Multi-dimensional interaction quality perception and resource optimization scheduling method for holographic communication system

Abstract

The invention discloses a holographic communication system-oriented multi-dimensional interaction quality perception and resource optimization scheduling method, which comprises the following steps of: firstly, constructing a system containing objective performance and subjective experience indexes according to a scene, acquiring network, holographic and user layer data, training a subjective experience representation model, and estimating the subjective experience indexes in real time; estimating indexes at the current moment and the next moment, constructing a design vector, and fusing physical, semantic and cognitive layer multi-source knowledge to obtain an enhanced input vector; reasoning the dynamic weight of the index through a utility function and a fuzzy rule, and calculating an interaction quality comprehensive score; screening indexes to be optimized, converting the indexes to be optimized into a multi-objective optimization problem by taking a comprehensive score reaching the standard as a constraint, and searching a compromise optimal scheme; and finally, converting the scheme into a resource configuration scheme to realize self-adaptive scheduling of network and terminal resources. Accurate mapping of objective indexes and subjective experience is realized, scenes and user requirements are dynamically adapted, and immersion experience and communication performance of holographic communication are improved.

Description

Technical Field

[0001] This invention relates to, specifically to, a method for multidimensional interactive quality perception and resource optimization scheduling for holographic communication systems. Background Technology

[0002] Holographic communication is an emerging communication method for remote interactive experiences and is considered one of the important application scenarios for 6G wireless communication. Through holographic communication technology, users can transmit and present high-resolution 3D holographic images in real time, creating a realistic and immersive communication experience. This technology has broad application prospects in fields such as remote conferencing, education and training, healthcare, and entertainment. However, achieving high-quality holographic interactive communication places extremely high demands on communication systems, including ultra-high bandwidth, ultra-low latency, highly reliable network assurance, and effective support for the user's subjective experience. Existing communication quality assurance methods are mainly divided into two categories: Quality of Service (QoS) and Quality of Experience (QoE). QoS methods focus on network-layer performance indicators such as throughput, latency, jitter, and packet loss rate, using network resource scheduling to meet these indicators, thereby indirectly ensuring service quality. This type of method only focuses on network-layer performance and cannot reflect changes in the user's subjective experience during immersive interaction, nor can it perceive the differentiated needs of different interactive tasks for different experience dimensions. QoE methods, on the other hand, consider subjective factors such as user satisfaction, for example, using subjective opinion ratings derived from user evaluations to measure video experience quality. However, in immersive interactive scenarios such as holographic communication, the user experience is much richer. Beyond image clarity, subjective feelings such as immersion and presence become key factors determining experience quality. Traditional QoE assessment methods struggle to quantify these immersive metrics and are typically offline assessments, unable to be integrated into real-time network resource management. Furthermore, immersive holographic communication environments are highly dynamic and scenario-dependent. On one hand, user interaction behaviors, operation frequency, and experience preferences change in real time as the task progresses; on the other hand, network conditions and the wireless propagation environment fluctuate continuously due to interference, load changes, and other factors. User interaction behaviors and experience needs may change in real time, and network conditions and the environment (such as interference from other users and wireless channel conditions) are also constantly fluctuating. Existing QoS / QoE methods employ fixed weights or static optimization objectives, making it difficult to dynamically adjust the priority of different quality dimensions in the decision-making process based on changing scenarios. This leads to issues such as fluctuating experience or unbalanced resource allocation in complex interactive scenarios. On the one hand, it is impossible to coordinate resources uniformly from the user end to the network globally to adapt to real-time changes; on the other hand, it is impossible to adjust system operating parameters in a timely manner based on user feedback, making it difficult to consistently guarantee an ideal immersive experience in complex scenarios. Current technologies still do not adequately support immersive interactive communication and cannot meet the comprehensive requirements of 6G holographic interactive communication for high immersion, high presence, and high network reliability and low latency. Therefore, there is an urgent need for a method that can perceive multi-dimensional interaction quality in real time and dynamically coordinate end-to-end resources to improve the overall quality of immersive interactive communication. Summary of the Invention

[0003] The purpose of this invention is to provide a multi-dimensional interactive quality perception and resource optimization scheduling method for holographic communication systems. Based on the integration of user subjective experience and network objective indicators, it achieves global resource optimization scheduling to improve the quality of the user's immersive experience and ensure communication performance.

[0004] To achieve the above-mentioned technical objectives, the technical solution adopted by the present invention is as follows:

[0005] A multi-dimensional interactive quality perception and resource optimization scheduling method for holographic communication systems, the method comprising the following steps:

[0006] S1. Construct a corresponding performance index system based on the scenario characteristics of the holographic communication system. The performance index system for each scenario includes objective performance indicators of the communication network and subjective experience indicators to reflect changes in the user's perception state. In different scenarios, collect multi-dimensional index data of the network layer, holographic layer and user layer through the terminal device of the holographic communication system, and construct a representation model of subjective experience indicators for different scenarios based on the objective performance indicators of the communication network.

[0007] In the model building and parameter calibration phase, the representation model of subjective experience indicators is trained and calibrated by collecting multi-dimensional indicator data from the network layer, holographic layer, and user layer. In the system operation phase, the representation model estimates the subjective experience indicators in real time based on the collected indicator data from the network layer and holographic layer.

[0008] S2, for any scenario, predict the objective network index at time t+1 based on the objective network index at the current time t of the scenario, and predict the subjective experience index at time t and time t+1 respectively based on the representation model. The objective network index at time t and time t+1 of the scenario, together with the subjective experience index estimated by the representation model, constitute the multi-dimensional input vectors for the corresponding times; multi-source knowledge embedding terms are introduced. This is then fused with the multidimensional input vector to obtain the enhanced input vector. The multi-source knowledge embedding item It consists of physical layer knowledge, semantic layer knowledge, and cognitive layer knowledge. Physical layer knowledge is used to correct the baseline prediction value obtained from historical data based on the bandwidth limit, propagation delay constraints, and channel characteristic parameters of the communication link, so that the predicted subjective experience indicators meet the basic communication conditions. Semantic layer knowledge is used to apply semantic modulation to the prediction results of different indicators according to the semantic type or interaction scenario of the current interaction task, so that the predicted subjective experience indicators have different emphases on the evolution of interaction quality sensitive indicators under different interaction semantics. Cognitive layer knowledge is used to introduce user experience tolerance parameters constructed from historical interaction behavior or interaction quality feedback to modulate the predicted subjective experience indicators to reflect the differences in the perception of interaction quality disturbances among different users or user groups.

[0009] S3 defines an indicator utility function that maps each indicator value to a corresponding utility score value. Based on the environment and task context, a fuzzy control rule set is defined according to the contribution of each indicator to the overall interaction quality score. The dynamic weight of each indicator is inferred through fuzzy rules. The contribution of each indicator to the overall interaction quality score at the current time and the next time is calculated by combining the dynamic weights. The overall interaction quality score at the next time is obtained by dynamically weighting and fusing the indicators of each dimension at the next time.

