A method for synchronizing the actions of multiple AI pet devices based on an emotion state model.

Abstract

This invention discloses a method for synchronizing the actions of multiple AI pet hardware devices based on an emotion state model, comprising the following steps: Emotion state detection: collecting and calculating the raw data required for emotion state, calculating the current emotion value through the emotion state model, and determining the corresponding emotion level; Action sequence generation: selecting the corresponding action template from the action template library according to the emotion level, and performing linear interpolation adjustment based on the current emotion value to generate a complete action timeline sequence; Clock synchronization: achieving precise clock synchronization and action sequence distribution between devices; achieving microsecond-level clock calibration through an anchor point time synchronization protocol, and distributing the action sequence to all slave devices; Local action execution: receiving the synchronized action sequence, superimposing the device's unique personalized offset, and driving the hardware to execute the action. This invention dynamically adjusts action parameters according to the social context, enabling the device behavior to possess emotional levels and contextual awareness.

Description

Technical Field

[0001] This invention relates to the fields of intelligent toys, collaborative control of IoT devices, emotion computing and multi-device synchronization technology, and specifically to a method for synchronizing the actions of multiple AI pet hardware devices based on an emotion state model. Background Technology

[0002] With the rapid development of artificial intelligence (AI) and the Internet of Things (IoT) technologies, smart toys have evolved from single-device interaction to a new stage of multi-device collaborative interaction. Consumers' expectations for smart toys are no longer limited to standalone entertainment functions, but rather they hope that multiple devices can exhibit social behaviors similar to those of real biological groups, such as "walking together," emotional resonance, and synchronized movements.

[0003] However, existing technologies have significant shortcomings in multi-device collaboration. Most solutions rely on simple command synchronization or preset action patterns, lacking emotionally driven dynamic behavior and failing to simulate the emotional resonance and synchronized behavior in real-world biological social interactions. Specifically, existing solutions suffer from technical bottlenecks in the following aspects:

[0004] • Lack of an emotionally driven dynamic behavior generation mechanism

[0005] • The time synchronization accuracy between multiple devices is insufficient to support a visually consistent experience.

[0006] • All devices perform the exact same actions, lacking individual differences and exhibiting a distinctly mechanical feel.

[0007] • It relies on cloud servers for inter-device coordination and cannot function in offline scenarios.

[0008] • Lack of dynamic fluctuation mechanisms that simulate biological respiratory rhythms.

[0009] The prior art closest to this invention includes the following.

[0010] 1. US11057236 - Multi-device synchronous control system

[0011] This patent proposes a centralized command distribution architecture where a master controller broadcasts action commands to multiple slave devices. All slave devices execute the same sequence of actions upon receiving the same command. The synchronization of this scheme relies on the arrival time of the broadcast commands and does not involve precise clock calibration.

[0012] 2. US20200082301 (IBM) - Social Learning System

[0013] IBM's proposed social learning system coordinates the behavior of multiple devices through a cloud server. Devices upload their status to the cloud, which calculates the appropriate responses based on pre-defined social rules and then distributes these responses to the devices for execution. This solution relies entirely on cloud computing and network connectivity.

[0014] 3. US10671903 - IoT Device Group Control Method

[0015] This patent describes a method for controlling a group of IoT devices based on preset scenarios. The system presets several behavioral scenarios (such as "welcome mode" and "party mode"). When the triggering conditions are met, all devices in the group switch to the same preset behavioral mode and execute a fixed sequence of actions.

[0016] 4. Existing BLE synchronization solutions

[0017] Common BLE device synchronization solutions in the industry employ time synchronization protocols similar to NTP (Network Time Protocol), measuring network latency and calibrating the clock through multiple message round trips. The synchronization accuracy of such solutions is typically in the 10-50 millisecond range.

[0018] Based on the above existing technologies, it mainly adopts the following three technical routes:

[0019] (1) Centralized control: One master device sends instructions, and other devices receive and execute the same actions. The advantage is that it is simple to implement, but the disadvantage is that it lacks flexibility and individual differences.

[0020] (2) Preset mode switching: The device switches between several preset behavior modes and has no dynamic generation capability. The advantage is high reliability, but the disadvantage is that the behavior mode is fixed and highly predictable.

[0021] (3) Cloud-based coordination: relies on cloud servers for coordination between devices. The advantage is strong computing power, but the disadvantages are network latency and offline unavailability.

