is:

[0049] ;

[0050] in, The discharge time is expressed in seconds (s). The bulk density of the feed currently in use is expressed in grams per cubic centimeter (g / cm3), and can be pre-calibrated through experiments. The average volumetric flow rate at the discharge port, expressed in cubic centimeters per second (cm³ / s), is determined by the motor speed and the discharge port size and is a constant that can be calibrated at the factory. Thus, the edge plate only needs to specify the motor's on-time to output the designated mass with relatively accurate accuracy.

[0051] The pet's food. To ensure that the diet is not always the same, but rather suited to the pet's current condition, the system performs a longer-term, higher-dimensional calculation in the cloud. First, the cloud calculates, based on a generally accepted pet energy requirement model, how much the pet should theoretically eat in a day. The formula is:

[0052] ;

[0053] in, The daily energy requirement for pets is expressed in kilocalories (kcal). This is the species energy constant, dimensionless, commonly used for small dogs, 70, and for cats, 100. The value is the pet's actual weight, expressed in kilograms (kg); an index of 0.75 is a common metabolic body mass index. The activity correction factor is dimensionless; 1.2 is suitable for quiet pets, 1.6 for active pets, and even higher for working dogs. After calculating the daily energy requirement, it needs to be divided by the calories in the food to determine the daily feed amount in grams.

[0054] ;

[0055] in, This refers to the total daily feeding amount, expressed in grams (g). The above formula represents the daily energy requirement, in kcal. This refers to the caloric value per unit mass of the feed used, expressed in kilocalories per gram (kcal / g). This formula ensures that the feeding amount is based on energy requirements, not just a random 300g. However, the feed is not consumed all at once throughout the day; the cloud computing system needs to... Divide meals into breakfast, lunch, and dinner, or multiple small meals; for this purpose, a time slot allocation function is introduced.

[0056] ;

[0057] in, For at any time The amount of food given in this meal is expressed in grams (g). Total daily amount, in grams; The time-slot assignment function is dimensionless; its summation across all time slots equals 1. To make the assignment learnable and not static, this function can be written in the form of softmax:

[0058] ;

[0059] in, This represents the weight of feeding preferences for the current time period, dimensionless; the denominator is the sum of the exponents of all time periods, ensuring that the proportion allocated to each time period is positive and the sum is 1. The cloud can adjust this based on the actual amount the pet has eaten over a past period. For example, always eating a very clean breakfast can improve your morning health. .

[0060] To make the allocation smarter, the cloud can also use a model combining a temporal convolutional network and an LSTM to predict eating behavior, concatenating features from the device and the collar camera to form an input vector:

[0061] ;

[0062] in, The input feature vector; This represents the current feed weight. The presence status of the pet; , , These are temperature, humidity, and light intensity, respectively. This is mealtime; Let the pet's weight be the starting point. The model's output is approximately when and how much to feed the next pet, which can be written as:

[0063] ;

[0064] in, The model output vector contains the predicted feeding amount and time; A neural network trained in the cloud; These are the weight parameters of the network, which are tuned using gradient descent during training.

[0065] ;

[0066] in, For the current parameter; The learning rate is dimensionless. The loss function is typically the square of the difference between the predicted feed amount and the actual feed amount. This represents the gradient of the loss function with respect to the parameters. The trained parameters are then sent to the device so that local decisions can also use the latest model.

[0067] Considering that home networks are always online, this implementation incorporates an emergency strategy. When the edge control board detects that data has not been successfully sent to the cloud within a certain period, it calculates a "non-problematic" feeding amount from the locally cached recent feeding records. The formula is:

[0068] ;

[0069] in, This is the amount to be fed offline, in grams. This is the average of the feeding amounts from the most recent feedings, in grams. and These are the actual feeding amounts for the previous and the one before that, respectively, in grams. This is a correction factor, typically between 0.2 and 0.5, dimensionless, used to slightly match the recent trend. This way, even if the cloud connection is lost, the device won't completely stop feeding or feed too much at once.

[0070] To determine whether the pet is actually eating, and not just a child reaching out or a cat simply passing by, this implementation places a simple camera in front of the device and uses a lightweight front-end recognition algorithm to calculate "how many frames in a given period of time actually show the pet's head in the bowl area," thereby obtaining a feeding confidence level.

[0071] ;

[0072] in, The confidence level for eating is dimensionless; The number of image frames in the bowl area that identify the head or mouth of the target pet within the sampling time, dimensionless; This represents the total number of image frames acquired during this period, dimensionless. If... If the value is less than 0.3, it can be assumed that the feeding did not actually occur, and the device will send an alarm to the user via the APP or record a "refusal to eat" event.

[0073] Because this device is an IoT device, it needs to transmit data with the cloud and mobile phones, so communication security must be ensured. This implementation uses MQTT as a lightweight message queue protocol, the transmission channel uses TLS encryption, and an HMAC-SHA256 signature is added at the application layer. The signature formula is:

[0074] ;

[0075] Here, Sig is the final generated message digest used for integrity verification; Msg is the message content to be transmitted; and Key is the unique key assigned to this device. The cloud also uses the same key to calculate the message upon receiving it. If the two results match, it means the message has not been tampered with. The user-side mobile application can view real-time weight, temperature, humidity, camera feed, and historical feeding records, and can also modify "how much to feed" and "how many times to feed." When the user adjusts the target amount in the app, the device needs a smooth tracking process. A simple feedback adjustment model can be used here:

[0076] ;

[0077] in, This is the control quantity currently issued to the executing agency, which can be understood as the maximum allowable single discharge quantity. The target feeding value set by the user on the app; This represents the actual amount of food fed as detected by the device. The feedback gain coefficient is dimensionless and used to control the adjustment speed. This way, when the user changes the parameters, the device doesn't change abruptly, but rather smoothly approaches the user's desired value.