[0010] S4. Based on the contribution calculation results, select the indicators to be optimized and construct the set of optimized indicators. With the comprehensive interaction quality score being greater than the preset score threshold as a constraint, activate the corresponding objective function set according to the set of optimized indicators, transform the task performance target optimization problem at the current moment into a multi-objective optimization problem that matches the interaction quality prediction results, and use the multi-objective optimization method to search for the optimal performance configuration scheme among the objective functions.

[0011] S5 transforms the found optimal performance configuration scheme into an executable system resource configuration scheme, and performs adaptive scheduling of network resources and terminal resources.

[0012] Furthermore, the scenarios include remote engineering collaboration, remote medical surgery, holographic social interaction, emergency command, security monitoring, and fire monitoring.

[0013] Step S1 further includes:

[0014] The scenario characteristics of the holographic communication system are analyzed to obtain a list of core requirements for each scenario, as well as objective performance indicators and subjective experience indicators for evaluating each core requirement; under different scenarios, multi-dimensional indicator data of the network layer, holographic layer and user layer are collected through the terminal devices of the holographic communication system.

[0015] Correlation analysis was performed on the collected multidimensional index data to match the objective performance index corresponding to each subjective experience index. Based on the correlation type, various models, including linear weighted fusion model, deep learning regression model, random forest ensemble model and fuzzy comprehensive evaluation model, were used to fit the representation model of each subjective experience index.

[0016] Furthermore, the network layer metrics include at least latency, jitter, packet loss rate, and bandwidth; the holographic layer metrics include at least spatial resolution, geometric fidelity, and frame rate; and the user layer metrics include at least user subjective feedback and user behavior data.

[0017] Furthermore, in step S2, external knowledge is combined with the multidimensional input vector. By fusing the input vectors, an enhanced input vector can be constructed. :

[0018]

[0019]

[0020] Among them, symbols Indicates feature concatenation operation; The first one obtained based on historical data The baseline forecast value for each indicator, Indicates the first The predicted values ​​of each indicator after incorporating external knowledge serve as the enhanced input vector. The One element; The correction amount for physical layer knowledge is determined based on the relationship between the current network environment parameters and the preset physical threshold. These are semantic layer modulation coefficients, determined based on information representing interaction semantics and task requirements in the task context parameters; The modulation amount for the cognitive layer is determined based on the user experience change trends reflected in historical interaction feedback parameters.

[0021] Step S3 further includes:

[0022] A fuzzy control rule set is designed based on the key factors affecting the interaction quality in holographic communication systems. , The key factors include at least network transmission status parameters, interaction performance parameters, and task context parameters; each rule Includes Each of the following prerequisites corresponds to at least one of the aforementioned key factors, and is used to describe the state range or level of that factor at the current moment.

[0023] A task context parameter vector is constructed based on the current environment and task context. This is used as an additional input to the preconditions of fuzzy rules to characterize the changes in rule activation intensity under different interaction scenarios. This indicates the number of environmental and task context parameters;

[0024] At any moment , by enhanced input vector and task context parameter vector Obtain fuzzy rules Dynamic membership degree:

[0025]

[0026] in, It is the first The number of prerequisites for each rule. For fuzzy rules The Middle The prerequisite at time The corresponding input variables are taken from the multidimensional index vector. Or environment and task context parameter vector These are used to characterize the network state, interaction performance, or interaction context features related to the precondition, respectively. For rules The Middle The fuzzy membership function corresponding to each precondition is used to classify the input variables. The mapping is to membership values; the fuzzy membership function adopts a predefined continuous monotonic function form.

[0027] Each indicator is obtained by weighted fusion of all fuzzy rules. Dynamic weights:

[0028]

[0029] Then time The overall scores for the interaction quality at time t+1 are as follows:

[0030]

[0031]

[0032] in, The membership function represents the knowledge context, reflecting the relative weights of different knowledge sources during the prediction stage. This is the weight adjustment coefficient for the indicator under knowledge constraints; Indicates the first The standard utility function of each indicator.

[0033] Step S4 further includes:

[0034] Based on time The overall interaction quality score at time t+1 is used to determine the trend of interaction quality changes. An optimization objective function set is constructed based on the changes in the contribution items of each dimension.

[0035] Calculate each indicator separately Changes in contribution:

[0036]

[0037] Among them, when At that time, the judgment of the first Each indicator dimension has a negative impact on the overall interaction quality score at the prediction time, and is based on... The magnitude of this indicator measures its impact on interaction quality fluctuations; construct a set of multi-objective optimization objective functions for the current scheduling period:

[0038]

[0039] in, The system resource allocation decision variables are determined by the performance targets corresponding to the indicator dimensions with significant negative impacts, and are optimized under the constraint that the comprehensive interaction quality score is not lower than a preset threshold.

[0040] In the multi-objective optimization process, the Pareto optimization method is used to search for the trade-off optimal solution among the objective functions, and the Pareto dominance relationship is defined as follows:

[0041] For any two feasible solutions and If satisfied

[0042]

[0043]

[0044] Then it is called a solution Dominant Solution A solution that is not dominated by any other solution is defined as a Pareto optimal solution. All Pareto optimal solutions constitute the Pareto optimal solution set, which serves as the candidate scheme set for subsequent resource allocation and scheduling.

[0045] Step S5 further includes:

[0046] Based on the optimal performance configuration scheme found, and according to the constraints of the available resources of the system and the change range of resource scheduling actions, several candidate resource configuration schemes are generated, and each candidate resource configuration scheme corresponds to a set of specific resource scheduling actions.

[0047] For each candidate resource scheduling action Predict its impact on the system state at the next moment and obtain the corresponding prediction index vector. And calculate the overall interaction quality score for the next time step. and the values ​​of each objective function in the predicted state. ;

[0048] The overall impact of all candidate resource scheduling actions on interaction quality, network performance, and resource consumption is quantitatively evaluated, and the candidate resource configuration scheme with the highest overall interaction quality score is selected as the optimal resource configuration scheme for the next time step.

[0049] Compared with the prior art, the beneficial effects of the present invention are as follows: Attached Figure Description

[0050] Figure 1 This is a flowchart of the multi-dimensional interactive quality perception and resource optimization scheduling method for holographic communication systems according to the present invention;

[0051] Figure 2 This is a schematic diagram of the structure of the multi-objective optimization module in this invention;

[0052] Figure 3 This is a schematic diagram of a simulation scene in an embodiment of the present invention;

[0053] Figure 4 This is a diagram showing the training and prediction results of the QoI prediction module in this embodiment of the invention;

[0054] Figure 5 This is a performance result diagram without optimization in the embodiments of the present invention;

[0055] Figure 6 This is a performance result diagram of the traditional QoS optimization method used in the embodiments of the present invention;

[0056] Figure 7 This is a graph showing the performance results of the optimization method using this solution in an embodiment of the present invention;

[0057] Figure 8 This is a performance comparison chart of the simulation environment in the embodiments of the present invention. Detailed Implementation

[0058] The embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0059] See appendix Figure 1 This invention discloses a multi-dimensional interactive quality perception and resource optimization scheduling method for holographic communication systems, the method comprising the following steps:

[0060] S1. Construct a corresponding performance index system based on the scenario characteristics of the holographic communication system. The performance index system for each scenario includes objective performance indicators of the communication network and subjective experience indicators to reflect changes in the user's perception state. In different scenarios, collect multi-dimensional index data of the network layer, holographic layer and user layer through the terminal device of the holographic communication system, and construct a representation model of subjective experience indicators for different scenarios based on the objective performance indicators of the communication network.

