Pet recall device and automatic recall method

Abstract

The application discloses a pet recall device and an automatic recall method, which comprises the following steps: collecting real-time related data of a pet and related data of an environment where the pet is located; pre-processing the collected data to obtain multi-modal data; processing the multi-modal data by using a trained deep learning model to obtain a feature vector; processing the feature vector by using a trained long short-term memory network to obtain a classification probability of a current behavior state and a prediction of a future behavior trend; generating an optimal recall strategy through hierarchical reinforcement learning, and executing the strategy by an execution component; collecting parameters of the optimal recall strategy, designing a loss function according to a difference between the parameters and actual feedback, and updating parameters of the deep learning model, the long short-term memory network and the hierarchical reinforcement learning by using an optimization algorithm. The application realizes real-time analysis and forward-looking decision of pet behavior and environmental situation, and can realize intelligent decision and execution of a recall instruction in various complex scenes.

Description

Technical Field

[0001] This invention relates to the field of smart pet devices and the Internet of Things (IoT) technology, specifically to a pet recall device and an automatic recall method. Background Technology

[0002] Existing automatic pet recall systems mostly rely on a single trigger mechanism, such as GPS / GPRS positioning or simple sound stimuli. These systems suffer from low recall success rates in complex and ever-changing real-world scenarios (such as high noise, weak signals, insufficient lighting, and volatile pet behavior) and struggle to adapt to changes in the environment and the pet's condition. For example, Chinese invention patent CN109446373B, entitled "A Semi-Free-Range Pet Management System and Method Based on an IoT Platform," primarily focuses on GPS / GPRS positioning and simple feedback, lacking prediction of future pet behavior trends and contextualized strategy optimization. Similarly, Chinese invention patent CN115883610A, entitled "An Intelligent Pet Recall System and Method," relies solely on passive recall based on simple sounds and stimuli, lacking forward-looking decision-making and personalized optimization for complex situations.

[0003] Current traditional solutions fail to fully utilize multi-source sensing data and advanced intelligent algorithms, rendering them ineffective against impending escape or anxiety behaviors, and failing to achieve true dynamic optimization in energy management and data security. Therefore, there is an urgent need for an automated pet recall system with predictive context awareness and reinforcement learning-driven dynamic adaptive strategies to improve recall success rates and user experience. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a pet recall device and an automatic recall method, which integrates multiple modules to achieve comprehensive perception and intelligent decision-making of pet behavior, emotions and environment, significantly improves the intelligence level and practicality of the recall system, and solves the problems of insufficient intelligence, poor environmental adaptability and low recall success rate in the prior art.

[0005] To solve the above technical problems, the present invention adopts the following technical solution:

[0006] A pet recall device includes:

[0007] Collar strap, outer shell, leash, actuator, pet-side smart device, and owner-side control device.

[0008] The outer shell has loopholes at both ends for attaching ropes. One end of each of the two collar straps is secured at the loopholes, and the other end of each collar strap is connected by a buckle.

[0009] The actuators include a loudspeaker, a vibrating device, and an LED. The LED is located on the outer surface of the housing, while the loudspeaker and the vibrating device are located inside the housing.

[0010] The pet-end smart device includes a context awareness module, a behavior recognition and predictive analysis module, a reinforcement learning strategy module, and a self-learning and incremental optimization module, all of which are located inside the outer casing.

[0011] The context awareness module, behavior recognition and predictive analysis module, reinforcement learning strategy module, and self-learning and incremental optimization module are integrated onto the same hardware platform.

[0012] The context-aware module includes a data acquisition unit and a data processing unit.

[0013] Furthermore, the pet-side smart device and the owner-side control device achieve data transmission and control command interaction through a wireless communication protocol.

[0014] Furthermore, the context awareness module and the behavior recognition and prediction analysis module transmit data via an internal bus, the behavior recognition and prediction analysis module and the reinforcement learning strategy module transmit data via an internal bus, and the self-learning and incremental optimization module transmit data via an internal bus to the behavior recognition and prediction analysis module and the reinforcement learning strategy module.

[0015] Furthermore, the data acquisition unit includes a GPS module, a Bluetooth beacon, an inertial measurement unit, a light sensor, a temperature sensor, a humidity sensor, a noise sensor, a heart rate sensor, and a body temperature sensor.

[0016] Furthermore, the master control device is a smartphone app or a smartwatch.

