Adaptive adjustment method for active matrix organic light emitting display

Abstract

The present application relates to the technical field of organic light emitting display data processing, and particularly relates to an adaptive adjustment method of active matrix organic light emitting display, wherein distributed light-sensitive sensor nodes are deployed to collect ambient light intensity and color temperature data, current stress parameter fluctuations captured by a programmable current sampling module and thermal distribution data of a temperature sensor network are combined, input into a dynamic aging model micro-service cluster for parameter coupling calculation, and time-varying degradation characteristic parameters are output; a containerized simulation micro-service constructs a virtual aging scene based on the parameters and generates a compensation coefficient matrix; a multi-objective loss function solver combines a delta E color difference threshold and power consumption constraints to optimize display driving parameters, the present application solves the compensation deviation problem caused by the coupling effect of current stress, temperature bias and material relaxation that cannot be adapted by a traditional static model, realizes dynamic balance of display brightness, color gamut and power consumption, and prolongs the service life of a display panel.

Description

Technical Field

[0001] This invention relates to the field of organic light-emitting display data processing technology, and more particularly to an adaptive adjustment method for active matrix organic light-emitting displays. Background Technology

[0002] The adaptive adjustment of active-matrix organic light-emitting diode (OLED) displays is based on ambient light sensors and a real-time feedback system. By detecting external light intensity and color temperature parameters, and combining this with a temperature compensation circuit, the system dynamically adjusts the pixel drive current and sub-pixel emission duration to maintain visual consistency in screen brightness and color temperature. In low ambient light conditions, the system reduces the drive voltage according to the gamma correction curve, while simultaneously optimizing the carrier injection efficiency of the organic light-emitting layer to balance power consumption and display accuracy. When high dynamic range content is detected, a pixel-level compensation algorithm synchronously corrects the gate voltage of the thin-film transistor (TFT) to compensate for brightness decay caused by organic material aging, and reduces ghosting effects in high-contrast scenes through dynamic refresh rate adjustment. This adjustment mechanism integrates photoelectric characteristic modeling and nonlinear optimization algorithms, effectively extending the lifespan of the display while maintaining color gamut coverage and ΔE color difference.

[0003] In the prior art, pixel-level compensation algorithms based on preset aging models are difficult to accurately adapt to the time-varying characteristics of the degradation rate of organic light-emitting units. Their static parameter mapping relationship cannot reflect the nonlinear aging trajectory under the combined effect of driving current stress, temperature bias and intrinsic material relaxation effect in real time, resulting in a cumulative deviation between the aging compensation coefficient and the actual pixel decay. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an adaptive adjustment method for active matrix organic light-emitting displays. This method solves the problem that the static parameter mapping relationship based on a preset aging model cannot adapt in real time to the nonlinear time-varying degradation characteristics of organic light-emitting units under the coupling effects of driving current stress, temperature bias, and material relaxation effects, resulting in a cumulative deviation between the aging compensation coefficient and the actual attenuation.

[0005] To solve the above-mentioned technical problems, the specific technical solution of the present invention is as follows:

[0006] The adaptive adjustment method for active matrix organic light-emitting displays provided by this invention includes:

[0007] Ambient light intensity and color temperature data are collected by deploying distributed photosensitive sensor nodes, and the ambient light intensity and color temperature data are transmitted to the dynamic aging model microservice cluster through the central data bus.

[0008] The fluctuation of current stress parameters in the drive circuit is captured by a programmable current sampling module, and the heat distribution data collected by the temperature sensor network is obtained. The fluctuation of current stress parameters and heat distribution data are input into the dynamic aging model microservice cluster for parameter coupling calculation, and the time-varying degradation characteristic parameters are output.

[0009] Based on the time-varying degradation characteristic parameters output by the microservice cluster of the dynamic aging model, a virtual aging scenario is constructed through containerized simulation microservices to generate a compensation coefficient matrix.

[0010] The compensation coefficient matrix is ​​input into a multi-objective loss function solver, and the parameters are optimized by combining the preset ΔE color difference threshold and power consumption constraints, and the optimized display driving parameters are output.