[0022] The disadvantages of the existing technologies along the above technical routes include:

[0023] 1. Lack of emotional drive: The existing solutions have fixed action parameters and cannot dynamically adjust the action performance according to social context (such as distance between devices, pairing time, interaction frequency, etc.). All devices perform the same preset actions in any situation, lacking emotional depth.

[0024] 2. Insufficient synchronization accuracy: Existing BLE NTP-like solutions have a synchronization accuracy of 10-50 milliseconds, while the human visual perception threshold for motion synchronization is approximately 30 milliseconds. When the time difference between the actions of two devices exceeds this threshold, users can clearly perceive the asynchrony, severely impacting the immersive experience of "traveling together".

[0025] 3. Strong sense of mechanical movement: All devices execute the exact same sequence of actions, with no individual differences. In real biological groups, even when acting synchronously, each individual will have subtle differences in timing, amplitude, and rhythm, and these differences are an important source of biological realism.

[0026] 4. Reliance on the cloud: Cloud-based coordination solutions cannot function in scenarios with high network latency or complete offline conditions, limiting their application scenarios. Smart toys are frequently used in environments with unstable networks (such as outdoors or while traveling).

[0027] 5. Lack of biological realism: Existing solutions lack the dynamic fluctuation mechanism to simulate the respiratory rhythm of organisms. The movements of real organisms are not completely periodic, but include respiratory amplitude fluctuations and tiny random perturbations. This organic dynamic change is a key element of biological realism. Summary of the Invention

[0028] The purpose of this invention is to provide a method for synchronizing the actions of multiple AI pet devices based on an emotional state model, thereby solving the following core technical problems:

[0029] 1. How to construct an emotional state model so that the actions of multiple AI pet devices can be dynamically adjusted according to the social context, thereby achieving emotion-driven collaborative behavior.

[0030] 2. How to achieve microsecond-level (within milliseconds) clock synchronization under BLE communication constraints to meet the stringent requirements of visual synchronization.

[0031] 3. How to introduce individual differences while maintaining an overall sense of synchronization, eliminating the mechanical feeling and enhancing the biological realism.

[0032] 4. How to design a decentralized device coordination architecture so that the system can still work normally in offline scenarios.

[0033] The technical solution of this invention is:

[0034] The method for synchronizing the actions of multiple AI pet devices includes the following steps:

[0035] S1. Emotional State Detection: Collect and calculate the raw data required for emotional state, calculate the current emotional value through the emotional state model, and determine the corresponding emotional level;

[0036] S2. Action Sequence Generation: Based on the emotion level, select the corresponding action template from the action template library, and perform linear interpolation adjustment in combination with the current emotion value to generate a complete action timeline sequence.

[0037] S3, Clock Synchronization and Distribution: Enables precise clock synchronization and action sequence distribution between devices; achieves microsecond-level clock calibration through anchor time synchronization protocol, and distributes action sequences to all slave devices via wireless broadcast;

[0038] S4. Local Action Execution: Receives the synchronized action sequence, overlays it with the device's unique offset, and drives the hardware to execute the action.

[0039] Preferably, in step S1, the raw data includes the distance between devices, pairing status, historical interaction data, and random stimuli, each expressed using a distance function. Pairing duration function Interaction count function Random stimulus function Calculate the scores for each dimension;

[0040] Current sentiment value It is calculated by weighted summation of four dimensions:

[0041] ;

[0042] in, , , , These are the weighting coefficients for distance, pairing duration, number of interactions, and random stimuli, respectively.

[0043] Based on the current sentiment value The scores are ranked from low to high, defining several progressive emotional levels.

[0044] Preferably, the emotion level employs a hysteresis fallback and frontier inhibition mechanism during state transitions:

[0045] Upgrade Delay: Once the sentiment value reaches the threshold for a higher sentiment level, the upgrade will not be performed until the first preset duration has elapsed.

[0046] Degradation delay: When the sentiment value drops to a lower sentiment level threshold, a second preset duration must be maintained before the degradation is executed.

[0047] Level jump limit: Each state transition can only jump one level.

[0048] Preferably, the action sequence generation in step S2 specifically includes the following steps:

[0049] S2-1. Select Template: Select the corresponding action template based on the current emotion level;

[0050] S2-2, Linear Interpolation Adjustment: If the emotion value is between two levels, then linear interpolation is performed on the action templates of the two adjacent levels to generate action parameters that are between the two.

