Human-computer interaction method and system based on personality emotion coupling and adaptive behavior rendering

Abstract

The application discloses a human-computer interaction method and system based on character emotion coupling and adaptive behavior rendering, relates to the technical field of robots, and comprises the following steps: receiving and simultaneously inputting signals, performing priority arbitration after fusion, and converting into an interactive dose vector IDV; performing character-emotion double-layer coupling calculation on the IDV, obtaining an original increment, and performing emotion mutual exclusion and attenuation constraint to obtain a final emotion state and character parameter; acquiring a basic action template, correcting the atomic action frame of the basic action template by using a four-dimensional dynamic correction coefficient to obtain final execution parameters, and delivering the final execution parameters to an execution mechanism. The application solves the response conflict problem of multi-sensor concurrent input, realizes that the character determines the emotion sensitivity and reflects individual differences, introduces a mutual exclusion constraint mechanism to avoid unreasonable emotion state superposition, and maps and penetrates the internal character and emotion parameters into the physical execution of each action frame in real time to solve the action mechanical problem.

Description

Technical Field

[0001] This invention relates to the field of robotics, specifically to a human-computer interaction method and system based on personality and emotion coupling and adaptive behavior rendering. Background Technology

[0002] In existing technologies, embedded companion robots with limited resources have the following problems:

[0003] (1) Multi-sensor perception and scheduling conflict: When multiple tactile sensors are activated at the same time, it is easy to cause behavioral response conflict, command coverage confusion or action lag.

[0004] (2) The robot’s emotional state and personality parameters are independent of each other, and the emotion and personality model are decoupled, resulting in the same response amplitude to individuals with different “personality” settings for the same stimulus, lacking individual differences;

[0005] (3) When robots deal with multiple emotions concurrently, they lack an effective mutual exclusion constraint mechanism, which can easily lead to unreasonable superposition of states such as "extreme happiness and extreme anger".

[0006] (4) Existing robot behavior output usually uses fixed parameters to execute preset animations, which cannot map the internal emotional state and personality parameters into differentiated physical execution parameters such as movement speed, amplitude, and pauses in real time, resulting in mechanical behavior. Summary of the Invention

[0007] The purpose of this invention is to provide a human-computer interaction method and system based on personality and emotion coupling and adaptive behavior rendering, which can solve the problems of multi-sensor concurrent input conflict, lack of individual differences in emotion calculation, unreasonable emotion superposition, and mechanical execution of actions in existing companion robots.

[0008] The present invention solves the above problems through the following technical solution:

[0009] A human-computer interaction method based on personality-emotion coupling and adaptive behavior rendering includes:

[0010] Step S1: Receive concurrent input signals from multimodal sensors, including one or more of tactile signals, microphone signals, gyroscope signals, and acceleration signals; perform perceptual fusion on the concurrent input signals to obtain interaction features; perform priority arbitration on the interaction features and combine them with the corresponding subsystem cooling status to determine the winning scenario type, i.e., the interaction type; combine the interaction type with the interaction features to generate an interaction dose vector (IDV);

[0011] Step S2: Perform personality-emotion dual-layer coupling calculation on the interaction dose vector IDV to obtain the original increment, and apply emotion mutual exclusion and attenuation constraints to the original increment to obtain the final emotional state and personality parameters.

[0012] Step S3: Based on the final emotional state and personality parameters, obtain the basic action template, correct the atomic action frames of the basic action template using a four-dimensional dynamic correction coefficient, obtain the final execution parameters, and send them to the execution mechanism.

[0013] This invention employs priority arbitration for the fused sensor signals, resolving the response conflict problem of concurrent input from multiple sensors; it realizes that personality determines emotional sensitivity, reflecting individual differences through personality-emotion coupling calculation; it introduces a mutual exclusion constraint mechanism to avoid the superposition of unreasonable emotional states; and it adopts a two-layer rendering architecture of basic template and dynamic correction, which maps the internal personality and emotion parameters in real time and permeates them into the physical execution of each action frame, solving the problem of mechanical motion.

[0014] As a further improvement of the present invention, step S1 further includes fusing the dispersed tactile signals into a global tactile intensity I_total:

[0015] I_total=Active(Head)+Active(Left)+Active(Right);

[0016] Where I_total represents the sum of the logic strengths of all activated sensors at the current moment; Active() is the sensor activation state function, which returns 1 when the sensor's sampled value exceeds the threshold, otherwise it returns 0; Head represents the head touch sensor, Left represents the left touch sensor, and Right represents the right touch sensor.

[0017] As a further improvement of the present invention, the interactive dose vector IDV:

[0018] IDV = {Interaction Type, I_total, T_duration, SpatialMode};

[0019] Where T_duration is the global duration: when I_total>0, the timer continues to accumulate until all sensors are released, and the global duration T_duration is obtained. After all sensors are released, the timer is cleared; SpatialMode is the spatial mode, including single-point mode and multi-point mode. When I_total=1, it is single-point mode, and when I_total≥2, it is multi-point mode.

[0020] The interaction types include stress / voice / system commands, multi-point / combined haptic feedback, single-point special interaction, and single-point basic interaction.

[0021] As a further improvement of the present invention, the method for performing personality-emotion dual-layer coupling calculation on the interaction dose vector IDV in step S2 is as follows:

[0022] Obtain the three-dimensional personality parameter set P:

[0023] P=[T_Energy,T_Affinity,T_Courage];

[0024] Among them, T_Energy is the vitality value, which affects the baseline of response speed and movement amplitude; T_Affinity is the affinity value, which affects the interaction threshold and pause time; and T_Courage is the courage value, which affects negative emotion sensitivity and stress branch.

[0025] Obtain the current four-dimensional emotion state parameter set E:

[0026] E=[E_Joy,E_Anger,E_Fear,E_Bored];

[0027] Among them, E_Joy is the happiness value; E_Anger is the anger value; E_Fear is the fear value; and E_Bored is the boredom value.

[0028] The baseline emotion change Delta is obtained based on the interaction type differences in IDV, and a personality parameter is introduced as a dynamic gain adjustment for emotion calculation: Effective_Delta.

[0029] Effective_Delta=min(N1,Delta×DR);

[0030] DR=1.0+k_p×(T_relevant / N2-N3);

[0031] Wherein, DR is the personality influence value; T_relevant is the personality parameter most relevant to the current emotion type: T_Energy is used when E_Joy increases, T_Courage is used when E_Fear increases, and T_Affinity is used when E_Affinity changes; N1 is the upper limit of a single emotion change; N2 is the upper limit of the personality parameter value; N3 is the personality parameter center offset, where DR=1.0 when T_relevant reaches the upper limit of a single emotion change; k_p is the personality-emotion coupling strength coefficient; the values ​​of T_Energy, T_Courage, and T_Affinity are obtained by looking up a table.

[0032] Then, a strong mutual exclusion pair suppression mechanism is applied:

[0033] When E_Joy increases by ΔJoy, E_Anger -= ΔJoy * a1, E_Bored -= ΔJoy * a2;

[0034] a1 is the mutual exclusion coefficient of E_Joy↔E_Anger; a2 is the mutual exclusion coefficient of E_Joy↔E_Bored;

[0035] Then, a weak mutual exclusion check is performed on the soft upper bound:

[0036] E_Anger and E_Fear are a weakly mutually exclusive pair. If E_Anger + E_Fear exceeds the soft upper limit N4, E_Anger and E_Fear are multiplied by ratio and compressed proportionally, where ratio = N4 / (E_Anger + E_Fear).

[0037] After applying differential time decay to the emotion states without event triggers in E_Joy, E_Anger, E_Fear, and E_Bored, the updated four-dimensional emotion state parameter set E is obtained.