[0061] In the model building and parameter calibration phase, the representation model of subjective experience indicators is trained and calibrated by collecting multi-dimensional indicator data from the network layer, holographic layer, and user layer. In the system operation phase, the representation model estimates the subjective experience indicators in real time based on the collected indicator data from the network layer and holographic layer.

[0062] S2, for any scenario, predict the objective network index at time t+1 based on the objective network index at the current time t of the scenario, and predict the subjective experience index at time t and time t+1 respectively based on the representation model. Construct the multi-dimensional input vectors for the corresponding time by combining the objective network index at time t and time t+1 of the scenario with the subjective experience index estimated by the representation model. and Introducing multi-source knowledge embedding items This is then fused with the multidimensional input vector to obtain the enhanced input vector. The multi-source knowledge embedding item It consists of physical layer knowledge, semantic layer knowledge, and cognitive layer knowledge. Physical layer knowledge is used to correct the baseline prediction value obtained from historical data based on the bandwidth limit, propagation delay constraints, and channel characteristic parameters of the communication link, so that the predicted subjective experience indicators meet the basic communication conditions. Semantic layer knowledge is used to apply semantic modulation to the prediction results of different indicators according to the semantic type or interaction scenario of the current interaction task, so that the predicted subjective experience indicators have different emphases on the evolution of interaction quality sensitive indicators under different interaction semantics. Cognitive layer knowledge is used to introduce user experience tolerance parameters constructed from historical interaction behavior or interaction quality feedback to modulate the predicted subjective experience indicators to reflect the differences in the perception of interaction quality disturbances among different users or user groups.

[0063] S3 defines an indicator utility function that maps each indicator value to a corresponding utility score value. Based on the environment and task context, a fuzzy control rule set is defined according to the contribution of each indicator to the overall interaction quality score. The dynamic weight of each indicator is inferred through fuzzy rules. The contribution of each indicator to the overall interaction quality score at the current time and the next time is calculated by combining the dynamic weights. The overall interaction quality score at the next time is obtained by dynamically weighting and fusing the indicators of each dimension at the next time.

[0064] S4. Based on the contribution calculation results, select the indicators to be optimized and construct the set of optimized indicators. With the comprehensive interaction quality score being greater than the preset score threshold as a constraint, activate the corresponding objective function set according to the set of optimized indicators, transform the task performance target optimization problem at the current moment into a multi-objective optimization problem that matches the interaction quality prediction results, and use the multi-objective optimization method to search for the optimal performance configuration scheme among the objective functions.

[0065] S5 transforms the found optimal performance configuration scheme into an executable system resource configuration scheme, and performs adaptive scheduling of network resources and terminal resources.

[0066] Holographic communication systems are used in a variety of scenarios, with common applications including remote engineering collaboration, remote medical surgery, holographic social networking, emergency command, security monitoring, and fire monitoring. The focus of the indicator systems differs across these scenarios. Therefore, in step S1, the scenario characteristics of the holographic communication system are first analyzed to obtain a list of core requirements for each scenario, as well as objective performance indicators and subjective experience indicators for evaluating each core requirement. Under different scenarios, multi-dimensional indicator data from the network layer, holographic layer, and user layer are collected through the terminal devices of the holographic communication system.

[0067] For example, in remote medical surgery scenarios, the core requirements are operational precision and real-time feedback. Objective performance indicators focus on ultra-low latency, microsecond-level jitter, zero packet loss, and stable bandwidth. Subjective experience indicators include the sense of presence during operation, image synchronization, and the perception of surgical operation error tolerance. For holographic social scenarios, the core requirements are immersive interaction and natural communication. Objective performance indicators emphasize spatial resolution, geometric fidelity, frame rate, and dynamic bandwidth adaptation capabilities. Subjective experience indicators include satisfaction with character image reproduction, voice-action synchronization perception, and environmental immersion. For emergency command scenarios, the core requirements are multi-terminal collaboration and real-time information sharing. Objective performance indicators include multi-link redundancy, anti-interference capability, and data transmission latency stability. Subjective experience indicators include the clarity of command transmission, the smoothness of multi-view switching, and the perception of emergency scenario response.

[0068] Each scenario's performance indicator system includes two core categories: first, objective performance indicators of the communication network, which directly reflect the technical parameters of network transmission and holographic presentation; and second, subjective user experience indicators, which depict changes in the user's perceptual state during immersive interaction, covering dimensions such as immersion, presence, operational smoothness, and content reproduction satisfaction.

[0069] In different scenarios, the terminal devices of the holographic communication system (including holographic display terminals, interactive control devices, network monitoring modules, etc.) synchronously collect multi-dimensional indicator data from the network layer, holographic layer, and user layer, forming a closed loop of end-to-end data acquisition. Specifically, the network monitoring module collects network layer indicator data in real time, including transmission latency, latency jitter, packet loss rate, link bandwidth utilization, channel signal-to-noise ratio, signal attenuation coefficient, etc., to ensure the capture of dynamic changes in network status; the built-in sensors and rendering module of the holographic display terminal collect holographic layer indicator data, including spatial resolution, geometric fidelity, texture mapping accuracy, frame rate, rendering latency, 3D model loading speed, etc., directly reflecting the presentation quality of holographic content; the sensors and subjective feedback entry points of the interactive devices collect user layer indicator data, including user behavior data (operation response time, viewpoint switching frequency, interaction command type) and subjective feedback data (scenario-based satisfaction rating, experience dimension level evaluation, problem feedback text), where subjective feedback is collected in real time through the terminal's built-in rating interface, and the frequency of behavior data collection is synchronized with the interaction actions. During the data collection process, timestamp alignment and format standardization were performed on various indicator data, and the data was uniformly stored as a structured dataset to ensure the spatiotemporal consistency of data at different levels and provide high-quality data support for subsequent model training.

[0070] Next, correlation analysis was performed on the collected multidimensional index data to match the objective performance index corresponding to each subjective experience index. Based on the correlation type, various models, including linear weighted fusion model, deep learning regression model, random forest ensemble model and fuzzy comprehensive evaluation model, were used to fit the representation model of each subjective experience index. Specifically, the collected multi-dimensional index data from the network layer, holographic layer, and user layer are preprocessed, including outlier removal (using the 3σ criterion), data normalization (mapping to the [0,1] interval), and feature selection (screening key influencing factors using Pearson correlation coefficient and mutual information entropy). Subsequently, for each subjective experience index, a strongly correlated objective performance index is matched (e.g., presence is strongly correlated with spatial resolution, latency, and geometric fidelity), and an appropriate model structure is selected according to the correlation type. For example, for linearly correlated scenarios (e.g., bandwidth utilization and perceived smoothness), a linear weighted fusion model is used; for nonlinear complex mapping scenarios (e.g., multiple indexes coupled to affect immersion), a deep learning regression model (e.g., a CNN-LSTM hybrid model) or a random forest ensemble model is used; for scenarios with strong ambiguity (e.g., the correlation between subjective satisfaction and multi-dimensional objective indicators), a fuzzy comprehensive evaluation model is used. The preprocessed dataset is divided into training and validation sets proportionally. The training set is used to iteratively train the representation model, optimizing the model parameters using gradient descent. The validation set is used to verify the model's output, calculating the error between the predicted subjective experience index and the actual user feedback. If the error exceeds a preset threshold, the feature input and model structure are adjusted, and retraining is performed until the accuracy requirements are met. Preferably, a cross-scene transfer learning mechanism can be introduced, using the parameters of a well-trained scene model as initial values ​​to transfer to the training of a new scene model, shortening the model convergence time. The representation model for the subjective experience index can be preset or calibrated according to the specific task scenario, terminal device capabilities, and network type.