[0051] S2-3, Assign an absolute timestamp: Assign an absolute timestamp relative to the start of the sequence to each frame;

[0052] S2-4, Format Encapsulation: Encapsulate the complete action frame sequence into a standard data packet format;

[0053] S2-5, Wireless Broadcast: Distribute the action sequence to all slave devices via wireless broadcast.

[0054] Preferably, the clock synchronization in step S3 achieves microsecond-level clock calibration through four stages:

[0055] S3-1, Synchronization Request: The slave device sends a synchronization request to the master device, and records the local sending time. ;

[0056] S3-2, Anchor Response: After receiving the request, the master device will carry the master device's clock timestamp in the next wireless connection event. Respond;

[0057] S3-3, Deviation Calculation: Calculate the clock deviation after receiving the response from the device. :

[0058] ;

[0059] in The round-trip time for a request and response;

[0060] S3-4, Linear Compensation: Smoothly adjust the local clock through progressive linear compensation.

[0061] ;

[0062] in, , These are the local clocks before and after the adjustment, respectively. The compensation coefficient is gradually increased from 0.1 to 1.0 to avoid abnormal execution of the action sequence due to clock jumps.

[0063] Preferably, the device-specific personalized offsets in step S4 include time offset, amplitude offset, and rhythm offset;

[0064] Time offset: Each device adds a fixed time offset when performing an action. This offset is determined by a fixed seed based on the device fingerprint at the factory, ensuring that the offset is consistent for the same device each time, while the offset is different for different devices.

[0065] Amplitude deviation: Each servo motor independently superimposes amplitude deviation within ±10%, so that the movement posture of the equipment presents natural individual differences;

[0066] Rhythm offset: The execution rhythm of the action is offset by ±5%, which creates subtle differences in the action speed of different devices, further enhancing the biological realism.

[0067] Preferably, the local action execution in step S4 is also designed with a breathing-style synchronization mode:

[0068] In breathing synchronization mode, the amplitude of movement A sinusoidal breathing modulation and random noise were superimposed:

[0069] ;

[0070] in, Indicates the basic range of motion. Indicates the amplitude of respiratory modulation. Indicates respiratory rate, This represents the unique phase offset for each device. This represents minute random noise, not exceeding ±3% of the basic motion amplitude.

[0071] Preferably, the method further includes a master device negotiation step, which triggers a master device re-election within the device group when preset conditions are met. The election is based on a preset priority strategy, which includes at least the device's online duration, current sentiment value, and device identifier.

[0072] Preferably, after a new master device is elected, a master device handover procedure is performed, which includes:

[0073] The current master device or a slave device that has detected an anomaly broadcasts a master device switchover declaration;

[0074] Devices within the group verify the validity of the switching declaration;

[0075] Switch the clock synchronization source to the new master device;

[0076] Reset clock synchronization parameters;

[0077] The new master device begins clock synchronization and action sequence distribution.

[0078] The wireless broadcasting method is Bluetooth Low Energy (BLE) broadcasting; the device distance in the real-time social context data is estimated by the BLE RSSI signal strength.

[0079] An AI pet device includes a processor, a memory, a wireless communication module, and an action execution module. The memory stores a computer program, which, when executed by the processor, enables the device to implement the method of synchronizing the actions of multiple AI pet hardware devices based on an emotional state model.

[0080] The advantages of this invention are:

[0081] Compared with the prior art, the present invention has the following significant technical advantages:

[0082] 1. Emotionally driven dynamic behavior vs. fixed preset patterns

[0083] Existing technologies employ fixed, preset action patterns, executing the same actions in all situations. This invention, through an emotional state model, dynamically adjusts action parameters based on real-time social contexts such as device distance, pairing duration, and interaction frequency, enabling device behavior to possess emotional depth and contextual awareness.

[0084] 2. Synchronization accuracy of ±1 millisecond vs. 10-50 milliseconds

[0085] Existing BLE NTP-like solutions have a synchronization accuracy of 10-50 milliseconds, exceeding the human visual perception threshold (approximately 30 milliseconds). This invention improves the synchronization accuracy to within ±1 millisecond through an anchor-point time synchronization protocol and a linear compensation algorithm, ensuring that users cannot perceive the time difference between device actions.

[0086] 3. Personalized offset eliminates the mechanical feel.

[0087] In existing technologies, all devices perform identical actions, resulting in a distinctly mechanical feel. This invention, through a three-dimensional personalized offset mechanism (time, amplitude, rhythm), introduces natural individual differences while maintaining overall synchronization, making the collaborative performance of multiple devices more closely resemble a real biological group.