[0038] As a further improvement of the present invention, the calculation method of the four-dimensional dynamic correction coefficient in step S3 is as follows:

[0039] M_Speed=clamp(α1_min,(b1+w1e×T_Energy+w1j×E_Joy) / N2,α1_max);

[0040] M_Hold=clamp(α2_min,(b2+w2a×T_Affinity+w2j×E_Joy) / N2,α2_max);

[0041] M_Amp=clamp(α3_min,(b3+w3e×T_Energy+w3j×E_Joy+w3a×E_Anger-w3f×E_Fear-w3b×E_Bored) / N2,α3_max);

[0042] M_Vol=clamp(α4_min,(b4+w4e×T_Energy+w4j×E_Joy+w4a×E_Anger-w4f×E_Fear) / N2,α4_max);

[0043] Where: M_Speed ​​is the speed correction coefficient, used to adjust the action execution speed; α1_min and α1_max are the lower limit and upper limit of the speed correction coefficient, respectively; b1 is the basic bias of the speed correction coefficient; w1e is the speed correction weight coefficient of T_Energy; w1j is the speed correction weight coefficient of E_Joy;

[0044] M_Hold is the pause correction coefficient, used to adjust the pause duration between action frames; α2_min and α2_max are the lower and upper limits of the pause correction coefficient, respectively; b2 is the base bias of the pause correction coefficient; w2a is the pause correction weight coefficient of T_Affinity; w2j is the pause correction weight coefficient of E_Joy.

[0045] M_Amp is the amplitude correction coefficient, used to adjust the amplitude of the servo motor's rotation angle; α3_min and α3_max are the lower and upper limits of the amplitude correction coefficient, respectively; b3 is the base offset of the amplitude correction coefficient; w3e is the amplitude correction weight coefficient of T_Energy; w3j is the amplitude correction weight coefficient of E_Joy; w3a is the amplitude correction weight coefficient of E_Anger; w3f is the amplitude correction weight coefficient of E_Fear; w3b is the amplitude correction weight coefficient of E_Bored.

[0046] M_Vol is the volume correction coefficient, used to adjust the playback volume of accompanying sound effects; α4_min and α4_max are the lower and upper limits of the volume correction coefficient, respectively; b4 is the base bias of the volume correction coefficient; w4e is the weight coefficient of T_Energy; w4j is the volume correction weight coefficient of E_Joy; w4a is the volume correction weight coefficient of E_Anger; w4f is the volume correction weight coefficient of E_Fear.

[0047] clamp(min, value, max) is a cutoff function that returns min when value is less than min, max when value is greater than max, and value itself otherwise.

[0048] The final execution parameters include:

[0049] Final_Speed=Base_Speed×M_Speed;

[0050] Final_Angle = clamp(lower physical limit, Base_Angle × M_Amp, upper physical limit);

[0051] Final_Hold=Base_Hold×M_Hold;

[0052] Final_Volume=Base_Volume×M_Vol;

[0053] Where Final_Speed ​​is the final execution speed level; Base_Speed ​​is the preset speed level in the basic action template;

[0054] Final_Angle is the final target angle of the servo motor; Base_Angle is the preset target angle of the servo motor in the basic motion template;

[0055] Final_Hold is the final inter-frame pause duration; Base_Hold is the preset inter-frame pause duration in the basic motion template;

[0056] Final_Volume is the final sound effect playback volume; Base_Volume is the preset accompanying sound effect volume in the basic action template;

[0057] Physical limits include pitch servo limit and yaw servo limit. Final servo target angle (Final_Angle) includes final pitch servo target angle (Final_Angle_Pitch) and final yaw servo target angle (Final_Angle_Yaw).

[0058] As a further improvement to the present invention, the logic of the priority arbitration is as follows:

[0059] Based on the predefined four-level priority P1~P4 and the independent cooldown periods of the tactile subsystem, motor subsystem, speech subsystem, and idle subsystem, determine:

[0060] If a subsystem is in a cooldown period, it will be cached and will not participate in this round of arbitration.

[0061] If there is an input signal with priority P1 among the concurrent input signals, then all cooldown restrictions are waived, it is treated as a valid interaction event, and the winning scenario type, i.e., the interaction type, is determined.

[0062] For signals that are not currently blocked by cooling, scan them in descending order of priority from P2 to P3 to P4. Once a match is found, stop scanning and output only the highest priority signal as a valid interaction event and determine the winning scenario type.

[0063] A human-computer interaction system based on personality and emotion coupling and adaptive behavior rendering includes:

[0064] The sensor fusion and priority arbitration module is configured to receive concurrent input signals from multimodal sensors, including one or more of tactile signals, microphone signals, gyroscope signals, and acceleration signals; to perform perceptual fusion on the concurrent input signals to obtain interaction features; and to perform priority arbitration on the interaction features and combine them with the corresponding subsystem cooling status to determine the winning scenario type, i.e., the interaction type.

[0065] An interactive dose vector generation module is configured to combine the interaction type with the interaction feature to generate an interactive dose vector IDV.

[0066] The dynamic emotion calculation and static personality evolution module is configured to perform personality-emotion dual-layer coupling calculation on the interaction dose vector IDV to obtain the original increment, and to apply emotion mutual exclusion and attenuation constraints to the original increment to obtain the final emotional state and personality parameters.

[0067] The adaptive behavior rendering module is configured to obtain a basic action template based on the final emotional state and personality parameters, correct the atomic action frames of the basic action template using a four-dimensional dynamic correction coefficient, obtain the final execution parameters, and send them to the execution mechanism.

[0068] As a further improvement of the present invention, the sensor fusion and priority arbitration module includes:

[0069] The perception fusion unit is configured to receive concurrent input signals, fuse the scattered tactile signals in the concurrent input signals into a global tactile intensity I_total, and combine the microphone signal, gyroscope signal and accelerometer signal to generate interactive features;

[0070] A priority arbitration unit is configured to perform priority arbitration on the interaction features to determine the winning scenario type;

[0071] The cooling timer group is configured to independently maintain the cooling timers of the four interactive subsystems: haptic, motion, voice, and idle. During the cooling period, if there is an input signal, it is recorded in the buffer. If there is still an input signal after the cooling period ends, matching is triggered immediately. When a subsystem is in the cooling period, it does not affect the normal response of other subsystems.

[0072] This invention solves the response conflict problem when multiple sensors input concurrently by using global haptic intensity fusion calculation and priority pyramid arbitration mechanism, combined with independent cooling windows for subsystems, and achieves orderly scheduling and coherent response in complex interactive scenarios.

[0073] As a further improvement of the present invention, the dynamic emotion calculation and static personality evolution module includes:

[0074] The personality evolution engine is configured to maintain a three-dimensional personality parameter set P:

[0075] P=[T_Energy,T_Affinity,T_Courage];

[0076] Among them, T_Energy is the vitality value, which affects the baseline of response speed and movement amplitude; T_Affinity is the affinity value, which affects the interaction threshold and pause time; and T_Courage is the courage value, which affects negative emotion sensitivity and stress branch.

[0077] The emotion computing engine is configured to maintain a four-dimensional emotion state parameter set E:

[0078] E=[E_Joy,E_Anger,E_Fear,E_Bored];

[0079] Among them, E_Joy is the happiness value; E_Anger is the anger value; E_Fear is the fear value; and E_Bored is the boredom value.

[0080] The three-dimensional personality parameter set P is used as a gating threshold and gain coefficient to participate in the calculation of emotion parameters to obtain the original increment, and the original increment is subjected to emotion mutual exclusion and attenuation constraints to obtain the final emotion state and personality parameters.

[0081] This invention addresses the lack of individual variability in emotion calculation by using a personality-emotion dual-layer coupling model based on interactive dose vectors. It employs long-term personality parameters as the gating threshold and gain adjustment parameters for short-term emotion calculation, thereby realizing a biomimetic mechanism in which personality determines emotional sensitivity.