[0011] The optimized display driving parameters are adjusted in real time through an edge and cloud collaborative architecture, and the compensation coefficient matrix is ​​synchronized to the online learning microservice. Incremental model updates are triggered based on the optimized display driving parameters and the compensation coefficient matrix.

[0012] Furthermore, in the adaptive adjustment method for active matrix organic light-emitting displays described in this invention, the data acquisition process of the distributed photosensitive sensor nodes includes:

[0013] Each sensor node independently collects ambient light parameters through a microservice architecture and uploads the ambient light parameters to the central data bus via a lightweight API protocol;

[0014] After receiving multi-source data including ambient light parameters, current stress parameter fluctuations and temperature distribution data, the central data bus calls the edge computing unit to perform abnormal pulse preprocessing on the current stress parameter fluctuations and generate filtered current stress time series data.

[0015] The temperature sensor network is based on a topology-aware dynamic orchestration strategy. It establishes a heat conduction path model based on preprocessed current stress time series data and corrects the spatial resolution deviation of the heat distribution data.

[0016] Furthermore, in the adaptive adjustment method for active matrix organic light-emitting displays described in this invention, the construction of the dynamic aging model microservice cluster includes:

[0017] The current stress accumulation calculation service, the temperature bias nonlinear function solver, and the material relaxation coefficient matrix generator are separated into independent microservices.

[0018] The data pipeline coupling between the current stress accumulation calculation service and the material relaxation coefficient matrix generator is achieved through an asynchronous message queue, wherein the filtered current stress time series data triggers the real-time correction of the material relaxation coefficient matrix.

[0019] The material relaxation coefficient matrix generator calls the pre-stored current density and mobility decay curves in the experimental database via RPC, and combines them with the corrected thermal distribution data to quantify the mobility decay error.

[0020] Furthermore, in the adaptive adjustment method for active matrix organic light-emitting displays described in this invention, the operation process of the containerized simulation microservice includes:

[0021] The LSTM network is deployed as a containerized inference microservice. After receiving the filtered current stress time series data stream, online weight fine-tuning is performed to generate an adaptive time series correlation prediction model.

[0022] An incremental learning algorithm is used to inject organic material response delay data into the training set of the adaptive time-series correlation prediction model to update the weight parameters of the prediction model.

[0023] The compensation strategy is computed in parallel based on the adaptive temporal correlation prediction model through a distributed simulation microservice, and the spatiotemporal correlation of the compensation parameters is maintained by a consistent hashing algorithm.

[0024] Furthermore, in the adaptive adjustment method for active matrix organic light-emitting displays described in this invention, the correction process of the material relaxation coefficient matrix includes:

[0025] The temperature bias parameters of the intrinsic relaxation effect of the material are updated based on the corrected heat distribution data output by the temperature sensor network.

[0026] The current density and mobility decay curves pre-stored in the experimental database are called, and the output results of the current stress accumulation calculation service are combined to generate the mobility decay compensation gradient.

[0027] The mobility decay compensation gradient is nonlinearly coupled with the current stress accumulation calculation result, and the corrected relaxation coefficient matrix is ​​output to the dynamic aging model microservice cluster.

[0028] Furthermore, in the adaptive adjustment method for active matrix organic light-emitting displays described in this invention, the process of constructing the virtual aging scene includes:

[0029] Based on the time-varying degradation parameters output by the microservice cluster of the dynamic aging model, a pixel-level aging compensation initial strategy is generated.

[0030] The pixel-level aging compensation initial strategy is divided into three parallel microservice tasks: gamma correction weight calculation, sub-pixel emissivity duration optimization, and TFT gate voltage adjustment gradient.

[0031] The three parallel microservice tasks are distributed to distributed computing nodes for collaborative execution via a service mesh, and the simulated state data of the virtual aging scenario is recorded using transaction logs.