[0088] 4. Breathing synchronization enhances the sense of biological realism.

[0089] Existing technologies perform actions deterministically and periodically, lacking a sense of organic movement. The breathing-based synchronization mode of this invention simulates the breathing rhythm and natural vibrations of real organisms through sine wave modulation and random noise, significantly enhancing the biological realism of multi-device collaboration.

[0090] 5. Decentralized design, available offline.

[0091] Existing cloud-based coordination solutions rely entirely on network connectivity and cannot function offline. This invention employs a decentralized P2P architecture, where all computation and coordination are performed locally on the device, requiring no cloud support and enabling normal operation even in completely offline scenarios.

[0092] 6. Scalable to multi-device social network topology

[0093] Existing solutions typically only support one-to-many star topologies. The master device negotiation mechanism and sentiment state model of this invention support flexible device topologies, can be extended to multi-device social networks, and enable more complex simulations of group social behaviors. Attached Figure Description

[0094] The present invention will be further described below with reference to the accompanying drawings and embodiments:

[0095] Figure 1 A flowchart illustrating a method for synchronizing the actions of multiple AI pet devices at the same frequency;

[0096] Figure 2 A flowchart generated for the action sequence;

[0097] Figure 3 A flowchart of the four stages of clock synchronization. Detailed Implementation

[0098] Example 1

[0099] This embodiment provides a method for synchronizing the actions of multiple AI pet hardware devices based on an emotional state model, enabling these devices to simulate the "companionship" behavior of a biological group. The method involves one device in the device group acting as the master device for coordination, with the other devices acting as slave devices. The AI ​​pet hardware devices typically include a microprocessor, memory, a Bluetooth Low Energy (BLE) communication module, a BLE RSSI module for distance sensing, multiple servos, RGB LEDs, and a speaker or buzzer.

[0100] like Figure 1 As shown, the method for synchronizing the actions of multiple AI pet devices based on an emotion state model in this embodiment mainly includes the following steps:

[0101] S1. Emotional State Detection:

[0102] The perception layer collects and calculates the raw data required for the emotional state, calculates the current emotional value through the emotional state model, and determines the corresponding emotional level.

[0103] The perception layer is responsible for collecting and calculating the raw data needed to determine emotional states, including:

[0104] • Distance between devices: The physical distance between devices is estimated using the BLE RSSI (Received Signal Strength Indicator) signal strength.

[0105] • Pairing status: Records the pairing relationship and pairing duration between devices;

[0106] •Historical interaction data: Statistics on the number of interactions between devices, the time of the most recent interaction, and other historical information;

[0107] • Random stimuli: Introduce controlled random factors to simulate the unpredictability of biological behavior.

[0108] S2. Action Sequence Generation: Based on the emotion level output by the perception layer, the decision layer selects the corresponding action template from the action template library, performs linear interpolation adjustment based on the current emotion value, and generates a complete action timeline sequence. This engine is the core of emotion-driven behavior, ensuring that action performance is closely related to emotional state;

[0109] S3. Clock Synchronization and Distribution: The synchronization layer is responsible for achieving precise clock synchronization and action sequence distribution between devices. It uses the BLE anchor time synchronization protocol to achieve microsecond-level clock calibration and distributes action sequences to all slave devices via BLE broadcast. This layer is also responsible for master device election and handover management.

[0110] S4. Local Action Execution: The execution layer receives the synchronized action sequence, overlays the device's unique personalized offsets (time offset, amplitude offset, rhythm offset), and drives hardware such as motors (servos), lights (LEDs), and sounds (buzzers / speakers) to execute actions.

[0111] Example 2

[0112] This embodiment provides a method for detecting emotional state, implementing step S1 in embodiment 1.

[0113] 1. Calculation functions for each dimension.

[0114] The device distance, pairing status, historical interaction data, and random stimuli in the raw data are respectively analyzed using distance functions. Pairing duration function Interaction count function Random stimulus function Calculate the scores for each dimension;

[0115] (1) Distance function: The distance score is calculated based on the BLE RSSI signal strength.

[0116] ;

[0117] When RSSI is -40dBm (very close range) = 60; when RSSI is -70dBm (medium distance), =0.

[0118] (2) Pairing duration function: Pairing duration The longer the score, the higher the score, with a maximum of 100.

[0119] ;

[0120] A perfect score of 100 is achieved after 50 seconds of pairing.

[0121] (3) Interaction count function: historical interaction count The more you play, the higher your score, with a maximum of 100.