[0017] Furthermore, the present invention also proposes an automatic recall method for pet recall devices, comprising:

[0018] S1. The data acquisition unit of the context awareness module collects the pet's location information, Bluetooth beacon signal strength, pet's acceleration and angular velocity, pet's heart rate and body temperature, noise level, light intensity, temperature and humidity of the pet's environment in real time.

[0019] S2. The data processing unit of the context-aware module preprocesses the data collected by the data acquisition unit to obtain the initial dynamic motion features and calculates the statistical features of the data within the time window. The initial dynamic motion features are divided into fixed-length time periods using the sliding window mechanism, and the features are normalized to obtain multimodal data.

[0020] S3. In the behavior recognition and prediction analysis module, the parameters of the deep learning model and the long short-term memory network are trained using the backpropagation algorithm with the loss function to obtain the trained deep learning model and the long short-term memory network; the trained deep learning model is used to process multimodal data to obtain feature vectors; the trained long short-term memory network is used to process the feature vectors to obtain the classification probability of the current behavior state and the prediction of future behavior trends.

[0021] S4. In the reinforcement learning strategy module, the optimal recall strategy is dynamically generated through hierarchical reinforcement learning, and the strategy is executed by the execution component. The strategy is adjusted and optimized in real time based on the pet's real-time response.

[0022] S5. In the self-learning and incremental optimization module, collect the parameters of the optimal recall strategy, design a loss function based on the difference between the parameters and the actual feedback, use the backpropagation algorithm to calculate the gradients of the deep learning model, long short-term memory network and hierarchical reinforcement learning parameters, and use the optimization algorithm to update the model parameters.

[0023] S6. Repeat steps S3-S5 to obtain the final optimal recall strategy.

[0024] S7. The master control device displays the data collected in the context awareness module and the instruction notification of the optimal recall strategy, and manually adjusts the optimal recall strategy according to the actual situation.

[0025] Furthermore, in step S2, the preprocessing includes: using low-pass filtering to remove high-frequency noise, using high-pass filtering to obtain motion signals whose standard deviation exceeds a preset threshold within 1 second, and then performing denoising and alignment.

[0026] Statistical characteristics include mean and variance.

[0027] Furthermore, in step S3, the convolutional layers in the trained deep learning model are used to extract local spatiotemporal features of multimodal data through local convolution operations, and pooling layers are used to reduce dimensionality to obtain feature maps. These feature maps are then converted into feature vectors through fully connected layers. These vectors include statistical and dynamic features of each window. The trained long short-term memory network is used to process the feature vectors, capturing the information transfer between time windows and the dynamic changes in pet behavior patterns. The feature vectors are then fused and dimensionality reduced through fully connected layers, and the classification probability of the current behavior state and the prediction of future behavior trends are output through the Softmax classification layer.

[0028] Furthermore, in step S4, the hierarchical reinforcement learning includes an upper-layer policy network and a lower-layer policy network. Multimodal data is input into the upper-layer policy network to obtain the parameter search range. Based on this parameter search range, the lower-layer policy network combines the prediction results of the behavior recognition and prediction analysis module and uses reinforcement learning algorithms to weight and score different parameter combinations to obtain the expected recall success rate, response time, and energy consumption. Within the parameter search range, the parameter combination with the highest weighted score or that reaches the set threshold is selected as the optimal recall strategy.

[0029] Based on the optimal recall strategy, if the entropy value of the predicted distribution is >0.5, the recall strategy is updated immediately; otherwise, the recall strategy is updated according to the set frequency.

[0030] Weighted score = w1 × recall success rate - w2 × response time - w3 × average energy consumption.

[0031] The formula for calculating the predicted distribution entropy value H(P) is:

[0032]

[0033] Where p(i) represents the probability of the i-th behavior category.

[0034] Furthermore, step S8 is included: adjusting the parameters of the models in the behavior recognition and prediction analysis module and the reinforcement learning strategy module according to the characteristics of different pets.

[0035] Compared with the prior art, the present invention, employing the above technical solution, has the following technical effects:

[0036] 1. This invention achieves real-time analysis and forward-looking decision-making on pet behavior and environmental situations through predictive context awareness and reinforcement learning-driven dynamic response strategies, enabling intelligent decision-making and execution of recall commands in various complex scenarios.

[0037] 2. The invention has a short response time, ensuring that pets can be recalled in a timely manner.

