An intelligent pet feeder based on eating weight curve learning and a control method thereof

Abstract

The application discloses an intelligent pet feeder based on food weight curve learning and a control method thereof. The feeder comprises a main shell, a grain storage bin, a spiral feeding mechanism, a grain outlet, a weighing food tray, a weighing sensor and a control board. The control method comprises the following steps: A1, collecting N times of food weight curves W(t); A2, calculating the slope dW / dt and the residual ratio r in real time; A3, dynamically feeding according to the slope ΔM=f(dW / dt, r); A4, abnormal feeding stop determination (dW / dt is approximately equal to 0 for T consecutive minutes); and A5, writing history and pushing an APP.

Description

Technical Field

[0001] This invention relates to the field of pet supplies and intelligent control technology, specifically to an intelligent pet feeder and its control method based on real-time weight curve analysis of the feeding process. Background Technology

[0002] Most existing smart pet feeders use a microcontroller and a weighing sensor to feed pets at fixed times or in fixed amounts: ① feeding in multiple sessions according to a preset schedule; ② dispensing a single serving of food in a preset amount (e.g., 30g / serving). The disadvantages of this approach are: ① Pets' daily appetite fluctuates, making it easy to overfeed or underfeed with a fixed amount; ② It cannot detect eating too fast or too slow habits; ③ In multi-cat households, a single serving can easily be monopolized by a dominant pet. Summary of the Invention

[0003] The purpose of this invention is to provide an adaptive feeding control method based on the real-time feeding weight curve W(t), which enables the feeder to dynamically determine when and how much food to add based on the pet's actual eating speed (dW / dt) and the proportion of food remaining in the dish (r), and has the ability to issue an alarm for abnormal eating.

[0004] Beneficial effects: ① Avoids excessive feeding at once, preventing food from being consumed alone; ② Feeds food in small portions as needed, reducing food oxidation and spoilage; ③ Provides early warning of decreased appetite, combined with home health management. Attached Figure Description

[0005] Figure 1 shows a schematic diagram of the feeder of the present invention on the left (1 main shell; 2 grain storage bin; 3 screw feeding mechanism; 4 grain outlet; 5 weighing tray; 51 weighing sensor; 6 control board); and a control flowchart on the right (steps A1 to A5). Detailed Implementation

[0006] Example 1: 5L intelligent grain feeder. The main shell (1) is injection molded from ABS, and the grain storage bin (2) has a magnetic sealing cover on the top; the screw feeding mechanism (3) is driven by a stainless steel auger screw with a diameter of 28mm, discharging 1.5g of grain per turn. Each of the four corners of the weighing tray (5) is equipped with a strain gauge sensor with a range of 5kg, which is output to the ESP32-S3 control board (6) via HX711 ADC.

[0007] Control method: Sample W(t) at 1Hz, smooth it using a 5-point moving average, and then calculate the difference to obtain dW / dt; Example parameters: S_min = 0.6g / s (active threshold), ε = 0.05g / s (fasting threshold), T = 12 minutes, N = 14 days. When r ≤ 0.20 and the moving average of dW / dt ≥ S_min, trigger function ΔM = 8|dW / dt| + 12(1-r) (grams) is used for supplementary feeding.

[0008] Comparative Case Study: Taking a 4kg adult domestic cat as an example, with an average daily food intake of 80g over 14 consecutive days. In the control group (using a timed, measured feeder, 4 times daily x 20g), food was fed too early between meals, resulting in leftover food clumping. In the experimental group (using this invention), the average number of feedings per meal was 2.7, with each meal ranging from 4.8 to 15.2g. The percentage of pets finishing their food within 30 minutes increased from 71% in the control group to 93%.

Claims

1. A smart pet feeder, characterized in that: It includes a main shell (1), a grain storage bin (2), a screw feeding mechanism (3) located below the grain storage bin, a grain outlet (4), a weighing tray (5) located below the grain outlet, at least two weighing sensors (51) located below the weighing tray, and a control board (6); the control board (6) is electrically connected to the screw feeding mechanism (3) and the weighing sensors (51) respectively.

2. The intelligent pet feeder according to claim 1, characterized in that: The weighing sensor (51) is a strain gauge type with a symmetrical arrangement at four corners, with a weighing limit of 2kg and a resolution of ≤0.5g.

3. The intelligent pet feeder according to claim 1, characterized in that: The control board (6) includes a Wi-Fi / BLE module, which can be used to set the total daily feeding amount and the maximum single feeding amount via a mobile APP.

4. A control method for the intelligent pet feeder according to any one of claims 1 to 3, characterized in that: Includes the following steps: A1. During each feeding process, the weight W(t) of the weighing tray (5) is collected at a sampling rate of 1Hz, and the historical curve W_k(t) of the last N times (N≥7) is collected; A2. The current feeding slope dW / dt is calculated in real time, and the residual ratio r=W_now / W_init is defined as the ratio of the current residual amount to the initial amount; A3. When r ≤ 0.20 and the moving average of dW / dt ≥ S_min (active threshold), the control board (6) drives the screw feeding mechanism (3) to replenish ΔM=f(dW / dt, r), where f is a monotonically increasing function ΔM=k_1×|dW / dt|+k_2×(1-r); A4. When |dW / dt| < ε (stop feeding threshold) and r > 0 for T consecutive minutes (T≥10), the feeding of this meal is stopped, and the remaining amount of food is recorded in history; A5. The current curve W(t) is recorded. The trigger parameters are written to storage and pushed to the APP via Wi-Fi / BLE.

5. The control method according to claim 4, characterized in that: In step A3, the coefficients k_1 and k_2 of the function f are obtained by self-learning through least squares fitting of the curvature of the feeding curves from the past N historical feeding cycles.

6. The control method according to claim 4, characterized in that: In step A4, the determination of whether a pet has stopped eating also incorporates camera liveness detection, and a pet is considered to have truly stopped eating only if its head is not detected for 5 consecutive minutes.

7. The control method according to claim 4, characterized in that: The daily total feeding limit protection mechanism: when the cumulative feeding amount exceeds 95% of the daily limit set by the APP, the function f in A3 will automatically decay to 0.

8. The control method according to claim 4, characterized in that: When the slope of the current curve deviates significantly from the average slope of the last N curves ±2σ, a suspected abnormal eating alert is sent to the APP.

Images