Method and device for adjusting dynamic rules of digital human based on causality

Abstract

The present application provides a method and device for dynamic rule adjustment of a digital human based on causality. The method comprises: obtaining historical data related to the interaction between the digital human and the user in a mixed reality scenario, wherein the historical data includes historical behavioral data and historical environmental data; determining the control step length and perception sliding window length of the digital human's behavioral decision based on the historical data, and constructing a behavior decision data matrix based on the control step length and perception sliding window length; performing causal structure extraction processing on the behavior decision data matrix to obtain a block matrix under causal constraints; generating a behavior adjustment rule set that satisfies the causal relationship based on the block matrix, and constructing a causality-driven multi-step behavior predictor based on the behavior adjustment rule set; and dynamically adjusting the behavior rules of the digital human based on the multi-step behavior predictor. The present application solves the technical problem that the digital human has poor adaptive behavior adjustment capabilities in complex interactive scenarios due to the lack of causal modeling.

Description

Technical Field

[0001] The present invention relates to the field of digital humans, and in particular to a method and device for adjusting dynamic rules of digital humans based on causality. Background Art

[0002] With the widespread adoption of mixed reality technology, digital humans, as the core medium for user interaction with virtual environments, have become crucial for enhancing the immersive experience through their adaptive behavioral capabilities. Current mainstream digital human systems often utilize predefined behavioral rule libraries or dynamic decision-making models based on statistical learning (such as deep reinforcement learning). However, these approaches have significant limitations in complex and ever-changing interactive scenarios. On the one hand, static rule libraries struggle to capture the diverse user behaviors in dynamic environments. When encountering unpredictable interaction scenarios, digital humans are prone to logical rigidity or inaccurate responses. On the other hand, while data-driven models can learn from historical interaction patterns, their reliance on statistical correlation rather than causality makes them susceptible to interference from environmental noise and confounding factors, leading to pseudo-correlation biases in behavioral decisions.

[0003] To address the above-mentioned problems, no effective solutions have been proposed so far. Summary of the Invention

[0004] The embodiments of the present invention provide a method and device for adjusting dynamic rules of digital humans based on causality, so as to at least solve the technical problem that the digital humans have poor adaptive behavior adjustment capabilities in complex interactive scenarios due to the lack of causal modeling.

[0005] According to one aspect of an embodiment of the present invention, a method for dynamic rule adjustment of a digital human based on causality is provided, comprising: obtaining historical data related to the interaction between a digital human and a user in a mixed reality scene, wherein the historical data includes historical behavior data and historical environment data; determining a control step size and a perception sliding window length of the digital human's behavior decision based on the historical data, and constructing a behavior decision data matrix based on the control step size and the perception sliding window length; performing causal structure extraction processing on the behavior decision data matrix to obtain a block matrix under causal constraints; generating a behavior adjustment rule set that satisfies the causal relationship based on the block matrix, and constructing a causality-driven multi-step behavior predictor based on the behavior adjustment rule set; and dynamically adjusting the behavior rules of the digital human based on the multi-step behavior predictor.

[0006] According to another aspect of an embodiment of the present invention, a device for dynamic rule adjustment of a digital human based on causal relationships is provided, comprising: an acquisition module configured to acquire historical data related to the interaction between a digital human and a user in a mixed reality scene, wherein the historical data includes historical behavior data and historical environment data; a construction module configured to determine the control step size and the perception sliding window length of the digital human's behavior decision based on the historical data, and to construct a behavior decision data matrix based on the control step size and the perception sliding window length; an extraction module configured to perform causal structure extraction processing on the behavior decision data matrix to obtain a block matrix under causal constraints; a generation module configured to generate a behavior adjustment rule set that satisfies the causal relationship based on the block matrix, and to construct a causal-driven multi-step behavior predictor based on the behavior adjustment rule set; and an adjustment module configured to dynamically adjust the behavior rules of the digital human based on the multi-step behavior predictor.