[0038] 3. This invention significantly improves the recall success rate in complex environments such as high-noise environments, low-light environments, and weak signal areas.

[0039] 4. This invention is the first to apply deep learning and adaptive strategies to the field of pet recall, resulting in a technological breakthrough and innovation. Attached Figure Description

[0040] Figure 1 This is an overall structural diagram of the present invention.

[0041] Figure 2 This is a structural diagram of the internal modules of the pet terminal device of the present invention.

[0042] Figure 3 This is a flowchart illustrating the overall implementation of the present invention. Detailed Implementation

[0043] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0044] To achieve the above objectives, the present invention proposes a pet recall device, such as... Figure 1 As shown, it includes:

[0045] The system comprises a collar, outer shell, leash, actuators, and a pet-side smart device and owner-side control device that transmit data and exchange control commands via wireless communication protocols (such as Bluetooth and Wi-Fi). Communication between devices requires two-way authentication to prevent unauthorized device access and data tampering, ensuring the security and reliability of the system.

[0046] The outer casing has loops at both ends for securing one end of two collar straps, and the other ends of the two collar straps are connected by buckles.

[0047] The actuators include a speaker, a vibration device, and an LED. The LED is located on the outer surface of the housing, while the speaker and vibration device are located inside the housing. The selection of multiple actuators provides multi-sensory feedback, enhancing the perceptibility and responsiveness of recall commands.

[0048] The pet-end smart device includes a context awareness module, a behavior recognition and predictive analysis module, a reinforcement learning strategy module, and a self-learning and incremental optimization module, all of which are located inside the outer casing.

[0049] like Figure 2 As shown, the context-aware module, behavior recognition and predictive analysis module, reinforcement learning strategy module, and self-learning and incremental optimization module are integrated onto the same hardware platform (e.g., the same main control board). The context-aware module and the behavior recognition and predictive analysis module communicate via an internal bus (e.g., SPI, I2C). 2 C) Data transmission is performed between the behavior recognition and prediction analysis module and the reinforcement learning strategy module via an internal bus (e.g., SPI, I...). 2 C) Data transmission is performed between the self-learning and incremental optimization module, the behavior recognition and prediction analysis module, and the reinforcement learning strategy module via an internal bus (e.g., SPI, I). 2 C) Perform data transmission.

[0050] The context awareness module includes a data acquisition unit and a data processing unit; the data acquisition unit includes a GPS module, a Bluetooth beacon, an inertial measurement unit (accelerometer and gyroscope), a light sensor, a temperature sensor, a humidity sensor, a noise sensor (microphone), a heart rate sensor, and a body temperature sensor.

[0051] The behavior recognition and predictive analysis module is used to process motion features using deep learning models, such as convolutional neural networks, to obtain behavior recognition results and future behavior trends.

[0052] The reinforcement learning strategy module is used to determine the most suitable recall strategy for the current situation based on the results of the behavior recognition and prediction analysis module, using hierarchical reinforcement learning. This strategy includes recall content, volume, frequency, and feedback method, and is then handed over to the execution component for execution.

[0053] The owner-side control device is a smartphone app or smartwatch, which allows them to monitor their pet's status, check the recall progress, and intervene manually at any time.

[0054] The smartphone app displays information such as the pet's location, behavior, and physiological data; allows users to set parameters such as the content, volume, and frequency of the recall sound; records the pet's behavior patterns and recall history for users to view and analyze; and allows users to send instant recall commands via the app.

[0055] Smartwatches are used to provide real-time notifications of pet status changes and recall commands; they also allow users to quickly adjust recall strategies while wearing the watch.

[0056] like Figure 3 As shown, the method for automatically recalling pets using a pet recall device includes:

[0057] S1. The data acquisition unit of the context awareness module collects the pet's location information, Bluetooth beacon signal strength, pet's acceleration and angular velocity, pet's heart rate and body temperature, noise level, light intensity, temperature and humidity of the pet's environment in real time.

[0058] S2. The data processing unit of the context-aware module preprocesses the data collected by the data acquisition unit to obtain the initial dynamic motion features. Specifically, this includes: using low-pass filtering to remove high-frequency noise, using high-pass filtering to obtain motion signals whose standard deviation exceeds a preset threshold within 1 second, such as the features of a pet running or jumping, so as to better identify and distinguish different behavior patterns (such as stillness vs. accelerated escape), and then performing noise reduction and alignment.