[0082] As a further improvement of the present invention, the adaptive behavior rendering module includes:

[0083] The behavior rendering unit is configured to obtain a basic action template based on the final emotional state and personality parameters, and to correct the atomic action frames of the basic action template using a four-dimensional dynamic correction coefficient to obtain the final execution parameters.

[0084] The scheduler is configured to physically limit the final execution parameters and then output control signals to the robot's actuator components using a single-action execution + single-slot pre-scheduling mechanism.

[0085] A robot includes a sensor assembly, an actuator assembly, a memory, and a main control processing unit (MCU). The memory stores a computer program, which, when executed by the MCU, implements the human-computer interaction method based on personality-emotion coupling and adaptive behavior rendering.

[0086] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0087] (1) The present invention adopts priority arbitration for the fused sensor signals, which solves the response conflict problem of multiple sensor concurrent inputs; through personality-emotion coupling calculation, it realizes that personality determines emotional sensitivity and reflects individual differences; introduces a mutual exclusion constraint mechanism to avoid the superposition of unreasonable emotional states; adopts a two-layer rendering architecture of basic template and dynamic correction, and maps the internal personality and emotion parameters in real time and permeates them into the physical execution of each action frame to solve the problem of mechanical action.

[0088] (2) The present invention can realize the complete processing link from the raw signal of multimodal sensor to the actuator drive signal on an embedded platform with severely limited computing power and storage resources. Attached Figure Description

[0089] Figure 1 This is a system block diagram of an embodiment of the present invention;

[0090] Figure 2 This is a flowchart of an embodiment of the present invention;

[0091] Figure 3 This is a flowchart of the global fusion process of distributed tactile sensors in an embodiment of the present invention;

[0092] Figure 4 This is a flowchart of the emotion update algorithm in an embodiment of the present invention;

[0093] Figure 5 This is a schematic diagram of the dual-channel behavior rendering pipeline in an embodiment of the present invention;

[0094] Figure 6 This is a comparison chart of the effects of correction coefficients on behavioral variations of the same basic action in embodiments of the present invention;

[0095] Figure 7 This is a timing diagram of the coverage semantic scheduling from single-point to multi-point interaction in an embodiment of the present invention. Detailed Implementation

[0096] The present invention will be further described in detail below with reference to embodiments, but the implementation of the present invention is not limited thereto.

[0097] See Figure 2 As shown, this embodiment of the invention provides a human-computer interaction method based on personality-emotion coupling and adaptive behavior rendering, including:

[0098] Step S1, Perceptual Fusion

[0099] It receives multi-channel sensor input and first performs global intensity calculation for the tactile channels, such as... Figure 3 As shown:

[0100] (1) Global strength calculation

[0101] It receives tactile signals (Head_ADC, Left_ADC, Right_ADC) detected by the head touch sensor, left touch sensor, and right touch sensor, and fuses them into a global tactile intensity I_total.

[0102] I_total=Active(Head)+Active(Left)+Active(Right)=∑Active(x);

[0103] Where I_total represents the sum of the logic strengths of all activated sensors at the current moment; Active(x) is the activation state function of sensor x. The activation state function returns 1 when the sensor's sampled value exceeds the threshold, otherwise it returns 0; Head represents the head touch sensor, Left represents the left touch sensor, and Right represents the right touch sensor.

[0104] (2) Global duration and seamless hand change

[0105] The timer continues to increment as long as any one of the head touch sensor, left touch sensor, or right touch sensor is active (I_total>0); it resets to zero when all sensors are released. It supports "seamless hand switching"—as long as the contact is uninterrupted, it is considered a single, continuous interaction.

[0106] (3) Spatial pattern classification

[0107] Based on I_total, it is classified into single-point mode SINGLE_POINT (I=1) and multi-point mode MULTI_POINT (I>=2). After combining the spatial mode and duration, the output interaction feature triple {I_total, T_duration, SpatialMode} is generated, where T_duration is the global duration and SpatialMode is the spatial mode.

[0108] Then, the motion events from the motion channel and the speech events from the speech channel are fused to obtain the result of three-channel perceptual fusion, i.e., the concurrent input signal, including:

[0109] (1) Interactive characteristics of the tactile channel output {I_total,T_duration,SpatialMode};

[0110] (2) The type of motion event output by the motion channel (fall, shake, inversion or no event);

[0111] (3) Voice event type output by the voice channel (wake word, command or no event).

[0112] Step S2, Arbitration + IDV

[0113] (1) Priority arbitration

[0114] like Figure 3 As shown, the processing result of the three-channel perception fusion is filtered by the cooling timers of each subsystem and then enters the four-level priority pyramid arbitration. The four-level priority is defined as shown in Table 1. A scene matching table is predefined, which uses I_total, T_duration, SpatialMode in the interaction feature triplet and the specific sensor activation combination as matching conditions. Each matching rule corresponds to an interaction type identifier. The arbitrator scans from high to low in order of P1>P2>P3>P4. Once a match is successful, subsequent detection is immediately interrupted, and only the highest priority signal matched is output as a valid interaction event to determine the winning scene type, i.e., the interaction type. The interaction type is combined with the interaction feature triplet to generate a standardized interaction dose vector IDV={interaction type, I_total, T_duration, SpatialMode}, which serves as the unified input interface for subsequent emotion calculation. The interaction types include stress / voice / system commands, multi-point / combined haptic, single-point special interaction, and single-point basic interaction.

[0115] Table 1. Definition of Level 4 Priority

[0116]

[0117] (2) Subsystem independent cooling timer

[0118] The cooldown timer group is located at the front end of the priority arbitration unit and functions as a signal filter. Whenever an interaction event in a subsystem successfully triggers an action execution, the cooldown timer for that subsystem starts counting. During the cooldown period, new signals generated by that subsystem do not participate in priority arbitration but are temporarily stored in a buffer. After the cooldown period ends, if there are still signals in the buffer, those signals re-enter the priority arbitration process. Each subsystem's cooldown timer is completely independent; a subsystem being in a cooldown period does not affect the normal response of other subsystems. P1 level events are exempt from all cooldown restrictions.

[0119] The four interaction subsystems—tactile, motion, voice, and idle—each maintain their own independent cooldown timers. For example, the tactile subsystem has a cooldown period of 800-1500ms and a sampling period of 1000ms; the motion subsystem has a cooldown period of 2000-10000ms and a sampling period of 50ms; the voice subsystem has a cooldown period of 1000-3000ms, with event-driven sampling; and the idle subsystem has a cooldown period controlled by a cycle and a sampling period of 1000ms.

[0120] Each subsystem's cooldown timer is completely independent and does not block others. During the cooldown period, inputs are still recorded in the buffer. If valid inputs are still available after the cooldown ends, a match is immediately triggered. P1-level scenarios are exempt from all cooldown restrictions. Non-P1 scenarios adhere to the non-interruption principle: the current action must be executed completely, and new inputs only overwrite the next action.

[0121] Step S3, State Calculation

[0122] This step addresses the lack of individual variability in emotion calculation by employing a two-layer coupling calculation of personality and emotion.

[0123] (1) Static personality parameters (long-term baseline)

[0124] The three-dimensional personality parameter group P=[T_Energy,T_Affinity,T_Courage] takes values ​​of [0,1000] and changes slowly. T_Energy (energy value, default 600) affects response speed and baseline movement amplitude; T_Affinity (affinity value, default 500) affects interaction threshold and pause time, graded into three stages: alienation period [0-300], adjustment period [301-700], and honeymoon period [701-1000]; T_Courage (courage value, default 500) affects sensitivity to negative emotions and stress response.