[0007] In an embodiment of the present invention, historical data related to interactions between a digital human and a user in a mixed reality scenario is obtained, where the historical data includes historical behavioral data and historical environmental data. Based on the historical data, the control step size and perception window length of the digital human's behavioral decision are determined, and a behavior decision data matrix is ​​constructed based on the control step size and perception window length. The behavior decision data matrix is ​​subjected to causal structure extraction processing to obtain a block matrix under causal constraints. Based on the block matrix, a set of behavior adjustment rules that satisfy causal relationships is generated, and based on the behavior adjustment rule set, a causality-driven multi-step behavior predictor is constructed. Based on the multi-step behavior predictor, the behavior rules of the digital human are dynamically adjusted. This solution solves the technical problem of the digital human's poor adaptive behavior adjustment ability in complex interactive scenarios due to the lack of causal modeling. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0009] Figure 1 is a flow chart of a method for adjusting dynamic rules of a digital human based on causality according to an embodiment of the present invention;

[0010] Figure 2 is a flow chart of another method for adjusting dynamic rules of digital humans based on causality according to an embodiment of the present invention;

[0011] Figure 3 is a flow chart of a method for constructing a behavior decision data matrix according to an embodiment of the present invention;

[0012] Figure 4is a flow chart of a method for generating a behavior adjustment rule set that satisfies a causal relationship according to an embodiment of the present invention;

[0013] Figure 5 2 is a schematic structural diagram of a device for adjusting dynamic rules of a digital human based on causality according to an embodiment of the present invention;

[0014] Figure 6 A schematic structural diagram of an electronic device suitable for implementing the embodiments of the present disclosure is shown. DETAILED DESCRIPTION

[0015] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0016] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0017] According to an embodiment of the present invention, a method embodiment of a method for dynamic rule adjustment of a digital human based on causality is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0018] Figure 1 The method for adjusting dynamic rules of digital humans based on causality according to an embodiment of the present invention is as follows. Figure 1 As shown, the method includes the following steps:

[0019] Step S102: Acquire historical data related to the interaction between the digital human and the user in the mixed reality scene, wherein the historical data includes historical behavior data and historical environment data.

[0020] Step S104: determining the control step length and the perception sliding window length of the digital human's behavior decision based on the historical data, and constructing a behavior decision data matrix based on the control step length and the perception sliding window length.

[0021] First, determine the control step size and perception sliding window length. For example, analyze the temporal relationship between the rate of change of interactive events and behavioral responses in the historical data to extract dynamic time features that reflect the digital human's behavioral response trends. Based on these dynamic time features, evaluate the digital human's behavioral stability index and the environmental disturbance amplitude. Based on these behavioral stability index and the environmental disturbance amplitude, determine the control step size and perception sliding window length.

[0022] Next, a behavior decision data matrix is ​​constructed. For example, based on the control step size and the perception window length, the historical data is segmented into sliding windows, and the input trajectory, state trajectory, and interference features within each sliding window are extracted to construct multiple joint data blocks that integrate behavior, environment, and response. These multiple joint data blocks are arranged in chronological order to form the behavior decision data matrix with a multi-time period causal structure.

[0023] Step S106 , performing causal structure extraction processing on the behavior decision data matrix to obtain a block matrix under causal constraints.

[0024] For example, the behavior decision data matrix is ​​subjected to lower triangular orthogonal decomposition to obtain a lower triangular causal transfer matrix representing the behavior response causal path; the lower triangular causal transfer matrix is ​​subjected to structural decoupling of the perturbation feedback path, the lower triangular blocks in the lower triangular causal transfer matrix are retained and the upper triangular blocks are shielded to obtain the block matrix under causality constraints.

[0025] Step S108 : generating a behavior adjustment rule set that satisfies the causal relationship based on the block matrix, and constructing a causal-driven multi-step behavior predictor based on the behavior adjustment rule set.

[0026] First, a behavior adjustment rule set is generated. For example, the block matrix is ​​processed by causal rule induction to extract the disturbance feedback paths and causal influence relationships of the digital human under different sensory input and historical behavior conditions. Based on the disturbance feedback paths and causal influence relationships, the behavior adjustment rule set containing behavior triggering conditions and corresponding behavior response strategies is constructed.