[0122]

[0123] A perfect score of 100 is achieved after 20 interactions.

[0124] (4) Random stimulus function: To generate random values ​​in the range of 0-20 using the Xorshift128+ pseudo-random number generator, introduce unpredictability into the system.

[0125] 2. Emotional value calculation formula.

[0126] Current sentiment value It is calculated by weighted summation of four dimensions:

[0127] ;

[0128] The weighting coefficients for each dimension are as follows:

[0129] (Distance weighting): The distance between devices is the most important emotion driver; the closer the distance, the higher the emotion value.

[0130] (Pairing Duration Weighting): The longer the pairing lasts, the higher the emotional value;

[0131] (Interaction Count Weighting): The more interactions in the past, the higher the sentiment score;

[0132] (Random stimulus weights): Introducing controlled randomness to simulate the unpredictability of biological behavior.

[0133] 3. Definition of emotional level.

[0134] This invention defines four progressively higher emotional levels based on the current emotional score E, with each level corresponding to different behavioral behaviors:

[0135] •E0 - Curiosity (Emotional Value 0-25): Slight head turning, slow breathing, low-frequency short sounds.

[0136] •E1 - Excitement (Emotional Value 26-50): Rapid head turning, faster light flashing, short mid-frequency sounds.

[0137] •E2 - Joy (Emotion Value 51-80): The whole body sways, the lights are bright and vibrant, and there are mid-to-high frequency sound effects.

[0138] •E3 - Ecstasy (Emotional Value 81-100): Large-scale full-body movements, rapid changes in all-light conditions, and high-frequency sound effects.

[0139] 4. State transition rules.

[0140] To prevent frequent jumps in emotional level near the boundary ("jittering"), this invention employs hysteresis fallback and frontier inhibition mechanisms:

[0141] • Upgrade Delay: After the sentiment value reaches a higher level threshold, an upgrade will only be performed after 1.5 seconds.

[0142] • Downgrade Delay: After the sentiment value drops to a lower threshold, a 3.0-second delay is required before the downgrade is performed.

[0143] • Level jump restriction: Each state transition can only jump one level (e.g., E0 cannot jump directly to E2).

[0144] This mechanism ensures that changes in emotion levels are smooth and stable, avoiding frequent state switching caused by sensor noise or transient interference.

[0145] Example 3

[0146] This embodiment provides a method for generating action sequences, implementing step S2 in embodiment 1.

[0147] 1. Action frame structure.

[0148] Each action frame contains data from four layers, and these layers are executed in parallel:

[0149] • Motion layer: target angles of three servos, data type uint8[3] (each angle 0-180 degrees);

[0150] • Light layer: Three-channel RGB brightness values, data type is uint16[3] (each channel 0-65535);

[0151] • Sound layer: tone frequency index and volume level, data type uint8[2];

[0152] • Timestamp: The absolute time offset relative to the start of the sequence, with a data type of uint16 (milliseconds).

[0153] Table 1 shows the field definitions for each layer of the action frame.

[0154] Table 1 Field definitions for each layer of the action frame

[0155]

[0156] The following is a simplified example of a typical motion frame sequence for the E2 Joy level:

[0157] Frame 1 (t=0ms): Motion layer [90, 90, 90], Light layer [65535, 50000, 0], Sound layer [8, 6];

[0158] Frame 2 (t=200ms): Motion layer [60, 120, 70], Light layer [65535, 60000, 10000], Sound layer [10, 7];

[0159] Frame 3 (t=400ms): Motion layer [120, 60, 110], Light layer [60000, 65535, 20000], Sound layer [12, 8];

[0160] Frame 4 (t=600ms): Motion layer [80, 100, 85], Light layer [50000, 65535, 30000], Sound layer [10, 7];

[0161] Frame 5 (t=800ms): Motion layer [100, 80, 95], Light layer [65535, 55000, 15000], Sound layer [8, 6];

[0162] Frame 6 (t=1000ms): Motion layer [90, 90, 90], Light layer [65535, 50000, 0], Sound layer [6,5];

[0163] In real-world applications, the density and complexity of frame sequences are far greater than in this example, typically involving dozens to hundreds of frames.

[0164] 2. Action sequence generation process.

[0165] like Figure 2 As shown, the action sequence generation specifically includes the following steps:

[0166] S2-1. Select Template: Select the corresponding action template based on the current emotion level;

[0167] S2-2, Linear Interpolation Adjustment: If the emotion value is between two levels (e.g., emotion value 60, between E1 and E2), then linear interpolation is performed on the action templates of the two adjacent levels to generate action parameters between the two.