[0125] (2) Dynamic emotional state parameters (short-term fluctuations)

[0126] The four-dimensional emotional state parameter set E=[E_Joy,E_Anger,E_Fear,E_Bored], where E_Joy is the joy value; E_Anger is the anger value; E_Fear is the fear value; and E_Bored is the boredom value; each value is [0,1000] and fluctuates in real time.

[0127] (3) Emotional gain regulation by personality gating

[0128] After extracting the IDV, the basic emotion change value Delta is obtained by looking up a table based on the interaction type. This invention introduces personality parameters as dynamic gain adjustment for emotion calculation, rather than using personality parameters only in the behavior rendering stage. The specific formula is:

[0129] Effective_Delta=min(500,Delta×DR);

[0130] DR=1.0+k_p×(T_relevant / 1000-0.5);

[0131] Among them, Effective_Delta is the dynamic gain adjustment; T_relevant is the personality parameter most relevant to the current emotion type: T_Energy for positive interaction (increased Joy), T_Courage for negative fear (increased Fear), and T_Affinity for social intimacy (changes in Affinity). k_p is the personality-emotion coupling strength sensor, with a value range of [0.2, 1.0] and a default of 0.5.

[0132] This mechanism realizes "personality determines emotional sensitivity": for highly energetic individuals with T_Energy=800, the DR of positive interaction is 1.0+0.5×(0.8-0.5)=1.15, which means the pleasure increment is amplified by 15%; for timid individuals with T_Courage=200, the DR of fear increment is actually reduced after calculation (this needs to be achieved by setting k_p to a negative value or reversing the direction of T_Courage).

[0133] (4) Strong mutual exclusion pair inhibition mechanism, as shown in Table 2.

[0134] Table 2 Suppression of Strong Mutual Exclusion Pairs

[0135]

[0136] (5) Weak mutual exclusion pairs with soft upper bound constraints

[0137] E_Anger and E_Fear are a weakly mutually exclusive pair with a soft upper limit of 1200. When the limit is exceeded, it is compressed proportionally according to the current ratio: ratio = 1200 / (E_Anger + E_Fear), with each multiplied by ratio.

[0138] (6) Differential time decay, as shown in Table 3

[0139] Table 3 Differential Time Decay Table

[0140]

[0141] (7) The emotional update process is as follows Figure 4 As shown:

[0142] Step 1: Personality Gains and Diminishing Margins

[0143] The basic emotion change Delta is obtained by looking up a table based on the interaction type in IDV. First, a personality parameter is introduced as a dynamic gain adjustment, and Effective_Delta = min(N1, Delta × DR) is calculated, where DR = 1.0 + k_p × (T_relevant / N2 - N3). Then, marginal reduction is applied: if the existing value of the currently updated emotion dimension exceeds the reduction threshold, Effective_Delta is automatically halved.

[0144] Step 2: Suppression of strong mutual exclusion pairs

[0145] Cross-inhibition is applied to strongly mutually exclusive emotions. When E_Joy increases by Delta, E_Anger decreases by Delta×a1, and E_Bored decreases by Delta×a2; when E_Anger increases, E_Joy is inhibited according to the corresponding coefficient; when E_Bored increases, E_Joy is also inhibited according to the corresponding coefficient. Here, a1 is the mutual exclusion coefficient between E_Joy and E_Anger, and a2 is the mutual exclusion coefficient between E_Joy and E_Bored.

[0146] Step 3: Check the soft upper bound of weak mutual exclusion pairs

[0147] Check if the sum of the values ​​of the weak mutex pair E_Anger and E_Fear exceeds the soft upper limit N4. If it does, compress it proportionally according to the current ratio: ratio = N4 / (E_Anger + E_Fear), where E_Anger and E_Fear are multiplied by ratio respectively.

[0148] Step 4: Interval Cutoff

[0149] The clamp function is used to truncate E_Joy, E_Anger, E_Fear, and E_Bored to the interval [0, N2].

[0150] The above four steps must be performed strictly in sequence and cannot be out of order. After the update is complete, perform differentiated exponential decay on the emotion dimensions that were not triggered by events during this period: E(t+1) = E(t) × decay_rate, where decay_rate is the decay rate. The decay rate is different for each dimension, with negative emotions decaying faster and positive emotions decaying slower. Unnatural decay of E_Bored is only deducted through interaction events, and the updated emotion state parameters are output.

[0151] Step S4, Behavior Rendering

[0152] This section addresses the issue of mechanical and lifeless behavioral performance. The core idea is: Final_Output = Base_Template × f(Personality P) × g(Emotion E).

[0153] (1) Basic movement template

[0154] It pre-stores 20+ basic motion templates, each consisting of a sequence of atomic motion frames. Each frame contains baseline parameters: target pose [Pitch, Yaw], speed level, transition pause time, and accompanying sound effect ID, all of which are read-only values.

[0155] (2) Calculation of four-dimensional dynamic correction coefficients

[0156] M_Speed=clamp(0.5,(600+T_Energy×0.4+E_Joy×0.2) / 1000,1.3);

[0157] M_Hold=clamp(0.3,(500+T_Affinity×0.5+E_Joy×0.3) / 1000,1.5);

[0158] M_Amp=clamp(0.2,(500+T_Energy×0.8+E_Joy×0.5+E_Anger×0.5-E_Fear×0.6-E_Bored×0.4) / 1000,1.8);

[0159] M_Vol=clamp(0.15,(400+T_Energy×0.4+E_Joy×0.2+E_Anger×0.4-E_Fear×0.3) / 1000,1.3);

[0160] in:

[0161] M_Speed ​​is a speed correction factor used to adjust the speed of action execution, with a value range of [0.5, 1.3].

[0162] M_Hold is the pause correction coefficient, used to adjust the pause duration between action frames, with a value range of [0.3, 1.5].

[0163] M_Amp is the amplitude correction coefficient, used to adjust the amplitude of the servo motor rotation angle, with a value range of [0.2, 1.8].

[0164] M_Vol is the volume correction coefficient, used to adjust the playback volume of accompanying sound effects, with a value range of [0.15, 1.3].

[0165] The clamp(min, value, max) function is a truncation function. It returns min when value is less than min, max when value is greater than max, and value itself otherwise.

[0166] The unified structure of the above four formulas is as follows: the weighted sum of the baseline bias, personality parameters, and emotion parameters is added, divided by N2 for normalization, and then truncated to the upper and lower limits using clamp. The personality parameters provide a long-term stable baseline offset (determining the robot's movement style tone), while the emotion parameters provide real-time dynamic modulation (reflecting the immediate impact of the current emotional state on the movement).

[0167] Dual-channel rendering pipeline, such as Figure 5As shown. The rendering pipeline receives two inputs in parallel: a 3D personality parameter set P and a 4D emotional state parameter set E. It reads three personality parameters—T_Energy, T_Affinity, and T_Courage—from the personality baseline channel, providing long-term stable correction offsets that determine the robot's movement style tone over long timescales. For example, a high-energy individual has a higher baseline speed and amplitude, while a high-affinity individual has a longer baseline pause time. It reads four emotional parameters—E_Joy, E_Anger, E_Fear, and E_Bored—from the real-time emotion channel, providing instantaneous dynamic modulation. For example, when the current level of pleasure is high, the speed and amplitude are further increased; when the current level of fear is high, the amplitude and volume are suppressed. The outputs of the two channels are combined into a 4D dynamic correction coefficient M.

[0168] M=[M_Speed,M_Hold,M_Amp,M_Vol].