[0032] Furthermore, in the adaptive adjustment method for active matrix organic light emission displays described in this invention, the optimization process of the multi-objective loss function solver includes:

[0033] Receive preset ΔE chromatic difference threshold, ghosting suppression index and power consumption constraints, and construct Pareto front analysis model;

[0034] By backpropagating the bidirectional communication channel between the microservice and the dynamic aging model microservice cluster, the coupling weight parameters of the Pareto front analysis model and the dynamic aging model are updated synchronously.

[0035] Differential privacy technology is used to de-identify the gradient sharing data output by the backpropagation microservice.

[0036] Furthermore, the adaptive adjustment method for active matrix organic light-emitting displays described in this invention includes the following execution process of the edge-cloud collaborative architecture:

[0037] Deploy a display driver chip register mapping microservice at the edge to execute the optimized display driver parameter adjustment instructions output by the multi-objective loss function solver in real time;

[0038] The version snapshot of the compensation coefficient matrix is ​​synchronized to the cloud historical database to establish a timestamp-based aging parameter version control link;

[0039] The system obtains actual display panel brightness decay feedback data through a subscription and publish model, triggering incremental model update operations for the online learning microservice.

[0040] Furthermore, in the adaptive adjustment method for active matrix organic light-emitting displays described in this invention, the incremental model update process includes:

[0041] After receiving the brightness attenuation feedback data, the online learning microservice generates a historical database update instruction and attaches the version identifier of the compensation coefficient matrix;

[0042] The update command is routed to the dynamic aging model microservice cluster via the service mesh, triggering parameter recalibration of the current stress accumulation calculation service and the material relaxation coefficient matrix generator.

[0043] The federated learning framework is used to aggregate the brightness decay feedback data and historical aging parameters across devices, and to update the weight coefficients of the global dynamic aging model.

[0044] Furthermore, the adaptive adjustment method for active matrix organic light-emitting displays according to the present invention further includes:

[0045] The distributed sensor microservice and edge preprocessing unit reduce the computational load on the central node through asynchronous data processing, and support the dynamic expansion of ultra-high resolution screens.

[0046] The containerized simulation microservice dynamically allocates computing resources based on an elastic resource scheduling strategy, thereby accelerating the verification efficiency of the compensation strategy for the virtual aging scenario.

[0047] The incremental learning and federated learning microservices synchronously optimize the time-varying adaptive parameters and cross-device generalization parameters of the dynamic aging model through a closed-loop data link.

[0048] The service mesh and version control middleware maintain state consistency among microservices based on simulated state data from transaction logs, preventing display parameter jumps caused by local microservice updates.

[0049] Beneficial effects of this invention;

[0050] This invention significantly improves the adaptive adjustment accuracy and efficiency of organic light-emitting display systems through the synergistic effect of a dynamic aging model microservice cluster and containerized simulation microservices. Distributed photosensitive sensors and multi-source data fusion technology capture ambient light, current stress, and thermal distribution parameters in real time. Combined with asynchronous message queues, dynamic parameter coupling between microservices is achieved, accurately quantifying the nonlinear correlation between material relaxation effects and temperature bias. An adaptive temporal prediction model based on LSTM networks and an incremental learning mechanism, through a spatiotemporal consistency generation strategy for the compensation coefficient matrix, effectively eliminates compensation biases caused by the mismatch between current stress accumulation and material aging rate in traditional static models. A federated learning framework under an edge-cloud collaborative architecture, combined with differential privacy technology, achieves privacy-preserving aggregation of aging data across devices, simultaneously optimizing the time-varying adaptability and generalization ability of the dynamic aging model. Service mesh and version control middleware maintain microservice state consistency through transaction logs, and a canary release strategy suppresses parameter jump risks, ultimately achieving a dynamic balance between display brightness, color gamut coverage, and power consumption, extending the lifespan of the display panel. Attached Figure Description

[0051] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on the drawings without creative effort.

[0052] Figure 1 A flowchart of an adaptive adjustment method for an active matrix organic light-emitting display provided in an embodiment of the present invention. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention. The technical solutions provided by various embodiments of this invention will be described in detail below with reference to the accompanying drawings. To better understand the objectives of this invention, it will be described in further detail below.