[0059] The statistical characteristics (including mean and variance) of the data within the time window are calculated. The initial dynamic motion characteristics are divided into fixed-length time periods (such as 2-second windows and 1-second steps) using a sliding window mechanism to ensure a balance between real-time performance and data continuity. The characteristics are then normalized to obtain multimodal data.

[0060] In module S3, the behavior recognition and prediction analysis module, the parameters of a convolutional neural network (CNN) and a long short-term memory (LSM) network are trained using a loss function (such as cross-entropy or mean squared error) via backpropagation. The trained CNN is then used to process multimodal data to obtain feature vectors. The trained LSM network is used to process these feature vectors to obtain the classification probability of the current behavior state and a prediction of future behavior trends. The specific content is as follows:

[0061] By utilizing the convolutional layers in the trained convolutional neural network, local spatiotemporal features of multimodal data are extracted through local convolutional operations. Pooling layers are then used to reduce dimensionality, resulting in feature maps. These feature maps are then transformed into feature vectors through fully connected layers. These vectors include statistical and dynamic features for each window. The trained long short-term memory network is used to process the feature vectors, capturing long-term dependencies (information transfer between time windows) and dynamic changes in pet behavior patterns. These features are then fused and dimensionality reduced through fully connected layers. Finally, a softmax classification layer outputs the classification probability of the current behavioral state and a prediction of future behavioral trends, accurately classifying various pet behavioral states, such as resting, free movement, wandering anxiety, escape tendency, and abnormal behaviors (illness, fatigue, cold), thus identifying potential escape or anxiety behaviors in advance.

[0062] S4. In the reinforcement learning strategy module, the optimal recall strategy is dynamically generated through hierarchical reinforcement learning. This strategy is then executed by the execution component, and the strategy is adjusted and optimized in real time based on the pet's real-time response (such as whether it returns to its owner). The specific content is as follows:

[0063] Layered reinforcement learning consists of an upper-layer policy network and a lower-layer policy network. Multimodal data (noise levels, light intensity, GPS signal strength, uncertainty in behavior prediction, etc.) are input into the upper-layer policy network to obtain a parameter search range (e.g., volume 50%–80%, frequency 1.5kHz–2kHz, LED blinking frequency 1–5 times / second, etc.). Based on this parameter search range, the lower-layer policy network combines the prediction results from the behavior recognition and prediction analysis module and uses reinforcement learning algorithms (such as Q-learning or Policy Gradient) to weight and score different parameter combinations, estimating the balance between recall success rate, energy consumption, and response time, and obtaining the expected recall success rate, response time, and energy consumption. Within the parameter search range, the parameter combination with the highest weighted score or reaching 0.8 is selected as the optimal recall strategy.

[0064] Based on the optimal recall strategy, if the entropy value of the predicted distribution is >0.5, the recall strategy is updated immediately; otherwise, the recall strategy is updated every 5 seconds.

[0065] Weighted score = w1 × recall success rate - w2 × response time - w3 × average energy consumption.

[0066] The formula for calculating the predicted distribution entropy value H(P) is:

[0067]

[0068] Here, p(i) represents the probability of the i-th behavior category. A higher entropy value indicates greater model uncertainty.

[0069] In this embodiment, the optimal recall strategy has the following parameters: recall success rate ≥ 80%, response time ≤ 2 seconds, average energy consumption < 0.1% * total battery capacity, volume = 70%, frequency = 1.8kHz, LED flashing = 3 times / second, vibration intensity = medium. w1 = 0.5, w2 = 0.3, w3 = 0.2.

[0070] In the S5 self-learning and incremental optimization module, the parameters of the optimal recall strategy are collected. Based on the difference between these parameters and actual feedback (including users' manual recall operations, recall success rate, and pet response data), a loss function (such as cross-entropy loss, mean squared error, etc.) is designed. The gradients of the deep learning model, long short-term memory network, and hierarchical reinforcement learning parameters are calculated using the backpropagation algorithm. The model parameters are updated using optimization algorithms (such as Adam, SGD), and online or offline fine-tuning is completed. After using the validation set or real-time monitoring of the effect of the new strategy, it is confirmed whether the update has improved the system performance. If it is effective, the updated model is applied.

[0071] When a large-scale change in data distribution is detected, or when the effectiveness of the system strategy declines significantly, a large-scale model update will be triggered (which can be done at night or when the battery is fully charged).