[0169] The Base_Speed, Base_Angle, Base_Hold, and Base_Volume values ​​in the basic motion template are used as read-only reference values. These are multiplied by their corresponding correction coefficients to obtain the final execution parameters, which are then truncated by physical limits before being output to the servo and speaker. Specifically, this includes:

[0170] Final_Speed=Base_Speed×M_Speed;

[0171] Final_Angle = clamp(lower physical limit, Base_Angle × M_Amp, upper physical limit);

[0172] Final_Hold=Base_Hold×M_Hold;

[0173] Final_Volume=Base_Volume×M_Vol;

[0174] in:

[0175] Final_Speed ​​is the final execution speed setting;

[0176] Final_Angle is the final target angle of the servo; the physical limits include the pitch servo limit and the yaw servo limit, which correspond to the final pitch target angle Final_Angle_Pitch and the final yaw target angle Final_Angle_Yaw in the final target angle Final_Angle, respectively. In this embodiment, the pitch servo limit is set to [-40°, +40°] and the yaw servo limit is set to [-90°, +90°], that is, Final_Angle_Pitch = clamp(-40, Base_Angle_Pitch × M_Amp, 40); Final_Angle_Yaw = clamp(-90, Base_Angle_Yaw × M_Amp, 90);

[0177] Final_Hold is the final inter-frame pause duration;

[0178] Final_Volume is the final sound effect playback volume;

[0179] Base_Speed ​​is the preset speed level in the basic action template;

[0180] Base_Angle is the preset servo target angle in the basic action template;

[0181] Base_Hold is the preset inter-frame pause duration in the basic action template;

[0182] Base_Volume is the preset volume of accompanying sound effects in the basic action template.

[0183] Differential effects of the same basic movement in different states, such as Figure 6 As shown, taking the basic action "confirmation nod" (Base_Speed=8, Base_Angle=10°, Base_Hold=0.3s, Base_Volume=0.5) as an example, it produces drastically different physical manifestations under three typical personality and emotional combinations:

[0184] Variant A – Hyperactive / Excitable (T_Energy=900, E_Joy=800):

[0185] The calculated values ​​are M_Speed=1.12, M_Amp=1.62, M_Hold=0.89, and M_Vol=0.92.

[0186] Final parameters: Final_Speed=8.96, Final_Angle=16.2°, Final_Hold=0.27s. Physically, it manifests as a rapid, large, and cheerful nod.

[0187] Variant B – Timid / Fearful Type (T_Energy=200, E_Fear=800):

[0188] The calculated values ​​are M_Speed=0.68, M_Amp=0.20, M_Hold=0.45, and M_Vol=0.35.

[0189] Final parameters: Final_Speed=5.44 seconds, Final_Angle=2.0°, Final_Hold=0.14 seconds. Physically, this manifests as an extremely slow, very slight, timid nodding.

[0190] Variant C – Cool / Calm Type (T_Affinity=200, E_Joy=100):

[0191] The calculated values ​​are M_Speed=0.70, M_Amp=0.62, M_Hold=0.45, and M_Vol=0.50.

[0192] Final parameters: Final_Speed=5.60, Final_Angle=6.2°, Final_Hold=0.14s. Physically, it manifests as a slow, slight, perfunctory nod.

[0193] The above three variations demonstrate that the same basic action template (occupying only 4 bytes of storage) can produce perceptibly different behavioral variations after being modified by a continuous 7-dimensional parameter space composed of personality parameters and emotional parameters, without the need to store action data separately for each personality and emotional state.

[0194] Step S5: Execute the scheduling

[0195] (1) Single action execution + single slot pre-scheduling method

[0196] The system does not maintain an action queue; at any given time, there can be at most one "current" action and one "next" action.

[0197] (2) Non-interruption / P1 interruption

[0198] In non-P1 scenarios, the currently executing slot must be fully executed; new input only overwrites the pending Next slot. When P1 is interrupted, the Current slot is immediately terminated and the Next slot is cleared. The overwrite semantic timing is as follows: Figure 7 As shown.

[0199] Take the process of users gradually upgrading from single-point head touching to three-point full-body interaction as an example:

[0200] T=0.0s: The user touches their head with one hand, and the tactile channel detects I_total=1 (single point). After arbitration, it is matched to TCH_H02 (head-touching joy), and the Current slot is assigned the value ACT_Enjoy_Up to start execution. At the same time, the tactile subsystem cooling timer starts (cooling time 800ms).

[0201] T=0.5s: The user adds touching the back of their left hand, and the tactile channel detects I_total=2 (multi-point), but at this time the tactile subsystem is still in the cooldown period (less than 800ms has elapsed). This signal is temporarily stored in the buffer and does not participate in this round of arbitration. The current action ACT_Enjoy_Up continues to execute unaffected.

[0202] T=0.8s: The haptic subsystem cooldown period ends, and the multi-point signals in the buffer immediately enter arbitration, matching to TCH_M02 (left snuggle), and the Next slot is assigned the value of left snuggle Snuggle_L. At this time, the Current action is still being executed, and Next simply waits without interrupting Current.

[0203] T=1.2s: The user adds the right hand (3 points), which is matched by arbitration as TCH_M04 (full body pounding), and the Next slot is overwritten as Sneeze. The original Snuggle_L is discarded—this is "overwrite semantics": new input always overwrites the old Next, only the latest intent is retained.

[0204] T=2.5s: The Current action ACT_Enjoy_Up completes, triggering the OnComplete callback. It checks if the Next slot is empty, promotes Sneeze to the new Current action, and clears the Next slot. The robot smoothly transitions from a head-patting, joyful posture to a sneezing posture without interruption or correction.

[0205] This timing sequence demonstrates two core characteristics of single-slot overlay scheduling: First, in non-P1 scenarios, the current action is fully executed and will not be interrupted by new input; second, the user's latest interaction intent is always retained in the Next slot (overlaying the old intent), ensuring that the action corresponding to the user's last operation is executed at OnComplete.

[0206] (3) Independent cooling timer

[0207] Furthermore, this invention also incorporates a multi-stage idle state mechanism, dividing the state into multiple stages based on the duration of no interaction, and combining this with an affinity personality parameter to trigger differentiated spontaneous behavior in an idle scheduling method. Each time the scheduler successfully elevates an action to Current execution, it immediately starts the cooldown timer corresponding to the interactive subsystem that generated the action. During the cooldown period, new signals generated by this subsystem do not participate in priority arbitration (they are filtered in step S2), and normal response resumes after the cooldown period ends. The cooldown timers of the four subsystems are completely independent and do not affect each other: tactile subsystem cooldown time 800~1500ms; motion subsystem cooldown time 2000~10000ms; voice subsystem cooldown time 1000~3000ms.

[0208] The idle subsystem employs periodic control. The complete working cycle of the cooling timer is as follows: cooling begins after step S5 triggers the action; during the cooling period, step S2 filters the subsystem's signals; after cooling is complete, the subsystem resumes participation in arbitration. This mechanism is similar to... Figure 1 The bidirectional connection corresponds to the cooling timer group and the priority arbitration unit.

[0209] The idle subsystem also maintains an idle cumulative duration, Timel, defined as the continuous inactivity time since the last valid interaction event, in seconds. When any interaction event is successfully triggered, Timel is immediately reset to zero and the timer restarts. The multi-stage idle state machine triggers different levels of spontaneous behavior based on the current value of Timel, and generates differentiated performance by combining affinity personality parameters. Upon receiving a user interaction signal at any stage, the idle state machine immediately exits, Timel is reset to zero, and the normal interaction response process resumes. See Table 4.

[0210] Table 4 Multi-stage Idle State Machine

[0211]

[0212] The present invention has the following advantages:

[0213] (1) Highly robust multi-sensor concurrent processing capability: Through global intensity fusion and four-level priority pyramid arbitration, the problem of instruction coverage and motion stuttering in traditional interactive robots during multi-point touch or compound motion is completely solved. The independent cooling mechanism of the subsystem reduces the average response latency of cross-modal interaction by about 60%~80%.