[0071] During the formal operation of the holographic communication system, the representation model does not rely on real-time subjective feedback from users. It can quickly output estimated values ​​of subjective experience indicators solely through real-time collection of network layer and holographic layer indicator data from terminal devices. By inputting network layer indicator data (latency, jitter, packet loss rate, etc.) and holographic layer indicator data (resolution, frame rate, fidelity, etc.) into the representation model in real time, and ensuring the input data update frequency matches the collection frequency, the model can rapidly perform calculations on the input data using the corresponding mapping relationship, outputting corresponding estimated values ​​of subjective experience indicators, including experience scores for each dimension and an overall experience rating. When newly collected user subjective feedback data becomes available, the model update mechanism can be automatically invoked to fine-tune the model parameters, ensuring that estimation accuracy is maintained even when network conditions or user needs change, thus achieving real-time and accurate quantification of subjective experience indicators.

[0072] Based on the obtained indicators, construct 3D input vector ,in Indicates the number of input metrics. This refers to input metrics, including subjective and objective metrics such as bandwidth, latency, immersion, and remote rendering.

[0073] In this embodiment, to evaluate the interaction quality during holographic communication (eight indicators: network bandwidth, communication latency, packet loss rate, jitter, visual quality, auditory quality, immersion, and remote presentation), the objective transmission status of the communication network is first quantitatively collected. These objective quality indicators include available bandwidth. End-to-end communication delay Packet loss rate And network jitter The aforementioned objective quality indicators are acquired in real time by the terminal device during the interaction process and are used to characterize the transmission capacity and stability of the current network environment.

[0074] Based on this, a subjective experience index representation model is constructed to calculate the user's subjective experience index according to the objective quality index. The subjective experience index representation model calculates visual quality according to the following formula:

[0075]

[0076] in, It is the feature bitrate for acceptable image quality (given by the target resolution, frame rate, and encoding configuration); It is the severity of the visual penalty imposed by stuttering.

[0077] The subjective experience index representation model calculates auditory quality according to the following formula:

[0078]

[0079] in, It is the characteristic bit rate of acceptable sound (given by the encoding configuration); It is the intensity of the auditory penalty imposed by the stutter; It is the jitter equivalent time delay coefficient.

[0080] The subjective experience index representation model calculates immersion based on the following formula:

[0081]

[0082] in, The weights for vision and hearing are respectively. ; It's the severity of the penalty that stuttering inflicts on immersion; It is the tolerance scale of immersion for latency.

[0083] The subjective experience index representation model calculates the remote presentation feeling based on the following formula:

[0084]

[0085] Where T is the half-saturation point of the remote presentation; It's the severity of the penalty that stuttering inflicts on immersion; It is the tolerance scale for latency in remote presentation.

[0086] After obtaining the aforementioned objective quality indicators and subjective experience indicators, the objective and subjective indicators are organized in a unified manner to construct a multi-dimensional input vector for subsequent comprehensive evaluation and prediction of interaction quality:

[0087]

[0088] The input vector is used for subsequent calculation of the overall interaction quality score, QoI prediction, and resource optimization scheduling decisions.

[0089] In step S2, external knowledge is combined with multidimensional input vectors. By fusing the input vectors, an enhanced input vector can be constructed. :

[0090]

[0091]

[0092] Among them, symbols Indicates feature concatenation operation; The first one obtained based on historical data The baseline forecast value for each indicator, Indicates the first The predicted values ​​of each indicator after incorporating external knowledge serve as the enhanced input vector. The One element; The correction amount for physical layer knowledge is determined based on the relationship between the current network environment parameters and the preset physical threshold. These are semantic layer modulation coefficients, determined based on information representing interaction semantics and task requirements in the task context parameters; The modulation amount for the cognitive layer is determined based on the user experience change trends reflected in historical interaction feedback parameters.

[0093] To enhance the model's cognitive and generalization abilities, multi-source knowledge embeddings are introduced. The multi-source knowledge embedding term is used to constrain the value range and trend of the QoI prediction results, avoiding deviations from the actual physical environment and the semantics of the interaction task, thereby ensuring that the prediction results can be used for subsequent resource scheduling decisions. External knowledge is fused with feature vectors to construct an enhanced input:

[0094]

[0095] Among them, symbols This indicates a feature splicing operation. It consists of knowledge at the physical, semantic, and cognitive levels. Specifically:

[0096] Physical layer knowledge is based on parameters such as the bandwidth limit of the communication link, propagation delay constraints, and channel characteristics to ensure that the prediction results meet the basic communication conditions. The physical layer correction amount is determined based on the relationship between the current network environment parameters and preset physical thresholds. For example, when the available bandwidth is higher than the service demand threshold, a positive value is used to correct image quality and resolution metrics upwards; when the link load level is high, a negative value is used to correct latency and packet loss rate metrics downwards; when the end-to-end latency is lower than the real-time requirement limit, a positive value is used to correct real-time performance and interactivity metrics upwards.

[0097] Semantic layer knowledge, based on the semantic type or interaction scenario of the current interaction task, applies semantic modulation to the prediction results of different indicators, causing the prediction model to emphasize different aspects of the evolution of interaction quality-sensitive indicators under different interaction semantics. The semantic modulation coefficients are determined based on information representing the interaction semantics and task requirements in the task context parameters. For example, when the task context parameters indicate a high real-time performance level, the latency and jitter-related indicators will... Take the larger value; the corresponding image quality-related indicators Take the smaller value; when the task context parameters indicate a high interaction intensity, the stability and continuity indicators correspond to... Take the larger value.

[0098] The cognitive layer knowledge involves incorporating user experience tolerance parameters built from historical interaction behaviors or interaction quality feedback to modulate the prediction results of experience-related indicators, reflecting the perceptual differences of different users or user groups in response to interaction quality disturbances. The cognitive layer modulation amount is determined based on the user experience change trends reflected by historical interaction feedback parameters. For example, when the number of interaction interruptions per unit time exceeds a preset threshold, a negative value is used to downwardly adjust indicators related to immersion or presence; when the fallback frequency or interaction failure rate is consistently low, a positive value is used to enhance the trend of stable user experience in the prediction results.