[0168] S2-3, Assign an absolute timestamp: Assign an absolute timestamp relative to the start of the sequence to each frame;

[0169] S2-4, Format Encapsulation: Encapsulate the complete action frame sequence into a standard data packet format;

[0170] S2-5, BLE Broadcast: Distribute action sequences to all slave devices via BLE broadcast.

[0171] In the example of linear interpolation in S2-2: when the sentiment value is 60 (E1 upper limit 50, E2 lower limit 51), the interpolation ratio = (60-50) / (80-50) = 33.3%. At this time, the amplitude of the action is a weighted average of 66.7% of the amplitude of the E1 template and 33.3% of the amplitude of the E2 template.

[0172] Example 4

[0173] This embodiment provides a clock synchronization method to implement step S3 in embodiment 1.

[0174] like Figure 3 As shown, this embodiment designs a precision clock synchronization protocol based on BLE, which achieves microsecond-level clock calibration through four stages, including the following steps:

[0175] S3-1, Synchronization Request: The slave device sends a synchronization request to the master device, and records the local sending time. ;

[0176] S3-2, Anchor Point Response: After receiving the request, the master device will include its clock timestamp in the next BLE connection event. Respond;

[0177] S3-3, Deviation Calculation: Calculate the clock deviation after receiving the response from the device. :

[0178] ;

[0179] RTT (Round Trip Time) is the round-trip time between a request and a response.

[0180] S3-4, Linear Compensation: Instead of applying the entire deviation at once, the device smoothly adjusts the local clock through progressive linear compensation.

[0181] ;

[0182] in, , These are the local clocks before and after the adjustment, respectively. The compensation coefficient is gradually increased from 0.1 to 1.0 to avoid abnormal execution of the action sequence due to clock jumps.

[0183] The timing sequence for BLE clock synchronization is illustrated below:

[0184] (1) From the equipment Send Sync Requests constantly and record them. ;

[0185] (2) The main equipment is in Requests are received at any time, in the next BLE connection event. Always carry Respond;

[0186] (3) Calculate the response received by the device at time T3. ;

[0187] (4) Calculation ;

[0188] (5) Apply progressive compensation: , It gradually increased from 0.1 to 1.0.

[0189] The synchronization accuracy target of this protocol is within ±1 millisecond, which is far superior to the 10-50 millisecond accuracy of existing BLE NTP-like solutions. This level of accuracy meets the human visual perception threshold for motion synchronization (approximately 30 milliseconds), ensuring that users cannot perceive the time difference between actions between devices.

[0190] To ensure the continuous reliability of clock synchronization, the present invention designs the following maintenance mechanism:

[0191] • Periodic synchronization: Automatically performs clock synchronization every 5 seconds to continuously calibrate accumulated errors;

[0192] • Event-driven synchronization: Actively trigger a synchronization before starting the distribution of a new action sequence;

[0193] • Packet loss detection: Synchronization packet loss is detected through the Event_Counter mechanism. Emergency resynchronization is triggered when continuous packet loss exceeds the threshold.

[0194] • Master device switching: When the master device is unavailable, the master device election and switching process is automatically triggered.

[0195] Example 5

[0196] This embodiment provides a personalized offset mechanism for superimposing the local action execution of step S3 in embodiment 1.

[0197] To eliminate the mechanical feel when multiple devices perform the same action, this invention introduces a three-dimensional personalized offset mechanism. Each device generates a unique device fingerprint at the factory, stored in non-volatile memory, to determine its personalized offset parameters.

[0198] (1) Time offset: Each device adds a fixed time offset when performing an action, with a range of ±50 milliseconds. This offset is determined by a fixed seed based on the device fingerprint at the factory, ensuring that the offset of the same device is consistent each time, while the offset of different devices is different.

[0199] (2) Amplitude offset: The amplitude offset of each of the three servos is independently superimposed, with a range of ±10%. For example, if the target angle of a certain servo is 90 degrees and the amplitude offset is +5%, then the actual executed angle is 90 + 90 × 5% = 94.5 degrees. The offset values ​​of the three servos are independent of each other, so that the movement posture of the equipment presents natural individual differences.

[0200] (3) Rhythm Shift: The execution rhythm of an action is superimposed with a shift of ±5%. For example, if the original duration of an action is 1000 milliseconds and the rhythm shift is -3%, the actual execution time is 970 milliseconds. The rhythm shift creates subtle differences in the action speed across different devices, further enhancing the biological realism.