[0072] S6. Repeat steps S3-S5 to obtain the final optimal recall strategy.

[0073] S7. The master control device displays the data collected in the context awareness module and the instruction notification of the optimal recall strategy, and manually adjusts the optimal recall strategy according to the actual situation.

[0074] S8. Based on the different pets' behavioral habits and physiological characteristics, dynamically adjust and optimize the parameters of the models in the behavior recognition and prediction analysis module and the reinforcement learning strategy module (such as the convolution kernel weights of the convolutional neural network, the gating parameters of the long short-term memory network, and the value function or policy function of the hierarchical reinforcement learning), ultimately changing the behavior recognition output and recall decision to ensure that the system exhibits higher accuracy and reliability in long-term use.

[0075] In the example:

[0076] High-noise environment: In noisy environments, the reinforcement learning strategy module automatically increases the volume of the recall sound or adjusts the audio frequency to overcome the interference of environmental noise and ensure the effective transmission of recall instructions.

[0077] Low-light environment: In nighttime or low-light conditions, the reinforcement learning strategy module combines LED flashing cues and sound recall to enhance the pet's perception of recall commands and avoid recall failures due to insufficient visual recognition. Specifically: if the light level is <5 lx (extremely dark), the LED flashing frequency is 5 times / second, and the volume gain is +30%; if the light level is 5 lx ≤ light level < 10 lx (dim), the LED flashing frequency is 3 times / second, and the volume gain is +20%; if the light level is >10 lx (relatively bright), the normal strategy is maintained (LED frequency = 1 time / second or off, volume gain = 0%~10%).

[0078] In areas with weak or lost GPS signals, the context awareness module automatically activates Bluetooth beacon-assisted positioning and performs position compensation via the inertial measurement unit (IMU) to improve positioning accuracy and ensure reliable retrieval indoors or in areas with complex signals. Specifically, when the GPS signal strength is <-120dBm or there are N consecutive packet losses (e.g., 3), the IMU + Bluetooth positioning compensation is activated. Within a short range, the relative displacement can be calculated based on the acceleration and angular velocity integrals output by the IMU, and the distance between the collar and the Bluetooth base station (phone) can be estimated by combining the Bluetooth wireless signal strength for multi-point positioning fusion. If the Bluetooth wireless signal strength is >-70dBm, the positioning error can be controlled within 2-5 meters; when an acceleration bias exceeding 0.1g or a gyroscope drift exceeding ±0.5° / s is detected, real-time correction is required.

[0079] This invention improves the recall success rate to 88% ± 3% in high-noise environments, 83% ± 2% in low-light environments, and 80% ± 2% in weak signal areas. The average response time is reduced to 1.5 seconds.

[0080] It also allows for module extensibility design, specifically:

[0081] It supports future expansion with more sensors and feedback mechanisms, such as odor emitters and temperature control devices, to enhance the system's functionality and applicability.

[0082] Remote updates and maintenance: OTA (Over-The-Air) technology supports remote software updates and maintenance, ensuring the system always runs with the latest features and optimizations.