[0214] (2) Individualized performance with real biological characteristics: The pioneering personality-emotion dual-layer coupling model enables personality parameters to participate in each emotion calculation as a dynamic gating mechanism. The robot with the "high courage" setting has a significantly lower stress response to negative stimuli than the robot with the "low courage" setting, thus achieving "a thousand machines, a thousand faces".

[0215] (3) Emotional evolution mechanism that conforms to psychological norms: strong / weak mutual exclusion pair constraints and heterogeneous exponential decay model, which effectively avoid the unreasonable superposition of multidimensional emotions. The four-step ordered update process ensures that the emotional state always meets all consistency constraints under any input sequence.

[0216] (4) Highly lifelike motion physics rendering effect: The four-dimensional dynamic correction coefficients infuse internal emotions and personality values ​​into the physical parameters such as speed, amplitude, and pauses of the motion in real time. More than 20 basic templates generate a continuous behavior spectrum in the 7-dimensional parameter space, reducing storage overhead by an order of magnitude.

[0217] (5) High system resource efficiency: The core calculation only involves addition, multiplication, comparison and truncation. On ARM Cortex-M4, a single processing time is <2ms, ROM <64KB, RAM <4KB, and it can run completely offline.

[0218] The following is the complete processing flow of this invention applied to the continuous stroking of a desktop companion intelligent robot.

[0219] Initial state:

[0220] T_Energy=600, T_Affinity=800 (honeymoon period);

[0221] T_Courage=500,E_Joy=200,E_Anger=0,E_Fear=0,E_Bored=300.

[0222] Step 1: Perception and Arbitration

[0223] The user continuously touches their head and left side of their body with both hands for 2 seconds. The system calculates I_total=2 (multi-point mode) and T_duration=2.0s. Priority matching is P2 level "intimate immersion" scene, generating:

[0224] IDV={Interaction type:TCH_M02,I_total:2,T_duration:2.0s,Mode:MULTI_POINT}.

[0225] Step 2: Emotion Calculation for Personality Gating

[0226] The base Delta for the TCH_M02 scene is Joy + 400, Affinity + 4. Since the current interaction is a positive intimacy type, T_relevant is set to T_Affinity = 800. DR is calculated as 1.0 + 0.5 × (800 / 1000 - 0.5) = 1.15. After gain adjustment, Effective_Delta_Joy = min(500, 400 × 1.15) = 460.

[0227] Step 3: Four Steps to Emotional Update

[0228] Step 1, Diminishing Margins:

[0229] Currently, E_Joy = 200 < 700, DR = 1.0, and Effective_Delta = 460. E_Joy is updated to 200 + 460 = 660.

[0230] Step 2, Strong Mutual Exclusion Restriction:

[0231] Joy increased by 460, Anger -= 460 × 0.5 = 230 → Anger = max(0, 0 - 230) = 0; Bored -= 460 × 0.6 = 276 → Bored = max(0, 300 - 276) = 24.

[0232] Step 3, Weak mutual exclusion check:

[0233] Anger + Fear = 0 + 0 = 0 < 1200, no compression is needed.

[0234] Step 4, Interval Cutoff:

[0235] E_Joy=660, E_Anger=0, E_Fear=0, E_Bored=24. All values ​​are within the range [0, 1000].

[0236] Step 4: Behavior Rendering

[0237] Retrieve the base template "Left Snuggle" (ACT_Snuggle_L): Base_Speed=3, Base_Angle=45°(Yaw), Base_Hold=1.5s.

[0238] Calculate the correction factor:

[0239] M_Speed=clamp(0.5,(600+600×0.4+660×0.2) / 1000,1.3)=clamp(0.5,0.972,1.3)=0.972;

[0240] M_Hold=clamp(0.3,(500+800×0.5+660×0.3) / 1000,1.5)=clamp(0.3,1.098,1.5)=1.098;

[0241] M_Amp=clamp(0.2,(500+600×0.8+660×0.5-24×0.4) / 1000,1.8)=clamp(0.2,1.300,1.8)=1.300.

[0242] Step 5: Physical Output

[0243] Final_Speed=3×0.972=2.92 (slightly slower, with a sluggish feel); Final_Angle=45°×1.300=58.5° (a large incline); Final_Hold=1.5×1.098=1.65s (a longer pause, for a more enjoyable experience). Accompanied by SND_13 (a soft murmur).

[0244] Effect: The robot moved to the left at a slightly slower speed and a large distance, and remained in the position for a longer period of time after reaching the position, showing that it was immersed in the enjoyment of being touched.

[0245] During the aforementioned intimate immersion action (at T=0.5s), the robot suddenly slipped off the table (acceleration > 4g).

[0246] P1 stress trigger: Immediately terminate Current Action (ACT_Snuggle_L), switch Current to ACT_Startle_Jump, and clear Next Action.

[0247] Emotion update: Fear +1000, actually increased by 500 after marginal reduction (current Fear = 0 < 700, DR = 1.0). Courage -100.

[0248] Effect: The robot instantly switched from a cuddling posture to a startled reaction, demonstrating the highest priority interruption characteristic of P1 stress response.

[0249] See Figure 1 This invention provides a human-computer interaction system based on personality and emotion coupling and adaptive behavior rendering, comprising:

[0250] The sensor fusion and priority arbitration module is configured to receive concurrent input signals from multimodal sensors, including one or more of tactile signals, microphone signals, gyroscope signals, and acceleration signals; perform perceptual fusion on the concurrent input signals to obtain interaction features; and perform priority arbitration on the interaction features and combine them with the corresponding subsystem cooling status to determine the winning scenario type, i.e., the interaction type.

[0251] An interactive dose vector generation module is configured to combine the interaction type with the interaction feature to generate an interactive dose vector IDV.

[0252] The dynamic emotion calculation and static personality evolution module is configured to perform personality-emotion dual-layer coupling calculation on the interaction dose vector IDV to obtain the original increment, and to apply emotion mutual exclusion and attenuation constraints to the original increment to obtain the final emotional state and personality parameters.

[0253] The adaptive behavior rendering module is configured to obtain a basic action template based on the final emotional state and personality parameters, correct the atomic action frames of the basic action template using a four-dimensional dynamic correction coefficient, obtain the final execution parameters, and send them to the execution mechanism.

[0254] As a further improvement of the present invention, the sensor fusion and priority arbitration module includes:

[0255] The perception fusion unit is configured to receive concurrent input signals, fuse the scattered tactile signals in the concurrent input signals into a global tactile intensity I_total, and combine the microphone signal, gyroscope signal and accelerometer signal to generate interactive features;

[0256] A priority arbitration unit is configured to perform priority arbitration on the interaction features to determine the winning scenario type;

[0257] The cooling timer group is configured to independently maintain the cooling timers of the four interactive subsystems: haptic, motion, voice, and idle. During the cooling period, if there is an input signal, it is recorded in the buffer. If there is still an input signal after the cooling period ends, matching is triggered immediately. When a subsystem is in the cooling period, it does not affect the normal response of other subsystems.

[0258] This invention solves the response conflict problem when multiple sensors input concurrently by using global haptic intensity fusion calculation and priority pyramid arbitration mechanism, combined with independent cooling windows for subsystems, and achieves orderly scheduling and coherent response in complex interactive scenarios.

[0259] As a further improvement of the present invention, the dynamic emotion calculation and static personality evolution module includes:

[0260] The personality evolution engine is configured to maintain a three-dimensional personality parameter set P:

[0261] P=[T_Energy,T_Affinity,T_Courage];

[0262] Among them, T_Energy is the vitality value, which affects the baseline of response speed and movement amplitude; T_Affinity is the affinity value, which affects the interaction threshold and pause time; and T_Courage is the courage value, which affects negative emotion sensitivity and stress branch.