[0027] Next, a multi-step behavior predictor is constructed. For example, based on the behavior adjustment rule set, a lower triangle prediction structure based on a disturbance feedback path is adopted to construct the multi-step behavior predictor.

[0028] Step S110: dynamically adjusting the behavior rules of the digital human based on the multi-step behavior predictor.

[0029] The multi-step behavior predictor outputs behavior predictions for multiple time steps in the future based on the sensory input within the current sliding window, historical behavior states, and environmental disturbance characteristics. Based on the behavior probability distribution, response delay interval, and disturbance sensitivity score contained in the prediction results, the digital human's behavioral rules are dynamically adjusted online.

[0030] Figure 2 is another method for adjusting dynamic rules of digital humans based on causality according to an embodiment of the present invention. Figure 2 As shown, the method includes the following steps:

[0031] Step S202: Acquire historical data related to the interaction between the digital human and the user in the mixed reality scene.

[0032] In this embodiment, a multimodal sensor system (including visual sensors, depth cameras, voice collectors, and environmental sensors) deployed in a mixed reality (MR) scene continuously collects user behavior (such as gaze trajectory, voice commands, gesture trajectory, etc.) and environmental information (such as scene lighting, object distribution, and spatial sound effects). This sensor data is fused via a synchronous fusion module to generate historical behavior and environmental data in a standardized format. The collection cycle can be set to 10 to 30 frames per second, dynamically adjusted based on the processing power of the MR device.

[0033] To ensure data accuracy and integrity, sensors are calibrated before data collection, and multi-source data is aligned using timestamps. The visual sensor uses YOLOv8+DeepSORT for person and object recognition and tracking. Depth information is used to construct a spatial reconstruction model using a point cloud library (such as PCL). Speech information is transcribed into text instructions using a pre-trained speech recognition model (such as Wav2Vec2.0). Finally, a time-series fusion network (such as the Transformer Encoder) is used to extract a unified cross-modal action perception vector sequence.

[0034] Step S204: determining the control step length and the perception sliding window length of the digital human's behavior decision based on historical data, and constructing a behavior decision data matrix based on the control step length and the perception sliding window length.

[0035] like Figure 3 As shown, the method for constructing the behavior decision data matrix includes the following steps:

[0036] Step S2042: extract dynamic time features.

[0037] In this embodiment, a time series feature vector group of interactive events is first constructed based on historical data, including the gaze point movement speed sequence v_gaze(t), the gesture direction change sequence v_gesture(t), the speech phoneme start and end sequence v_voice(t), etc.; at the same time, the response behavior delay sequence r_delay(t), the action start time r_start(t), the reaction peak time r_peak(t), etc. of the digital human in the corresponding time period are extracted.

[0038] The dynamic time warping (DTW) method is used to align the above interaction-response pairs, construct a similarity matrix, and calculate their timing offset index Δτ, which represents the typical delay structure of the digital human's reaction triggered by the interaction change.

[0039] Subsequently, the statistical characteristics (mean, standard deviation, local kurtosis, etc.) of these timing offset indicators are calculated for multiple rounds of interaction samples to form the dynamic time feature T_dyn, which reflects the trend stability and variability of the digital human's response timing.

[0040] Step S2044: Determine the control step size and the sensing sliding window length.

[0041] In this embodiment, a behavioral stability indicator S_behave is constructed, which includes the following components: response delay mean μ_delay, standard deviation σ_delay, response amplitude consistency ρ_act, and response pattern confidence score_pattern; at the same time, the environmental disturbance amplitude E_var is evaluated, and the light brightness fluctuation value ΔI(t), the ambient sound background change rate ΔS(t), and the displacement speed of objects in the scene v_obj(t) are collected.

[0042] T_dyn, S_behave and E_var form a feature set Z_feat and input it into a lightweight causal structure perception network (such as Causal-GRNN). The network has been trained with a labeled interaction dataset (such as Multimodal-MR-Sim) and can predict the adaptive control step size h (behavior update granularity) and perception sliding window length l (perception fusion time window) in the current scene.