[0099] Step S3 further includes:

[0100] A fuzzy control rule set is designed based on the key factors affecting the interaction quality in holographic communication systems. , The key factors include at least network transmission status parameters, interaction performance parameters, and task context parameters; each rule Includes There are several prerequisites; each prerequisite corresponds to at least one of the aforementioned key factors, used to describe the state range or level of that factor at the current moment; wherein, the prerequisites used to characterize the network transmission state or interaction performance correspond to a multi-dimensional index vector. The specific indicator parameters include at least one or more of latency, packet loss rate, bandwidth, jitter, visual quality, auditory quality, or immersion; the preconditions used to characterize the task context correspond to the task context parameter vector. The specific parameters in the task context include at least one or more of the following: real-time level, interaction intensity, or task stage status.

[0101] A task context parameter vector is constructed based on the current environment and task context. This is used as an additional input to the preconditions of fuzzy rules to characterize the changes in rule activation intensity under different interaction scenarios. The number of environmental and task context parameters is represented; each precondition is mapped to the corresponding input parameter through a preset fuzzy membership function to characterize the degree of influence of the parameter on the interaction quality when it is in different state intervals, thus constituting the premise part of the fuzzy rule.

[0102] At any moment Fuzzy rules The membership degree of activation level is:

[0103]

[0104] in, For rules The Middle The prerequisite at time The corresponding input variables, which are derived from the multidimensional index vector according to the type of the precondition. Or environment and task context parameter vector Selected from; when the aforementioned preconditions are used to characterize network state or interaction performance features, Taken from the multidimensional index vector When the aforementioned preconditions are used to characterize contextual features such as interactive scenarios, task stages, or real-time requirements, Taken from the task context parameter vector ; For rules The Middle The fuzzy membership function corresponding to each precondition is used to classify the input variables. The fuzzy membership function is mapped to a membership value, with a range of [0,1]. It adopts a predefined continuous monotonic function form. For example, when latency is in a high membership interval and real-time requirements are high, the membership degree of rules related to low latency will be enhanced; when the network state is stable and the focus is on immersive experience, the membership degree of rules related to user experience will be enhanced.

[0105] In this embodiment, to evaluate the interaction quality during holographic communication, eight indicators are selected as input indicators: network bandwidth, communication latency, packet loss rate, jitter, visual quality, auditory quality, immersion, and remote presentation. A multi-dimensional indicator vector is constructed as follows:

[0106]

[0107] in, Indicates network bandwidth; Indicates communication delay; Indicates packet loss rate; Indicates shaking; Indicates visual quality; Indicates hearing quality; Indicates a sense of immersion; It conveys a sense of remote presentation.

[0108] Simultaneously, construct the task context parameter vector:

[0109]

[0110] in, Indicates the real-time level of the interactive task; This indicates the current stage of the task.

[0111] During the rule-building phase, a fuzzy control rule set was designed to address the impact of the eight interaction quality indicators on the overall interaction quality score under different task scenarios. One of the fuzzy rules... The definition is as follows:

[0112] If the communication latency is high, the packet loss rate is high, and the task real-time level is high, then the weight of indicators related to latency and packet loss rate in the interaction quality assessment should be increased.

[0113] The rule includes three preconditions, each of which establishes a definite correspondence with a specific dimension in the multidimensional indicator vector or task context parameter vector:

[0114] The first prerequisite is high communication latency. This prerequisite is used to describe the communication latency parameter in the network transmission state.

[0115] During the rule-building phase, this prerequisite is combined with the multidimensional indicator vector. The dimension representing communication delay Establish the correspondence; therefore, during system operation, the input variables for this prerequisite are:

[0116]

[0117] The second prerequisite is a high packet loss rate. This prerequisite describes the packet loss rate parameter in the network transmission state.

[0118] During the rule-building phase, this prerequisite is combined with the multidimensional indicator vector. Dimensions representing packet loss rate Establish the correspondence; therefore, during system operation, the input variables for this prerequisite are:

[0119]

[0120] The third prerequisite: the task real-time level is high. This prerequisite is used to describe the real-time context of the interactive task.

[0121] During the rule-building phase, this prerequisite is combined with the task context parameter vector. Dimensions representing the real-time performance level Establish the correspondence; therefore, during system operation, the input variables for this prerequisite are:

[0122]

[0123] Under different environments and task contexts, the membership degrees corresponding to different preconditions will change, leading to dynamic adjustments in the activation level of rules. Based on dynamic membership degrees, various indicators... The dynamic weights are obtained by weighted fusion of all fuzzy rules:

[0124]

[0125] in, The membership function represents the knowledge context, reflecting the relative weights of different knowledge sources during the prediction stage. This is the weight adjustment coefficient for the indicator under knowledge constraints.

[0126] Based on this, the overall score for interaction quality at any time t is:

[0127]

[0128] To obtain the forecast indicators for the next moment under the current resource allocation conditions. Based on this, the overall score for predicted interaction quality is calculated as follows:

[0129]

[0130] Comparing the two scores above allows us to determine the trend of interaction quality over time, which is then used for adaptive updates in subsequent multi-objective optimization. Based on the multi-dimensional quality assessment, the overall interaction quality score is obtained by weighting the sub-utilities of each dimension with dynamic weights. Therefore, at the current and predicted times, the contribution of each dimension to QoI can be compared dimension by dimension. Specifically, for any one dimension... ,have:

[0131]

[0132] Among them, when At that time, the judgment of the first Each indicator dimension has a negative impact on the overall interaction quality score at the prediction time, and is based on... The magnitude of the value measures the degree of influence of this indicator on the fluctuation of interaction quality.

[0133] Based on the contribution results of the above-mentioned indicators, the system transforms the current task performance objective into a multi-objective optimization problem that matches the QoI prediction results. Specifically, the system determines the objective function that needs to be prioritized for optimization in the current period based on the contribution of the QoI prediction results, and constructs a set of multi-objective optimization functions accordingly.

[0134] In this multi-objective optimization problem, the system selects multiple optimization objective functions based on actual interaction requirements. , , ... , In this multi-objective optimization problem, the system selects multiple objective functions based on actual interaction requirements. For example, To maximize the overall QoI score , To minimize network latency, To minimize one or more of the following: terminal energy consumption.

[0135] It should be noted that the set of multi-objective optimization functions mentioned is not for a fixed set of objectives. Instead of performing constant solutions, it dynamically determines the current set of effective objective functions and their optimization directions based on the QoI prediction results.

[0136] In a specific embodiment, the system uses the overall interaction quality score as the objective function. As the core objective; when the prediction results are obtained Below the preset threshold When a significant downward trend is observed, the system selects the objective function corresponding to the performance dimension that most significantly leads to the decrease in QoI score, based on the aforementioned QoI contribution decomposition results, as the key optimization objective for the current scheduling cycle. For example, when the contribution of network latency-related metrics to QoI decreases significantly, the objective function that minimizes network latency will be selected. Incorporate into the current set of optimization objectives; when the indicator dimension related to terminal energy consumption or resource load has the most significant impact on QoI, minimize the objective function of terminal energy consumption. Include it in the current set of optimization objectives.