[0263] The emotion computing engine is configured to maintain a four-dimensional emotion state parameter set E:

[0264] E=[E_Joy,E_Anger,E_Fear,E_Bored];

[0265] Among them, E_Joy is the happiness value; E_Anger is the anger value; E_Fear is the fear value; and E_Bored is the boredom value.

[0266] The three-dimensional personality parameter set P is used as a gating threshold and gain coefficient to participate in the calculation of emotion parameters to obtain the original increment, and the original increment is subjected to emotion mutual exclusion and attenuation constraints to obtain the final emotion state and personality parameters.

[0267] This invention addresses the lack of individual variability in emotion calculation by using a personality-emotion dual-layer coupling model based on interactive dose vectors. It employs long-term personality parameters as the gating threshold and gain adjustment parameters for short-term emotion calculation, thereby realizing a biomimetic mechanism in which personality determines emotional sensitivity.

[0268] As a further improvement of the present invention, the adaptive behavior rendering module includes:

[0269] The behavior rendering unit is configured to obtain a basic action template based on the final emotional state and personality parameters, and to correct the atomic action frames of the basic action template using a four-dimensional dynamic correction coefficient to obtain the final execution parameters.

[0270] The scheduler is configured to physically limit the final execution parameters and then output control signals to the robot's actuator components using a single-action execution + single-slot pre-scheduling mechanism.

[0271] This invention also provides a robot, including a sensor assembly, an actuator assembly, a memory, and a main control processing unit (MCU). The memory stores a computer program, and when the computer program is executed by the MCU, it implements the human-computer interaction method based on personality and emotion coupling and adaptive behavior rendering.

[0272] Although the present invention has been described herein with reference to illustrative embodiments, the above embodiments are merely preferred embodiments of the present invention, and the implementation of the present invention is not limited to the above embodiments. It should be understood that those skilled in the art can devise many other modifications and implementations, which will fall within the scope and spirit of the principles disclosed in this application.

Claims

1. A human-computer interaction method based on personality and emotion coupling and adaptive behavior rendering, characterized in that, include: Step S1: Receive concurrent input signals from multimodal sensors, including one or more of tactile signals, microphone signals, gyroscope signals, and acceleration signals; perform perceptual fusion on the concurrent input signals to obtain interactive features; Priority arbitration is performed on the interaction features and combined with the corresponding subsystem cooling status to determine the winning scenario type, i.e., the interaction type; the interaction type and the interaction features are combined to generate an interaction dose vector (IDV); Step S2: Perform personality-emotion dual-layer coupling calculation on the interaction dose vector IDV to obtain the original increment, and apply emotion mutual exclusion and attenuation constraints to the original increment to obtain the final emotional state and personality parameters. Step S3: Based on the final emotional state and personality parameters, obtain the basic action template, correct the atomic action frames of the basic action template using a four-dimensional dynamic correction coefficient, obtain the final execution parameters, and send them to the execution mechanism. The method for performing personality-emotion dual-layer coupling calculation on the interaction dose vector IDV in step S2 is as follows: Obtain the three-dimensional personality parameter set P: P=[T_Energy,T_Affinity,T_Courage]; Among them, T_Energy is the vitality value, which affects the baseline of response speed and movement amplitude; T_Affinity is the affinity value, which affects the interaction threshold and pause time; and T_Courage is the courage value, which affects negative emotion sensitivity and stress branch. Obtain the current four-dimensional emotion state parameter set E: E=[E_Joy,E_Anger,E_Fear,E_Bored]; Among them, E_Joy is the happiness value; E_Anger is the anger value; E_Fear is the fear value; and E_Bored is the boredom value. The baseline emotion change Delta is obtained based on the interaction type differences in IDV, and a personality parameter is introduced as a dynamic gain adjustment for emotion calculation: Effective_Delta. Effective_Delta=min(N1,Delta×DR); DR=1.0+k_p×(T_relevant / N2-N3); Wherein, DR is the personality influence value; T_relevant is the personality parameter most relevant to the current emotion type: T_Energy is used when E_Joy increases, T_Courage is used when E_Fear increases, and T_Affinity is used when E_Affinity changes; N1 is the upper limit of a single emotion change; N2 is the upper limit of the personality parameter value; N3 is the personality parameter center offset, where DR=1.0 when T_relevant reaches the upper limit of a single emotion change; k_p is the personality-emotion coupling strength coefficient; the values ​​of T_Energy, T_Courage, and T_Affinity are obtained by looking up a table. Then, a strong mutual exclusion pair suppression mechanism is applied: When E_Joy increases by ΔJoy, E_Anger -= ΔJoy * a1, E_Bored -= ΔJoy * a2; a1 is the mutual exclusion coefficient of E_Joy↔E_Anger; a2 is the mutual exclusion coefficient of E_Joy↔E_Bored; Then, a weak mutual exclusion check is performed on the soft upper bound: E_Anger and E_Fear are a weakly mutually exclusive pair. If E_Anger + E_Fear exceeds the soft upper limit N4, E_Anger and E_Fear are multiplied by ratio and compressed proportionally, where ratio = N4 / (E_Anger + E_Fear). After applying differential time decay to the emotion states without event triggers in E_Joy, E_Anger, E_Fear, and E_Bored, the updated four-dimensional emotion state parameter set E is obtained.

2. The human-computer interaction method based on personality-emotion coupling and adaptive behavior rendering according to claim 1, characterized in that, Step S1 further includes fusing the dispersed tactile signals into a global tactile intensity I_total: I_total=Active(Head)+Active(Left)+Active(Right); Where I_total represents the sum of the logic strengths of all activated sensors at the current moment; Active() is the sensor activation state function, which returns 1 when the sensor's sampled value exceeds the threshold, otherwise it returns 0; Head represents the head touch sensor, Left represents the left touch sensor, and Right represents the right touch sensor.

3. The human-computer interaction method based on personality-emotion coupling and adaptive behavior rendering according to claim 2, characterized in that, The interactive dose vector IDV: IDV = {Interaction Type, I_total, T_duration, SpatialMode}; Where T_duration is the global duration: when I_total>0, the timer continues to accumulate until all sensors are released, and the global duration T_duration is obtained. After all sensors are released, the timer is cleared; SpatialMode is the spatial mode, including single-point mode and multi-point mode. When I_total=1, it is single-point mode, and when I_total≥2, it is multi-point mode. The interaction types include stress response / voice / system commands, multi-point / combined haptic feedback, single-point special interaction, and single-point basic interaction.