[0043] In order to enhance the generalization ability of estimation, adversarial sample training based on perturbation distribution simulation (such as adding high-frequency lighting changes or sudden sound interference) is introduced in the model training stage to improve the model's adaptability to extreme scenarios.

[0044] Step S2046: Constructing a behavior decision data matrix.

[0045] Set the perception sliding window length to l and the control step size to h. Use the sliding window method to gradually divide the historical data sequence into several time periods. Each time period is represented as [t, t+l], with an interval of h steps.

[0046] For each sliding window segment, we extract: 1) input trajectory u[t:t+l]: including user instructions, environmental changes, etc.; 2) state trajectory x[t:t+l]: including the state of the digital human, such as position, posture, expression parameters, etc.; 3) interference features d[t:t+l): such as noise in the scene, lighting changes, and abnormal user behavior.

[0047] The three types of data are normalized and time-aligned to form a joint data block F_i that integrates the behavior, environment, and response. The set of joint data blocks corresponding to all sliding window segments is {F_1, F_2, ..., F_N}, where N depends on the number of sliding windows.

[0048] In order to improve structural consistency and noise resistance, in some embodiments, the following processing can be further added when constructing each data block: using an occlusion-aware mechanism to perform multimodal interpolation on missing / outliers; introducing temporal positional embedding of timestamps within a sliding window to maintain temporal continuity; using a cross-channel attention module to perform weighted fusion of u, x, and d to generate a context-aware multi-channel joint vector as multiple joint data blocks.

[0049] Subsequently, the multiple joint data blocks are arranged in chronological order to form the behavior decision data matrix with a multi-time period causal structure. Specifically, the joint data blocks {F_1, ..., F_N} generated by the N sliding window segments are stacked in their original chronological order to form a behavior decision data matrix D of dimension N × M, where M is the multidimensional fusion feature dimension of each data block.

[0050] To enhance the expressiveness of the causal structure of the matrix, this embodiment maintains strict temporal order during the construction process and uses lower-triangular masking to explicitly encode the history-to-current causal direction.

[0051] Step S206: Perform causal structure extraction processing on the behavior decision data matrix D to obtain a block matrix under causal constraints.

[0052] This example introduces a causal structure orthogonal decomposition algorithm. First, a lower triangular orthogonal decomposition (LT-SVD) is performed on the behavioral decision data matrix D to obtain the lower triangular causal transfer matrix L_causal and the residual term E_res. A perturbation feedback path identifier (consisting of a multi-head self-attention mechanism) is used to structurally decouple L_causal, extracting the lower triangular block L_lower and masking the upper triangular block to construct the block matrix B_cause under causality constraints.

[0053] In order to improve the robustness of modeling, a perturbation path compression regularization term can be added during the decomposition process. The objective function is as follows:

[0054] Min ||D - L_lower||² + λ1 * ||L_lower|| + λ2 * ||E_res||²,

[0055] Among them, the first term is the causal fitting error, the second term is the perturbation path sparsity regularization term, and the third term is the residual control term. In the formula, λ1 is the sparsity regularization coefficient, which is used to balance the weights of the main fitting term and the sparsity term. The larger λ1 is, the fewer non-zero causal paths there will be in L_lower. λ2 is the residual penalty term coefficient, which is used to control the penalty intensity of the fitting error. The larger λ2 is, the stricter the model's reconstruction requirements for D. The objective function in this embodiment not only significantly weakens the spurious correlation path, but also has scalability to adapt to different task scales.

[0056] In other embodiments, in order to deal with the numerical stability problem in the high-dimensional matrix decomposition process, a block matrix recursive decoupling strategy can also be adopted, which only processes a part of the causal path in the time window each time, performs segmented compression, and then reorganizes the overall causal graph through weighted averaging.

[0057] Step S208: Generate a set of behavior adjustment rules that satisfy the causal relationship based on the block matrix.

[0058] like Figure 4 As shown, the method for generating a behavior adjustment rule set that satisfies causal relationships includes the following steps:

[0059] Step S2082: extract the disturbance feedback path and the causal influence relationship.