[0137] Based on the above comprehensive interaction quality score As a constraint (e.g., maintain) To ensure basic interaction quality, a set of multi-objective optimization functions is constructed:

[0138]

[0139] in, This represents the threshold for the overall QoI score, ensuring that the system meets basic interaction quality standards. It is the feasible solution space that satisfies all constraints. The specific task requirements determine the direction of optimization (maximization or minimization).

[0140] When the prediction results indicate that the overall interaction quality score has stably met the threshold requirements and there are no obvious experience bottlenecks, the system dynamically adjusts the optimization focus to ensure... Under the premise of [missing information], the set of objective functions is switched to resource efficiency-related objectives to reduce resource consumption or energy expenditure.

[0141] Therefore, within different scheduling cycles, the set of objective functions and their optimization directions for the multi-objective optimization problem can be updated with the QoI prediction module, realizing the dynamic derivation of task performance objectives.

[0142] The multi-objective optimization module, under the constraints of the currently enabled set of objective functions, employs a multi-objective optimization method to search for the optimal resource allocation scheme that compromises among the objective functions. In practice, the system does not pursue the extreme value of a single objective. Instead, while ensuring that the overall interaction quality score meets the threshold constraint, it comprehensively considers the interrelationships between multiple performance objectives and selects the optimal resource allocation scheme that achieves a compromise across all objective functions from the candidate solutions. Specifically, if a solution... A solution is considered optimal if it is impossible to improve any objective without decreasing other objective values. The dominance relation is defined as:

[0143]

[0144]

[0145] in, This represents a set of objective functions, each used to evaluate different performance metrics of the system. It can represent the overall QoI score. Indicates network latency. This indicates the terminal energy consumption. This indicates that it corresponds to the optimal solution. The objective function value, where the superscript is... This represents the optimal value. Specifically, , and These represent the best QoI score, best latency, and best energy consumption, respectively. If a solution... It is superior to another solution in at least one objective. If a solution is not inferior to the latter in all other objectives, then it is called a solution. Dominant Solution A solution is called a Pareto optimal solution if it is not dominated by any other solution. The set of all Pareto optimal solutions constitutes the Pareto front, representing the trade-off boundary between conflicting objectives. The set of solutions not dominated by other solutions can be represented as:

[0146]

[0147] in, express Dominate These Pareto optimal solutions constitute a subset of the solution space, and the corresponding objective function values ​​are... The Pareto front is defined. The Pareto optimal front set is defined as follows:

[0148]

[0149] Appendix Figure 2 The multi-objective optimization module outputs several Pareto optimal solution candidates and provides the performance score of each candidate on the objective function, which serves as a set of candidate solutions for subsequent resource allocation and scheduling.

[0150] Finally, based on the optimal performance configuration scheme found through compromise, and considering the constraints of the available system resources and the magnitude of resource scheduling action changes, several candidate resource configuration schemes are generated. Each candidate resource configuration scheme corresponds to a set of specific resource scheduling actions, including: selecting different configuration levels within the allowed bandwidth or bitrate range; adjusting scheduling priority or redundancy intensity while meeting QoS constraints; and switching different energy consumption or computing load states on the terminal side. The candidate resource configuration schemes correspond to different values. For each candidate resource scheduling action Predict its impact on the system state at the next moment and obtain the corresponding prediction index vector. And calculate the overall interaction quality score for the next time step. and the values ​​of each objective function in the predicted state. The algorithm quantitatively evaluates the comprehensive impact of all candidate resource scheduling actions on interaction quality, network performance, and resource consumption, and selects the candidate resource configuration scheme with the highest comprehensive interaction quality score as the optimal resource configuration scheme for the next moment, thereby achieving adaptive scheduling of network resources and terminal resources in different interaction stages.

[0151] The system selects the optimal resource allocation scheme that achieves a trade-off across all objective functions based on the set of objective functions enabled in the current scheduling cycle and their optimization directions. For example, when QoI prediction indicates a risk of deterioration in the user experience, the system prioritizes the option that can significantly improve the experience. Alternatively, improve resource allocation schemes that address the primary bottleneck indicators; when QoI predictions show that interaction quality is consistently meeting standards, prioritize resource allocation schemes that ensure... Under the premise of reducing network latency or terminal power consumption, a resource allocation scheme is proposed. After selecting the optimal resource allocation scheme, the system maps it to the final resource scheduling action vector. The system then distributes the data to the network and terminal sides for execution. To avoid frequent resource configuration changes that could cause fluctuations in the user experience, the system can limit the magnitude of changes in resource scheduling actions between adjacent scheduling cycles, ensuring a continuous and smooth resource adjustment process.

[0152]

[0153] After completing the resource scheduling action, the system enters the next scheduling cycle and performs QoI evaluation and prediction again based on the updated system state, thus forming a continuous dynamic resource allocation closed loop. Through this end-to-end adaptive resource control method, this invention can continuously improve the user's immersive experience quality and ensure communication performance during holographic communication interaction.

[0154] The effects of the present invention will be further explained below with reference to simulation experiments.

[0155] 1. Simulation conditions and parameter settings:

[0156] In one embodiment of the invention, the simulation is conducted in an immersive virtual environment designed for fire detection tasks to verify the performance of the proposed multidimensional interactive quality assessment and optimization method. This simulation environment is based on... Figure 3 The spherical screen projection system shown, combined with a virtual reality (VR) terminal and a virtual drone with multimodal perception capabilities, simulates the real process of a user performing a three-dimensional holographic interactive task in a remote environment. In this embodiment, the following simulation conditions and parameter configurations are set:

[0157] Task Setup: The virtual drone performs fire zone detection and risk monitoring tasks in a simulated space. Users can observe, make decisions, and control the drone's behavior in real time through a VR interactive terminal.

[0158] Network conditions: Communication link parameters include bandwidth (range 5–100 Mbps), latency (10–200 ms), packet loss rate (0–10%), and jitter (0–50 ms). All network data are generated based on real cellular network measurement samples, and interference is introduced at random time intervals to simulate network fluctuations.

[0159] QoI evaluation metrics: Eight core metrics are monitored in real time during the simulation process, including network bandwidth, communication latency, packet loss rate, jitter, visual quality, auditory quality, immersion, and remote presentation.

[0160] Interactive dimension control: The simulator dynamically adjusts image quality (such as video encoding rate), audio reproduction accuracy, motion tracking latency, and error control parameters to reproduce the communication experience changes in typical dynamic interactive scenarios.

[0161] Subjective experience collection: Multiple volunteers participated in the task execution, and their task completion time, operational accuracy, and immersion experience scores were recorded. The obtained subjective scores were used to train the representation model of subjective experience indicators and calibrate the comprehensive QoI score. .

[0162] Platform support: The simulation is based on a distributed holographic interactive communication platform, which supports adjustable network parameters and QoI feedback, ensuring that the relationship between resource allocation strategies and changes in user experience is observable and verifiable.

[0163] The simulation scenario described is merely an example to illustrate the effectiveness of the method of the present invention in complex immersive interactive tasks. The present invention is not limited to fire monitoring scenarios, but can also be applied to holographic interactive communication scenarios such as telemedicine, industrial collaboration, and immersive education.