4. The human-computer interaction method based on personality-emotion coupling and adaptive behavior rendering according to claim 1, characterized in that, The calculation method for the four-dimensional dynamic correction coefficient in step S3 is as follows: M_Speed=clamp(α1_min,(b1+w1e×T_Energy+w1j×E_Joy) / N2,α1_max); M_Hold=clamp(α2_min,(b2+w2a×T_Affinity+w2j×E_Joy) / N2,α2_max); M_Amp=clamp(α3_min,(b3+w3e×T_Energy+w3j×E_Joy+w3a×E_Anger-w3f×E_Fear-w3b×E_Bored) / N2,α3_max); M_Vol=clamp(α4_min,(b4+w4e×T_Energy+w4j×E_Joy+w4a×E_Anger-w4f×E_Fear) / N2,α4_max); Where: M_Speed ​​is the speed correction coefficient, used to adjust the action execution speed; α1_min and α1_max are the lower limit and upper limit of the speed correction coefficient, respectively; b1 is the basic bias of the speed correction coefficient; w1e is the speed correction weight coefficient of T_Energy; w1j is the speed correction weight coefficient of E_Joy; M_Hold is the pause correction coefficient, used to adjust the pause duration between action frames; α2_min and α2_max are the lower and upper limits of the pause correction coefficient, respectively; b2 is the base bias of the pause correction coefficient; w2a is the pause correction weight coefficient of T_Affinity; w2j is the pause correction weight coefficient of E_Joy. M_Amp is the amplitude correction coefficient, used to adjust the amplitude of the servo motor's rotation angle; α3_min and α3_max are the lower and upper limits of the amplitude correction coefficient, respectively; b3 is the base offset of the amplitude correction coefficient; w3e is the amplitude correction weight coefficient of T_Energy; w3j is the amplitude correction weight coefficient of E_Joy; w3a is the amplitude correction weight coefficient of E_Anger; w3f is the amplitude correction weight coefficient of E_Fear; w3b is the amplitude correction weight coefficient of E_Bored. M_Vol is the volume correction coefficient, used to adjust the playback volume of accompanying sound effects; α4_min and α4_max are the lower and upper limits of the volume correction coefficient, respectively; b4 is the base bias of the volume correction coefficient; w4e is the weight coefficient of T_Energy; w4j is the volume correction weight coefficient of E_Joy; w4a is the volume correction weight coefficient of E_Anger; w4f is the volume correction weight coefficient of E_Fear. clamp(min, value, max) is a cutoff function that returns min when value is less than min, max when value is greater than max, and value itself otherwise. The final execution parameters include: Final_Speed=Base_Speed×M_Speed; Final_Angle = clamp(lower physical limit, Base_Angle × M_Amp, upper physical limit); Final_Hold=Base_Hold×M_Hold; Final_Volume=Base_Volume×M_Vol; Where Final_Speed ​​is the final execution speed level; Base_Speed ​​is the preset speed level in the basic action template; Final_Angle is the final target angle of the servo motor; Base_Angle is the preset target angle of the servo motor in the basic motion template; Final_Hold is the final inter-frame pause duration; Base_Hold is the preset inter-frame pause duration in the basic motion template; Final_Volume is the final sound effect playback volume; Base_Volume is the preset accompanying sound effect volume in the basic action template.

5. The human-computer interaction method based on personality-emotion coupling and adaptive behavior rendering according to claim 1, characterized in that, The logic of the priority arbitration is as follows: Based on the predefined four-level priority P1~P4 and the independent cooldown periods of the tactile subsystem, motor subsystem, speech subsystem, and idle subsystem, determine: If a subsystem is in a cooldown period, it will be cached and will not participate in this round of arbitration. If there is an input signal with priority P1 among the concurrent input signals, then all cooldown restrictions are waived, it is treated as a valid interaction event, and the winning scenario type, i.e., the interaction type, is determined. For signals that are not currently blocked by cooling, scan them in descending order of priority from P2 to P3 to P4. Once a match is found, stop scanning and output only the highest priority signal as a valid interaction event and determine the winning scenario type.

6. A human-computer interaction system based on personality and emotion coupling and adaptive behavior rendering, characterized in that, include: The sensor fusion and priority arbitration module is configured to receive concurrent input signals from multimodal sensors, including one or more of tactile signals, microphone signals, gyroscope signals, and acceleration signals, and to perform perceptual fusion on the concurrent input signals to obtain interactive features. The interaction features are prioritized and arbitrated, and the winning scenario type, i.e., the interaction type, is determined by combining the corresponding subsystem cooling status. An interactive dose vector generation module is configured to combine the interaction type with the interaction feature to generate an interactive dose vector IDV. The dynamic emotion calculation and static personality evolution module is configured to perform personality-emotion dual-layer coupling calculation on the interaction dose vector IDV to obtain the original increment, and to apply emotion mutual exclusion and attenuation constraints to the original increment to obtain the final emotional state and personality parameters. The adaptive behavior rendering module is configured to obtain a basic action template based on the final emotional state and personality parameters, correct the atomic action frames of the basic action template using a four-dimensional dynamic correction coefficient, obtain the final execution parameters, and send them to the execution mechanism. The dynamic emotion calculation and static personality evolution module includes: The personality evolution engine is configured to maintain a three-dimensional personality parameter set P: P=[T_Energy,T_Affinity,T_Courage]; Among them, T_Energy is the vitality value, which affects the baseline of response speed and movement amplitude; T_Affinity is the affinity value, which affects the interaction threshold and pause time; and T_Courage is the courage value, which affects negative emotion sensitivity and stress branch. The emotion computing engine is configured to maintain a four-dimensional emotion state parameter set E: E=[E_Joy,E_Anger,E_Fear,E_Bored]; Among them, E_Joy is the happiness value; E_Anger is the anger value; E_Fear is the fear value; and E_Bored is the boredom value. The original increment is obtained by using a three-dimensional personality parameter set P as a gating threshold and gain coefficient in the calculation of emotion parameters. Emotional mutual exclusion and attenuation constraints are then applied to the original increment to obtain the final emotional state and personality parameters, specifically including: The baseline emotion change Delta is obtained based on the interaction type differences in IDV, and a personality parameter is introduced as a dynamic gain adjustment for emotion calculation: Effective_Delta. Effective_Delta=min(N1,Delta×DR); DR=1.0+k_p×(T_relevant / N2-N3); Wherein, DR is the personality influence value; T_relevant is the personality parameter most relevant to the current emotion type: T_Energy is used when E_Joy increases, T_Courage is used when E_Fear increases, and T_Affinity is used when E_Affinity changes; N1 is the upper limit of a single emotion change; N2 is the upper limit of the personality parameter value; N3 is the personality parameter center offset, where DR=1.0 when T_relevant reaches the upper limit of a single emotion change; k_p is the personality-emotion coupling strength coefficient; the values ​​of T_Energy, T_Courage, and T_Affinity are obtained by looking up a table. Then, a strong mutual exclusion pair suppression mechanism is applied: When E_Joy increases by ΔJoy, E_Anger -= ΔJoy * a1, E_Bored -= ΔJoy * a2; a1 is the mutual exclusion coefficient of E_Joy↔E_Anger; a2 is the mutual exclusion coefficient of E_Joy↔E_Bored; Then, a weak mutual exclusion check is performed on the soft upper bound: E_Anger and E_Fear are a weakly mutually exclusive pair. If E_Anger + E_Fear exceeds the soft upper limit N4, E_Anger and E_Fear are multiplied by ratio and compressed proportionally, where ratio = N4 / (E_Anger + E_Fear). After applying differential time decay to the emotion states without event triggers in E_Joy, E_Anger, E_Fear, and E_Bored, the updated four-dimensional emotion state parameter set E is obtained.

7. The human-computer interaction system based on personality and emotion coupling and adaptive behavior rendering according to claim 6, characterized in that, The sensor fusion and priority arbitration module includes: The perception fusion unit is configured to receive concurrent input signals, fuse the scattered tactile signals in the concurrent input signals into a global tactile intensity I_total, and combine the microphone signal, gyroscope signal and accelerometer signal to generate interactive features; A priority arbitration unit is configured to perform priority arbitration on the interaction features to determine the winning scenario type; The cooling timer group is configured to independently maintain the cooling timers of the four interaction subsystems: haptic, motion, voice, and idle. If there is an input signal during the cooling period, it is recorded in the buffer. If there is still an input signal after the cooling period ends, matching is triggered immediately.

8. The human-computer interaction system based on personality and emotion coupling and adaptive behavior rendering according to claim 6, characterized in that, The adaptive behavior rendering module includes: The behavior rendering unit is configured to obtain a basic action template based on the final emotional state and personality parameters, and to correct the atomic action frames of the basic action template using a four-dimensional dynamic correction coefficient to obtain the final execution parameters. The scheduler is configured to physically limit the final execution parameters and then output control signals to the robot's actuator components using a single-action execution + single-slot pre-scheduling mechanism.

Images