[0060] In this embodiment, a perturbation feedback path is a causal path in which the perceptual input variable does not directly influence the behavioral response, but rather indirectly regulates it through the intermediary of the perturbation feature variable. First, a path tracing graph is constructed for each pair of variables in the block matrix B_cause. A structural decomposition algorithm is used to identify intermediate links in multi-hop paths involving perturbation feature variables. If the joint weight of the path "perceptual input → perturbation feature → behavioral variable" exceeds a threshold, the path is marked as a perturbation feedback path.

[0061] Next, we calculate the causal weight matrix W_c between the decision-making variables and the input variables in the block matrix B_cause. This matrix represents the strength of the causal influence between each pair of variables and is numerically stable through weighted summation and normalization of the causal graph structure.

[0062] To identify potential causal behavior types, unsupervised clustering methods are used to cluster the column vectors or subspaces of W_c. Clustering algorithms can include spectral clustering, the EM algorithm based on Gaussian mixture models, or sparse subspace clustering based on mutual information distance metrics. Each cluster represents a behavioral response pattern with a similar causal structure. The final output is a collection of causal paths (i.e., causal influence relationships), each of which corresponds to a set of candidate behavioral rules.

[0063] Step S2084: constructing the behavior adjustment rule set including behavior triggering conditions and corresponding behavior response strategies based on the disturbance feedback path and the causal influence relationship.

[0064] Based on the extracted perturbation feedback paths and causal influence relationships, the dynamic relationships between the input variables, interfering mediating variables, and behavioral outputs involved in each causal path are analyzed one by one, and converted into rule entries in an interpretable structural form. For each perturbation feedback path, the joint conditions of its causal chain are extracted, such as "voice intonation change" combined with "sudden increase in background noise" can constitute a causal condition for predicting "voice response delay." If the path appears frequently in the training samples and has significant causal weight, it is refined into a formal rule.

[0065] When constructing rules, multi-dimensional constraints are used to represent the triggering boundaries and behavioral strategy outputs of each rule, including: 1) perception input conditions: generated by fuzzy logic partitioning of perception vectors; 2) historical behavior status: extracting subcategories from the historical behavior label space; 3) interference feedback status: determined by the activity distribution of disturbance path nodes; 4) behavioral response strategy: including behavior category, action parameters, response probability range, etc.; 5) execution condition additional items: such as confidence threshold, priority level, time delay tolerance, etc.

[0066] All generated rules are organized into a hierarchical rule set, divided into a regular rule layer (handling basic stable paths), a disturbance adjustment layer (handling rules with intermediate feedback), and an anomaly avoidance layer (handling extreme / sudden inputs). Furthermore, a disturbance feedback scoring function can be introduced to weight each rule based on environmental fluctuations during operation, enabling online dynamic adjustment capabilities.

[0067] The resulting behavior adjustment rule set can not only characterize the response logic in complex MR scenarios, but also directly support subsequent predictive scheduling and multi-strategy behavior control module calls.

[0068] Step S210: construct a causal-driven multi-step behavior predictor based on the behavior adjustment rule set.

[0069] This embodiment uses a lower triangular prediction structure to construct a recursive multi-step prediction network. The network input is the joint behavior-perception features within the current sliding window, and the output is the probability distribution of behavior for the next T steps. Each prediction step uses a gated recurrent unit (such as a GRU) to encode the previous step state and automatically selects a prediction path based on the causal triggering rules in Rule_base.

[0070] In order to enhance the model's multimodal alignment capabilities, a Cross-ModalAttention module is introduced into the network to assign dynamic weights to the visual, speech, and environmental channels. The multi-task loss function is used in the training phase:

[0071] L_total = αL_pred + βL_rule + γL_cons

[0072] Among them, L_pred is the behavior prediction error, L_rule is the rule deviation penalty, L_cons is the perturbation path consistency loss, and the hyperparameters α, β, and γ are automatically adjusted through Bayesian optimization.