[0201] Example 6

[0202] At the execution level, to simulate the respiratory rhythm of real organisms, this invention designs a respiratory synchronization mode. In this mode, the amplitude of the movement... It is not constant, but rather superimposed with a sinusoidal breathing modulation and random noise:

[0203] ;

[0204] Parameter descriptions:

[0205] • The basic range of motion is determined by the emotional level and the motion template.

[0206] • The amplitude of respiratory modulation is 15% of the base amplitude.

[0207] • : Respiratory rate, corresponding to 18-48 breaths per minute, within the normal biological respiratory rate range;

[0208] • The unique phase shift of each device causes the breathing rhythms of different devices to be not completely synchronized.

[0209] • Tiny random noise to simulate the natural jitters in biological movements.

[0210] The visual effect of the breathing-style synchronization mode is that while multiple devices maintain an overall sense of synchronization, each exhibits subtle fluctuations in amplitude and phase differences, creating an organic and dynamic effect similar to a group of real organisms breathing in sync. This effect is far superior to the mechanical appearance of all devices performing the exact same movements.

[0211] Example 7

[0212] This embodiment proposes a master device negotiation and switching mechanism.

[0213] 1. Master equipment negotiation mechanism.

[0214] This invention employs a decentralized master device election mechanism to ensure the system continues to function normally even when any device is offline. The election strategy is based on the following priority:

[0215] (1) The device that enters the synchronization state first shall become the master device;

[0216] (2) If they enter at the same time, the device with the higher emotional value will be given priority;

[0217] (3) If the sentiment values ​​are the same, the device with the smallest MAC address takes priority;

[0218] (4) The main equipment monitors its own status through a heartbeat detection mechanism and initiates a switchover when an abnormality occurs.

[0219] 2. Master equipment switching process.

[0220] When it is necessary to switch master devices, the following procedure should be followed:

[0221] (1) Declaration: The current master device or the slave device that has detected an anomaly broadcasts a master device switchover declaration;

[0222] (2) Verification: Other devices in the group verify the legality of the switch statement (priority comparison);

[0223] (3) Switch clock source: All devices switch their clock synchronization source to the new master device;

[0224] (4) Reset synchronization parameters: Clear the old synchronization state and reset the alpha compensation coefficient to 0.1;

[0225] (5) New master device starts distribution: The new master device starts clock synchronization and action sequence distribution.

[0226] Example 8

[0227] This embodiment proposes an AI pet device, including a processor, a memory, a wireless communication module, and an action execution module. The memory stores a computer program. When the computer program is executed by the processor, the device can realize the method of synchronizing the actions of multiple AI pet hardware devices at the same frequency as described in any one of embodiments 1-7.

[0228] The above embodiments are only for illustrating the technical concept and features of the present invention, and are intended to enable those skilled in the art to understand the content of the present invention and implement it accordingly. They should not be construed as limiting the scope of protection of the present invention. All modifications made according to the spirit and essence of the main technical solution of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for synchronizing the actions of multiple AI pet devices based on an emotion state model, characterized in that, Includes the following steps: S1. Emotional State Detection: Collect and calculate the raw data required for emotional state, calculate the current emotional value through the emotional state model, and determine the corresponding emotional level; S2. Action Sequence Generation: Based on the emotion level, select the corresponding action template from the action template library, and perform linear interpolation adjustment in combination with the current emotion value to generate a complete action timeline sequence. S3, Clock Synchronization and Distribution: Enables precise clock synchronization and action sequence distribution between devices; achieves microsecond-level clock calibration through anchor time synchronization protocol, and distributes action sequences to all slave devices via wireless broadcast; S4. Local Action Execution: Receives the synchronized action sequence, overlays it with the device's unique offset, and drives the hardware to execute the action.

2. The method for synchronizing the actions of multiple AI pet devices at the same frequency according to claim 1, characterized in that, In step S1, the raw data includes the distance between devices, pairing status, historical interaction data, and random stimuli, each expressed using a distance function. Pairing duration function Interaction count function Random stimulus function Calculate the scores for each dimension; Current sentiment value It is calculated by weighted summation of four dimensions: ； in, , , , These are the weighting coefficients for distance, pairing duration, number of interactions, and random stimuli, respectively. Based on the current sentiment value The scores are ranked from low to high, defining several progressive emotional levels.