[0078] To ensure the system becomes increasingly accurate with use, this implementation also incorporates an error-driven self-learning update process. Once a feeding actually occurs, the weighing sensor can obtain an estimate of the "actual amount consumed." A comparison can be made between the cloud and local data to identify and predict the values. Even if they are different, there will be errors:

[0079] ;

[0080] in, This is the error in the amount of feed given, in grams; This represents the actual consumption amount, in grams (g). This is the predicted feed amount, in grams. Then, use this error to adjust the model weights or the local weight table:

[0081] ;

[0082] in, These are the old weight or parameter values; The updated weights or parameter values; The correction rate is dimensionless and used to control the impact of a single error on the overall model. The updated parameters are uploaded to the cloud, and then distributed to other devices of the same model or used for global model retraining. This allows the device to gradually adapt to the pet's real-world needs over long-term operation.

[0083] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A control system for an IoT-based intelligent pet feeding device, characterized in that, It includes a data acquisition module, an edge computing module, an execution control module, a cloud-based intelligent analysis module, a mobile terminal control module, and a secure communication module; The data acquisition module is used to obtain the pet's eating status and environmental parameters, including feed weight, initial weight, eating time, pet proximity status, ambient temperature, humidity, light intensity, and camera recognition information. The edge computing module is used to analyze the collected data, determine the real-time feeding demand based on changes in feed weight, the pet's proximity status, and environmental correction parameters, and adjust the feeding time according to feed density and discharge flow rate. The execution control module is used to drive the discharge motor or the electronically controlled valve, and release the feed according to the control signal of the edge computing module; The cloud-based intelligent analysis module is used to calculate the daily energy requirement and total feeding amount based on the pet's weight, activity intensity, and feed calorie parameters, and dynamically adjust the feeding ratio and control parameters for each time period in combination with the pet's historical feeding data. The cloud-based intelligent analysis module sends the optimized parameter weights to the edge computing module to update the local decision-making logic; The mobile terminal control module is used to display the device status and feeding records, and to receive remote commands from the user; The secure communication module is used to establish encrypted communication channels between the device, the cloud, and the mobile terminal, and to verify data integrity.

2. The control system for an IoT-based intelligent pet feeding device according to claim 1, characterized in that, The cloud-based intelligent analysis module determines the pet's daily energy requirements based on a metabolic weight model, and the energy requirements are positively correlated with the pet's weight and activity level. The cloud-based intelligent analysis module also calculates the total daily feed amount based on the unit calorie parameters of the feed, providing basic data for subsequent feeding allocation and behavior prediction.

3. The control system for an IoT-based intelligent pet feeding device according to claim 1, characterized in that, The cloud-based intelligent analysis module allocates the total daily feeding amount according to the proportion of time periods; the feeding ratio of each time period is adaptively adjusted according to the pet's eating habits, activity time, and environmental conditions to keep the total ratio normalized; the cloud-based intelligent analysis module continuously corrects the weight value of each time period through historical behavior data to form an individualized dynamic feeding plan.

4. The control system for an IoT-based intelligent pet feeding device according to claim 2, characterized in that, The cloud-based intelligent analysis module employs a behavior prediction model that combines temporal convolutional networks and long short-term memory networks to predict and analyze pets' eating trends. The model's input features include feed weight, pet proximity status, ambient temperature and humidity, light intensity, feeding duration, and pet weight. The output is the next feeding time and recommended feeding amount. The model parameters are iteratively updated using an error backpropagation algorithm to continuously improve prediction accuracy.

5. The control system for an IoT-based intelligent pet feeding device according to claim 1, characterized in that, When the edge computing module experiences communication failures with the cloud or network interruptions, it automatically activates an offline emergency strategy. The emergency strategy generates an emergency feeding amount based on the historical feeding records cached locally and after trend correction of the average values ​​of the most recent feedings, so that the system can still maintain stable operation in an offline state.

6. The control system for an IoT-based intelligent pet feeding device according to claim 1, characterized in that, The camera recognition module identifies the frequency of the target pet's appearance in the feeding area by collecting image frame sequences of the feeding area; when the detected frequency is lower than a set threshold, the system determines that the pet is not eating normally and sends an alarm message to the user through the mobile terminal.

7. The control system for an IoT-based intelligent pet feeding device according to claim 1, characterized in that, The secure communication module employs a message queue telemetry transmission protocol and a transport layer security encryption mechanism to achieve encrypted data transmission between the device, cloud, and mobile terminal; and ensures the authenticity and integrity of the data source through a signature verification method based on hash message authentication codes.

8. The control system for an IoT-based intelligent pet feeding device according to claim 1, characterized in that, The mobile terminal control module has a feedback adjustment function, which performs gain correction based on the deviation between the target feeding amount set by the user and the actual feedback amount from the device. The feedback adjustment process enables smooth adjustment of feeding parameters to prevent pets from experiencing dietary disorders due to sudden changes.

9. The control system for an IoT-based intelligent pet feeding device according to claim 1, characterized in that, The system has a self-learning function. When there is a deviation between the actual feeding amount detected by the weighing sensor and the predicted feeding amount, the error value is calculated and the model parameters are updated accordingly. The updated parameters are trained and optimized in the cloud and then sent back to the edge computing module to correct the local control strategy and improve the prediction accuracy.

Images