[0083] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. An automatic recall method for a pet recall device, characterized in that, include: S1. The data acquisition unit of the context awareness module collects the pet's location information, Bluetooth beacon signal strength, pet's acceleration and angular velocity, pet's heart rate and body temperature, noise level, light intensity, temperature and humidity of the pet's environment in real time. S2. The data processing unit of the context-aware module preprocesses the data collected by the data acquisition unit to obtain the initial dynamic motion characteristics and calculates the statistical characteristics of the data within the time window. The initial dynamic motion features are segmented into fixed-length time periods using a sliding window mechanism, and the features are normalized to obtain multimodal data. S3. In the behavior recognition and prediction analysis module, the parameters of the deep learning model and the long short-term memory network are trained using the backpropagation algorithm with the loss function to obtain the trained deep learning model and the long short-term memory network; the trained deep learning model is used to process multimodal data to obtain feature vectors; the trained long short-term memory network is used to process the feature vectors to obtain the classification probability of the current behavior state and the prediction of future behavior trends. S4. In the reinforcement learning strategy module, an optimal recall strategy is generated through hierarchical reinforcement learning, and this strategy is then executed by the execution component. The strategy is adjusted and optimized in real-time based on the pet's real-time response. Specifically: Hierarchical reinforcement learning includes an upper-layer policy network and a lower-layer policy network. Multimodal data is input into the upper-layer policy network to obtain the parameter search range. Based on this parameter search range, the lower-layer policy network combines the prediction results of the behavior recognition and prediction analysis module and uses reinforcement learning algorithms to weight and score different parameter combinations to obtain the expected recall success rate, response time, and energy consumption. Within the parameter search range, the parameter combination with the highest weighted score or that reaches a set threshold is selected as the optimal recall strategy. Based on the optimal recall strategy, if the entropy value of the predicted distribution is >0.5, the recall strategy is updated immediately; otherwise, the recall strategy is updated at the set frequency. Weighted score = w1 × recall success rate - w2 × response time - w3 × average energy consumption; The formula for calculating the predicted distribution entropy value H(P) is: ； Where p(i) represents the probability of the i-th behavior category; S5. In the self-learning and incremental optimization module, collect the parameters of the optimal recall strategy, design the loss function based on the difference between the parameters and the actual feedback, use the backpropagation algorithm to calculate the gradients of the deep learning model, long short-term memory network and hierarchical reinforcement learning parameters, and use the optimization algorithm to update the model parameters. S6. Repeat steps S3-S5 to obtain the final optimal recall strategy; S7. The master control device displays the data collected in the context awareness module and the instruction notification of the optimal recall strategy, and manually adjusts the optimal recall strategy according to the actual situation.

2. The automatic recall method for the pet recall device according to claim 1, characterized in that, In step S2, the preprocessing includes: using low-pass filtering to remove high-frequency noise, using high-pass filtering to obtain motion signals whose standard deviation exceeds a preset threshold within 1 second, and then performing noise reduction and alignment. Statistical characteristics include mean and variance.

3. The automatic recall method for the pet recall device according to claim 2, characterized in that, In step S3, the convolutional layers in the trained deep learning model are used to extract local spatiotemporal features of multimodal data through local convolution operations, and the dimensionality is reduced by pooling layers to obtain feature maps. These feature maps are then converted into feature vectors through fully connected layers. These vectors include the statistical and dynamic features of each window. The trained Long Short-Term Memory network is used to process the feature vectors, capturing the information transfer between time windows and the dynamic changes in pet behavior patterns. The data is then fused and reduced in dimensionality through a fully connected layer, and the classification probability of the current behavior state and the prediction of future behavior trends are output through a Softmax classification layer.

4. The automatic recall method for the pet recall device according to claim 1, characterized in that, It also includes step S8: adjusting the parameters of the models in the behavior recognition and prediction analysis module and the reinforcement learning strategy module according to the characteristics of different pets.

5. A pet recall device applied to the automatic recall method of claim 1, characterized in that, include: Collar strap, outer shell, leash, actuator, pet-side smart device and owner-side control device; The outer shell has loopholes at both ends for attaching ropes. One end of each of the two collar straps is secured at the loopholes, and the other end of each collar strap is connected by a buckle. The actuators include a loudspeaker, a vibrating device, and an LED. The LED is disposed on the outer surface of the housing, while the loudspeaker and the vibrating device are disposed inside the housing. The pet-end smart device includes a context awareness module, a behavior recognition and prediction analysis module, a reinforcement learning strategy module, and a self-learning and incremental optimization module, all of which are located inside the outer casing. The context awareness module, behavior recognition and prediction analysis module, reinforcement learning strategy module, and self-learning and incremental optimization module are integrated onto the same hardware platform; The context-aware module includes a data acquisition unit and a data processing unit.

6. The pet recall device according to claim 5, characterized in that, The pet-side smart device and the owner-side control device communicate via a wireless communication protocol to exchange data and control commands.

7. The pet recall device according to claim 5, characterized in that, The context awareness module and the behavior recognition and prediction analysis module transmit data via an internal bus. The behavior recognition and prediction analysis module and the reinforcement learning strategy module transmit data via an internal bus. The self-learning and incremental optimization module transmits data via an internal bus to the behavior recognition and prediction analysis module and the reinforcement learning strategy module.

8. The pet recall device according to claim 5, characterized in that, The data acquisition unit includes a GPS module, Bluetooth beacon, inertial measurement unit, light sensor, temperature sensor, humidity sensor, noise sensor, heart rate sensor, and body temperature sensor.

9. The pet recall device according to claim 5, characterized in that, The master control device is a smartphone app or a smartwatch.

Images