3. The method for synchronizing the actions of multiple AI pet devices at the same frequency according to claim 2, characterized in that, The emotional level employs a hysteresis fallback and frontier inhibition mechanism during state transitions. Upgrade Delay: Once the sentiment value reaches the threshold for a higher sentiment level, the upgrade will not be performed until the first preset duration has elapsed. Degradation delay: When the sentiment value drops to a lower sentiment level threshold, a second preset duration must be maintained before the degradation is executed. Level jump limit: Each state transition can only jump one level.

4. The method for synchronizing the actions of multiple AI pet devices at the same frequency according to claim 1, characterized in that, Step S2, which generates the action sequence, specifically includes the following steps: S2-1. Select Template: Select the corresponding action template based on the current emotion level; S2-2, Linear Interpolation Adjustment: If the emotion value is between two levels, then linear interpolation is performed on the action templates of the two adjacent levels to generate action parameters that are between the two. S2-3, Assign an absolute timestamp: Assign an absolute timestamp relative to the start of the sequence to each frame; S2-4, Format Encapsulation: Encapsulate the complete action frame sequence into a standard data packet format; S2-5, Wireless Broadcast: Distribute the action sequence to all slave devices via wireless broadcast.

5. The method for synchronizing the actions of multiple AI pet devices at the same frequency according to claim 1, characterized in that, Step S3, clock synchronization, achieves microsecond-level clock calibration through four stages: S3-1, Synchronization Request: The slave device sends a synchronization request to the master device, and records the local sending time. ; S3-2, Anchor Response: After receiving the request, the master device will carry the master device's clock timestamp in the next wireless connection event. Respond; S3-3, Deviation Calculation: Calculate the clock deviation after receiving the response from the device. : ； in The round-trip time for a request and response; S3-4, Linear Compensation: Smoothly adjust the local clock through progressive linear compensation. ； in, , These are the local clocks before and after the adjustment, respectively. The compensation coefficient is gradually increased from 0.1 to 1.0 to avoid abnormal execution of the action sequence due to clock jumps.

6. The method for synchronizing the actions of multiple AI pet devices at the same frequency according to claim 1, characterized in that, The device-specific personalized offsets in step S4 include time offset, amplitude offset, and rhythm offset; Time offset: Each device adds a fixed time offset when performing an action. This offset is determined by a fixed seed based on the device fingerprint at the factory, ensuring that the offset is consistent for the same device each time, while the offset is different for different devices. Amplitude deviation: Each servo motor independently superimposes amplitude deviation within ±10%, so that the movement posture of the equipment presents natural individual differences; Rhythm offset: The execution rhythm of the action is offset by ±5%, which creates subtle differences in the action speed of different devices, further enhancing the biological realism.

7. The method for synchronizing the actions of multiple AI pet devices at the same frequency according to claim 1, characterized in that, The local action execution in step S4 is also designed with a breathing-style synchronization mode: In breathing synchronization mode, the amplitude of movement A sinusoidal breathing modulation and random noise were superimposed: ； in, Indicates the basic range of motion. Indicates the amplitude of respiratory modulation. Indicates respiratory rate, This represents the unique phase offset for each device. This represents minute random noise, not exceeding ±3% of the basic motion amplitude.

8. The method for synchronizing the actions of multiple AI pet devices at the same frequency according to claim 1, characterized in that, The method also includes a master device negotiation step, which triggers a master device re-election within the device group when preset conditions are met. The election is based on a preset priority strategy, which includes at least the device's online duration, current sentiment value, and device identifier.

9. The method for synchronizing the actions of multiple AI pet devices at the same frequency according to claim 8, characterized in that, After a new master device is elected, a master device handover procedure is performed, which includes: The current master device or a slave device that has detected an anomaly broadcasts a master device switchover declaration; Devices within the group verify the validity of the switching declaration; Switch the clock synchronization source to the new master device; Reset clock synchronization parameters; The new master device begins clock synchronization and action sequence distribution.

10. The method for synchronizing the actions of multiple AI pet devices at the same frequency according to claim 1, characterized in that, The wireless broadcasting method is Bluetooth Low Energy (BLE) broadcasting; the device distance in the real-time social context data is estimated by the BLE RSSI signal strength.

11. An AI pet device, characterized in that, The device includes a processor, a memory, a wireless communication module, and an action execution module. The memory stores a computer program, which, when executed by the processor, enables the device to implement a method for synchronizing the actions of multiple AI pet hardware devices as described in any one of claims 1-9.

Images