[0073] During the training process, the Monte Carlo perturbation resampling mechanism is used to improve the robustness to uncertain scenarios, while multi-source data enhancement (such as background changes and noise superposition) is introduced to improve model generalization.

[0074] Step S212: Dynamically adjust the behavior rules of the digital human based on the multi-step behavior predictor.

[0075] During operation, the MR device collects current behavior-perception features every second and feeds them into a predictor to generate a candidate behavior sequence for the next T steps. Each candidate sequence is evaluated for benefits (e.g., task completion rate, user response satisfaction), and the behavior with the highest expected benefit is selected. The digital human's current behavior rule table is dynamically updated and pushed to the behavior execution module.

[0076] Dynamic updates of behavioral rules are implemented through a rule cache. The system categorizes rules into three types based on their frequency of change: high-frequency, low-frequency, and immediate rules. Different refresh rates and expiration policies are configured for each category. Instant rules support rapid insertion and replacement within 15ms to ensure responsiveness in high-interaction scenarios.

[0077] Furthermore, when environmental disturbances exceed the stability threshold, a behavioral rule relearning mechanism is triggered: the system sends current behavior-environment data back to the behavioral decision module to retrain the lightweight prediction model, enabling online behavioral rule fine-tuning. This mechanism relies on the edge-deployed TinyML model to quickly perform micro-parameter adjustments, avoiding frequent migration and updates of the overall model and improving the long-term operational stability of the system.

[0078] Through the above embodiments, the present invention not only improves the causal modeling ability and behavior prediction accuracy of digital humans in complex MR environments, but also significantly enhances the adaptability of digital humans to sudden disturbances and complex user behaviors, and has higher interaction naturalness and system robustness.

[0079] This application also provides a digital human dynamic rule adjustment device based on causality, such as Figure 5 As shown, it includes: an acquisition module 52, configured to acquire historical data related to the interaction between the digital human and the user in the mixed reality scene, wherein the historical data includes historical behavior data and historical environment data; a construction module 54, configured to determine the control step size and the perception sliding window length of the digital human's behavior decision based on the historical data, and construct a behavior decision data matrix based on the control step size and the perception sliding window length; an extraction module 56, configured to perform causal structure extraction processing on the behavior decision data matrix to obtain a block matrix under causal constraints; a generation module 58, configured to generate a behavior adjustment rule set that satisfies the causal relationship based on the block matrix, and construct a causal-driven multi-step behavior predictor based on the behavior adjustment rule set; an adjustment module 59, configured to dynamically adjust the behavior rules of the digital human based on the multi-step behavior predictor.

[0080] It should be noted that the causal relationship-based digital human dynamic rule adjustment device provided in the above embodiment is only illustrated by the division of the above-mentioned functional modules. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the causal relationship-based digital human dynamic rule adjustment device provided in the above embodiment and the causal relationship-based digital human dynamic rule adjustment method embodiment are based on the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0081] Figure 6 Schematic diagram of the structure of an electronic device suitable for implementing the embodiment of the present disclosure is shown. Figure 6 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0082] like Figure 6As shown, the electronic device includes a central processing unit (CPU) 1001, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 1002 or the program loaded from the storage part 1008 into the random access memory (RAM) 1003. Various programs and data required for system operation are also stored in the RAM 1003. The CPU 1001, ROM 1002 and RAM 1003 are connected to each other via a bus 1004. An input / output (I / O) interface 1005 is also connected to the bus 1004.

[0083] The following components are connected to the I / O interface 1005: an input section 1006 including a keyboard, a mouse, and the like; an output section 1007 including devices such as a cathode ray tube (CRT), a liquid crystal display (LCD), and a speaker; a storage section 1008 including a hard disk; and a communication section 1009 including a network interface card such as a LAN card or a modem. The communication section 1009 performs communication processing via a network such as the Internet. A drive 1010 is also connected to the I / O interface 1005 as needed. A removable medium 1011, such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory, is installed in the drive 1010 as needed, so that computer programs read therefrom can be installed into the storage section 1008 as needed.