[0164] Table 1 shows the parameter configuration of the subjective experience index representation model in the embodiments.

[0165]

[0166] 2. Simulation content:

[0167] Appendix Figure 4This chart compares the actual and predicted values ​​of the QoI prediction module based on Transform for eight key indicators: network bandwidth, communication latency, packet loss rate, jitter, visual quality, auditory quality, immersion, and telepresence. It demonstrates the prediction accuracy of the module in typical task scenarios. The radar chart reflects the model's accurate characterization of the coupling relationships between various quality dimensions. The prediction results are highly consistent with the actual observations, especially for key indicators such as packet loss rate, communication latency, and visual quality, where the prediction error is extremely small. This verifies that the prediction method proposed in this invention possesses good stability and accuracy under the fusion modeling of subjective and objective indicators. This predictive capability provides a reliable basis for subsequent dynamic resource allocation and system optimization.

[0168] To further verify the effectiveness of the multi-dimensional interaction quality assessment and optimization method proposed in this invention in improving interaction quality in holographic communication scenarios, the following three comparative test schemes were designed:

[0169] a) No optimization scheme: In this scheme, the system does not adopt any optimization strategy. The UAV flies at a fixed speed along a preset path. The video encoding rate and other parameters are set to default values. Network resources are evenly distributed in the system. As a baseline without communication quality assurance, it is impossible to achieve effective control over the interaction quality.

[0170] b) Traditional network QoS optimization scheme: This scheme only optimizes based on network layer QoS indicators and adopts a threshold-based heuristic scheduling strategy. When the communication latency is detected to exceed 50ms, the system automatically reduces parameters such as video bitrate to ensure network transmission stability. However, it does not consider the user's subjective interactive experience feedback and cannot achieve comprehensive optimization in the interactive dimension.

[0171] c) A QoI-based perception optimization scheme (the scheme of this invention): This scheme adopts the complete closed-loop optimization framework proposed in this invention, which integrates multi-dimensional interaction quality score function, Pareto multi-objective optimization and predictive feedback mechanism. In the process of resource scheduling, it jointly considers network status and user experience, dynamically adjusts transmission parameters to stabilize interaction quality, and realizes feedforward control and active optimization adjustment of indicators through QoI prediction mechanism.

[0172] To ensure the fairness of the comparison results, all experiments were conducted on the same simulation platform and under the same mission conditions. The UAVs performed the same mission procedures and flight paths to evaluate the differences in the actual effects of each scheme in terms of interaction quality assurance.

[0173] Appendix Figure 5The system exhibits significant fluctuations in eight core indicators within 30 seconds in the unoptimized scheme. In particular, around the 15th second, after the drone approaches the fire source, the network environment is affected by interference, communication latency and packet loss rate rise sharply, visual and auditory quality decline rapidly, and subjective scores such as immersion and presence also drop significantly, ultimately leading to a low level of interaction quality score and slow overall recovery.

[0174] Appendix Figure 6 This test assessed the performance changes of eight core metrics in a traditional QoS scheme over a 30-second period. The system mitigated communication fluctuations during interference by adjusting network metrics such as bandwidth, latency, and packet loss rate. Particularly between the 15th and 25th seconds, the system proactively reduced the video bitrate to ensure stable communication, significantly decreasing latency fluctuations and maintaining an average latency score of approximately 0.48. However, due to the compromised video clarity, the visual and immersive experience scores remained low, and the subjective experience was not significantly improved.

[0175] Appendix Figure 7 This invention utilizes the performance changes of eight core indicators within 30 seconds using the proposed QoI perception optimization scheme. When the system detects early signs of interference, it adjusts the video encoding strategy in advance, dynamically compressing data to alleviate potential network pressure, based on feedback from the QoI prediction module. During peak interference periods (approximately from the 15th to the 22nd second), the system effectively controls communication latency and packet loss rate, keeping the average latency at approximately 0.47, while maintaining an average visual quality score above 0.65. Furthermore, it achieves a performance trade-off among multiple indicators through Pareto optimal configuration. As interference diminishes, the system quickly restores image quality, resulting in significantly better immersion and user satisfaction compared to the previous two schemes.

[0176] Appendix Figure 8 A quantitative comparison of the comprehensive performance of the three schemes at the task level was conducted. The results show that the experimental group using the QoI perception optimization scheme proposed in this invention outperformed the two comparative schemes in terms of immersion level and interaction smoothness. During the experiment, the drone video footage showed almost no lag, and participants were able to identify and respond to fire events more quickly. Compared with the unoptimized scheme, the QoI optimization scheme reduced the average task response time by 2.4 seconds, decreased the false alarm rate by 60%, shortened the overall task completion time by 33%, and improved the fire source detection accuracy by 45%.

[0177] Based on the simulation results and performance analysis above, it is evident that the multi-dimensional interaction quality perception and resource optimization scheduling method for holographic communication systems proposed in this invention can effectively improve interaction stability and resource allocation in complex dynamic network environments. Compared to existing optimization methods, the proposed solution not only possesses stronger subjective and objective indicator perception capabilities and a superior system response strategy, but also proactively predicts interaction quality fluctuations and adjusts key indicator parameters in advance under resource constraints and sudden interference, maintaining the continuity and consistency of the user's immersive experience.

[0178] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0179] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A multi-dimensional interactive quality perception and resource optimization scheduling method for holographic communication systems, characterized in that, The method includes the following steps: S1. Construct a corresponding performance index system based on the scenario characteristics of the holographic communication system. The performance index system for each scenario includes objective performance indicators of the communication network and subjective experience indicators to reflect changes in the user's perception state. In different scenarios, collect multi-dimensional index data of the network layer, holographic layer and user layer through the terminal device of the holographic communication system, and construct a representation model of subjective experience indicators for different scenarios based on the objective performance indicators of the communication network. In the model building and parameter calibration phase, the representation model of subjective experience indicators is trained and calibrated by collecting multi-dimensional indicator data from the network layer, holographic layer, and user layer. In the system operation phase, the representation model estimates the subjective experience indicators in real time based on the collected indicator data from the network layer and holographic layer. S2, for any scenario, predict the objective network index at time t+1 based on the objective network index at the current time t of the scenario, and predict the subjective experience index at time t and time t+1 respectively based on the representation model. The objective network index at time t and time t+1 of the scenario, together with the subjective experience index estimated by the representation model, constitute the multi-dimensional input vectors for the corresponding times; multi-source knowledge embedding terms are introduced. This is then fused with the multidimensional input vector to obtain the enhanced input vector. The multi-source knowledge embedding item It consists of physical layer knowledge, semantic layer knowledge, and cognitive layer knowledge. Physical layer knowledge is used to correct the baseline prediction value obtained from historical data based on the bandwidth limit, propagation delay constraints, and channel characteristic parameters of the communication link, so that the predicted subjective experience indicators meet the basic communication conditions. Semantic layer knowledge is used to apply semantic modulation to the prediction results of different indicators according to the semantic type or interaction scenario of the current interaction task, so that the predicted subjective experience indicators have different emphases on the evolution of interaction quality sensitive indicators under different interaction semantics. Cognitive layer knowledge is used to introduce user experience tolerance parameters constructed from historical interaction behavior or interaction quality feedback to modulate the predicted subjective experience indicators to reflect the differences in the perception of interaction quality disturbances among different users or user groups. S3 defines an indicator utility function that maps each indicator value to a corresponding utility score value. Based on the environment and task context, a fuzzy control rule set is defined according to the contribution of each indicator to the overall interaction quality score. The dynamic weight of each indicator is inferred through fuzzy rules. The contribution of each indicator to the overall interaction quality score at the current time and the next time is calculated by combining the dynamic weights. The overall interaction quality score at the next time is obtained by dynamically weighting and fusing the indicators of each dimension at the next time. S4. Based on the contribution of each indicator to the overall interaction quality score, select the indicators to be optimized and construct an optimized indicator set. With the overall interaction quality score being greater than a preset score threshold as a constraint, activate the corresponding objective function set according to the optimized indicator set, transform the task performance target optimization problem at the current moment into a multi-objective optimization problem that matches the interaction quality prediction result, and use a multi-objective optimization method to search for the optimal performance configuration scheme among the objective functions. S5 transforms the found optimal performance configuration scheme into an executable system resource configuration scheme, and performs adaptive scheduling of network resources and terminal resources.