[0084] The above is only a preferred embodiment of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for adjusting dynamic rules of digital humans based on causality, characterized in that: include: Acquiring historical data related to the interaction between the digital human and the user in the mixed reality scene, wherein the historical data includes historical behavior data and historical environment data; Determining a control step length and a perception sliding window length of the digital human's behavior decision based on the historical data, and constructing a behavior decision data matrix based on the control step length and the perception sliding window length; Performing causal structure extraction processing on the behavior decision data matrix to obtain a block matrix under causal constraints; generating a behavior adjustment rule set satisfying causal relationships based on the block matrix, and constructing a causal-driven multi-step behavior predictor based on the behavior adjustment rule set; Based on the multi-step behavior predictor, the behavior rules of the digital human are dynamically adjusted.

2. The method according to claim 1, characterized in that Determining the control step length and the perception sliding window length of the digital human's behavioral decision based on the historical data includes: Analyzing the temporal relationship between the change rate of the interactive events and the behavioral responses in the historical data, and extracting dynamic time features reflecting the behavioral response trends of the digital human; Based on the dynamic time characteristics, the behavior stability index of the digital human and the environmental disturbance amplitude are evaluated, and based on the behavior stability index and the environmental disturbance amplitude, the control step size and the perception sliding window length are determined.

3. The method according to claim 1, characterized in that Constructing a behavior decision data matrix based on the control step size and the perception sliding window length includes: Based on the control step size and the perception sliding window length, the historical data is segmented into sliding windows, and the input trajectory, state trajectory and interference features within each sliding window are extracted respectively to construct multiple joint data blocks that integrate behavior, environment and response; The multiple joint data blocks are arranged in chronological order to form the behavior decision data matrix with a multi-time period causal structure.

4. The method according to claim 3, characterized in that Perform causal structure extraction processing on the behavior decision data matrix to obtain a block matrix under causal constraints, including: Performing a lower triangular orthogonal decomposition on the behavior decision data matrix to obtain a lower triangular causal transfer matrix representing the behavior response causal path; The lower triangular causal transfer matrix is ​​subjected to structural decoupling of a disturbance feedback path, the lower triangular blocks in the lower triangular causal transfer matrix are retained and the upper triangular blocks are shielded, so as to obtain the block matrix under causality constraint.

5. The method according to claim 4, characterized in that Generating a behavior adjustment rule set that satisfies the causal relationship based on the block matrix includes: Performing causal rule induction processing on the block matrix to extract the disturbance feedback path and causal influence relationship of the digital human under different perceptual input and historical behavior conditions; Based on the disturbance feedback path and the causal influence relationship, the behavior adjustment rule set including behavior triggering conditions and corresponding behavior response strategies is constructed.

6. The method according to claim 1, characterized in that Based on the behavior adjustment rule set, a causal-driven multi-step behavior predictor is constructed, including: based on the behavior adjustment rule set, adopting a lower triangle prediction structure based on a disturbance feedback path to construct the multi-step behavior predictor.

7. A digital human dynamic rule adjustment device based on causality, characterized in that: include: An acquisition module is configured to acquire historical data related to the interaction between the digital human and the user in the mixed reality scene, wherein the historical data includes historical behavior data and historical environment data; A construction module is configured to determine a control step length and a perception sliding window length of the digital human's behavior decision based on the historical data, and to construct a behavior decision data matrix based on the control step length and the perception sliding window length; an extraction module configured to perform causal structure extraction processing on the behavior decision data matrix to obtain a block matrix under causal constraints; a generation module configured to generate a behavior adjustment rule set satisfying causal relationships based on the block matrix, and to construct a causal-driven multi-step behavior predictor based on the behavior adjustment rule set; An adjustment module is configured to dynamically adjust the behavior rules of the digital human based on the multi-step behavior predictor.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the method according to any one of claims 1 to 6.

9. A computer device, characterized in that: include: A memory and a processor, wherein the memory stores a computer program; the processor is configured to execute the computer program stored in the memory, wherein when the computer program is executed, the processor executes the method according to any one of claims 1 to 6.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images