2. The multi-dimensional interactive quality perception and resource optimization scheduling method for holographic communication systems according to claim 1, characterized in that, The scenarios include remote engineering collaboration, remote medical surgery, holographic social interaction, emergency command, security monitoring, and fire monitoring.

3. The multi-dimensional interactive quality perception and resource optimization scheduling method for holographic communication systems according to claim 1, characterized in that, Step S1 further includes: The scenario characteristics of the holographic communication system are analyzed to obtain a list of core requirements for each scenario, as well as objective performance indicators and subjective experience indicators for evaluating each core requirement; under different scenarios, multi-dimensional indicator data of the network layer, holographic layer and user layer are collected through the terminal devices of the holographic communication system. Correlation analysis was performed on the collected multidimensional index data to match the objective performance index corresponding to each subjective experience index. Based on the correlation type, various models, including linear weighted fusion model, deep learning regression model, random forest ensemble model and fuzzy comprehensive evaluation model, were used to fit the representation model of each subjective experience index.

4. The multi-dimensional interactive quality perception and resource optimization scheduling method for holographic communication systems according to claim 1, characterized in that, The network layer metrics include at least latency, jitter, packet loss rate, and bandwidth; the holographic layer metrics include at least spatial resolution, geometric fidelity, and frame rate; and the user layer metrics include at least user subjective feedback and user behavior data.

5. The multi-dimensional interactive quality perception and resource optimization scheduling method for holographic communication systems according to claim 1, characterized in that, In step S2, external knowledge is combined with multidimensional input vectors. By fusing the input vectors, an enhanced input vector can be constructed. : ； ； Among them, symbols Indicates feature concatenation operation; The first one obtained based on historical data The baseline forecast value for each indicator, Indicates the first The predicted values ​​of each indicator after incorporating external knowledge serve as the enhanced input vector. The One element; The correction amount for physical layer knowledge is determined based on the relationship between the current network environment parameters and the preset physical threshold. These are semantic layer modulation coefficients, determined based on information representing interaction semantics and task requirements in the task context parameters; The modulation amount for the cognitive layer is determined based on the user experience change trends reflected in historical interaction feedback parameters.

6. The multi-dimensional interactive quality perception and resource optimization scheduling method for holographic communication systems according to claim 1, characterized in that, Step S3 further includes: A fuzzy control rule set is designed based on the key factors affecting the interaction quality in holographic communication systems. , The key factors include at least network transmission status parameters, interaction performance parameters, and task context parameters; each rule Includes Each of the following prerequisites corresponds to at least one of the aforementioned key factors, and is used to describe the state range or level of that factor at the current moment. A task context parameter vector is constructed based on the current environment and task context. This is used as an additional input to the preconditions of fuzzy rules to characterize the changes in rule activation intensity under different interaction scenarios. This indicates the number of environmental and task context parameters; At any moment , from the enhanced input vector and task context parameter vector Obtain fuzzy rules Dynamic membership degree: ； in, It is the first The number of prerequisites for each rule. For fuzzy rules The Middle The prerequisite at time The corresponding input variables are taken from the multidimensional index vector. Or environment and task context parameter vector These are used to characterize the network state, interaction performance, or interaction context features related to the precondition, respectively. For rules The Middle The fuzzy membership function corresponding to each precondition is used to classify the input variables. The mapping is to membership values; the fuzzy membership function adopts a predefined continuous monotonic function form. Each indicator is obtained by weighted fusion of all fuzzy rules. Dynamic weights: ； Then time The overall scores for the interaction quality at time t+1 are as follows: ； ； in, The membership function represents the knowledge context, reflecting the relative weights of different knowledge sources during the prediction stage. This is the weight adjustment coefficient for the indicator under knowledge constraints; Indicates the first The standard utility function of each indicator.

7. The multi-dimensional interactive quality perception and resource optimization scheduling method for holographic communication systems according to claim 1, characterized in that, Step S4 further includes: Based on time The overall interaction quality score at time t+1 is used to determine the trend of interaction quality changes. An optimization objective function set is constructed based on the changes in the contribution items of each dimension. Calculate each indicator separately Changes in contribution: ； Among them, when At that time, the judgment of the first Each indicator dimension has a negative impact on the overall interaction quality score at the prediction time, and is based on... The magnitude of this indicator measures its impact on interaction quality fluctuations; construct a set of multi-objective optimization objective functions for the current scheduling period: ； in, The system resource allocation decision variables are determined by the performance targets corresponding to the indicator dimensions with significant negative impacts, and are optimized under the constraint that the comprehensive interaction quality score is not lower than a preset threshold. In the multi-objective optimization process, the Pareto optimization method is used to search for the trade-off optimal solution among the objective functions, and the Pareto dominance relationship is defined as follows: For any two feasible solutions and If satisfied ， ； Then it is called a solution Dominant Solution A solution that is not dominated by any other solution is defined as a Pareto optimal solution. All Pareto optimal solutions constitute the Pareto optimal solution set, which serves as the candidate scheme set for subsequent resource allocation and scheduling.

8. The multi-dimensional interactive quality perception and resource optimization scheduling method for holographic communication systems according to claim 1, characterized in that, Step S5 further includes: Based on the optimal performance configuration scheme found, and according to the constraints of the available resources of the system and the change range of resource scheduling actions, several candidate resource configuration schemes are generated, and each candidate resource configuration scheme corresponds to a set of specific resource scheduling actions. For each candidate resource scheduling action Predict its impact on the system state at the next moment and obtain the corresponding prediction index vector. And calculate the overall interaction quality score for the next time step. and the values ​​of each objective function in the predicted state. ; The comprehensive impact of all candidate resource scheduling actions on interaction quality, network performance, and resource consumption is quantitatively evaluated, and the candidate resource configuration scheme with the highest comprehensive interaction quality score is selected as the optimal resource configuration scheme for the next time step.

Images