[0054] Please see Figure 1 The adaptive adjustment method for active matrix organic light-emitting displays provided by the present invention includes:

[0055] Step S101: Collect ambient light intensity and color temperature data by deploying distributed photosensitive sensor nodes, and transmit the ambient light intensity and color temperature data to the dynamic aging model microservice cluster through the central data bus;

[0056] Step S102: The programmable current sampling module is used to capture the fluctuation of current stress parameters in the driving circuit and to obtain the heat distribution data collected by the temperature sensor network. The fluctuation of current stress parameters and the heat distribution data are input to the dynamic aging model microservice cluster for parameter coupling calculation and output time-varying degradation characteristic parameters.

[0057] Step S103: Based on the time-varying degradation characteristic parameters output by the dynamic aging model microservice cluster, a virtual aging scenario is constructed through containerized simulation microservices to generate a compensation coefficient matrix.

[0058] Step S104: Input the compensation coefficient matrix into the multi-objective loss function solver, combine the preset ΔE color difference threshold and power consumption constraints to optimize the parameters, and output the optimized display driving parameters.

[0059] Step S105: Real-time adjustment of the optimized display driving parameters is performed through the edge and cloud collaborative architecture, and the compensation coefficient matrix is ​​synchronized to the online learning microservice. Incremental model update is triggered based on the optimized display driving parameters and the compensation coefficient matrix.

[0060] The adaptive adjustment method for active matrix organic light-emitting displays provided by this invention achieves dynamic optimization of display parameters through multi-level data processing and a microservice architecture. First, distributed photosensitive sensor nodes are deployed. Each node independently collects ambient light intensity and color temperature data based on the microservice architecture and transmits the raw data to a central data bus via a lightweight API protocol. The central data bus integrates an edge computing unit to perform abnormal pulse filtering on the fluctuations in the current stress parameters of the driving circuit captured by the current sampling module. Simultaneously, it receives thermal distribution data generated by a temperature sensor network based on a topology-aware dynamic orchestration strategy, forming a multi-source heterogeneous sensor data stream.

[0061] The processed current stress parameter fluctuations and thermal distribution data are input into a dynamic aging model microservice cluster. This microservice cluster consists of a current stress accumulation calculation service, a temperature bias nonlinear function solver, and a material relaxation coefficient matrix generator. Data pipeline coupling between microservices is achieved through an asynchronous message queue. The filtered current stress time-series data triggers real-time correction of the material relaxation coefficient matrix. Combined with pre-stored current density-mobility decay curves in the experimental database, the mobility decay error is quantified, and time-varying degradation characteristic parameters are output. These time-varying degradation characteristic parameters characterize the nonlinear aging trajectory of the organic light-emitting unit under the combined effects of current stress, temperature bias, and material relaxation effects.

[0062] Based on the aforementioned time-varying degradation characteristic parameters, a containerized simulation microservice constructs a virtual aging scenario. This simulation microservice deploys an LSTM network as a containerized inference unit, receives time-series current data streams, performs online weight fine-tuning, and generates an adaptive time-series correlation prediction model. Distributed computing nodes execute gamma correction weight calculation, sub-pixel emission duration optimization, and TFT gate voltage adjustment gradient solving tasks in parallel. A consistent hashing algorithm is used to maintain the spatiotemporal correlation of the compensation parameters, and transaction logs are used to record the simulated state data of the virtual aging scenario, forming a compensation coefficient matrix.

[0063] The compensation coefficient matrix is ​​input into a multi-objective loss function solver, and a Pareto front analysis model is constructed by combining a preset ΔE color difference threshold and power consumption constraints. The coupling weight parameters are synchronously updated through a bidirectional communication channel between the backpropagation microservice and the dynamic aging model microservice cluster. Differential privacy technology is used to desensitize the gradient sharing data, outputting optimized display driving parameters that meet the requirements of color gamut coverage, ghosting suppression, and power consumption balance.

[0064] A collaborative edge-cloud architecture enables real-time adjustment of display driver parameters. At the edge, a microservice for mapping display driver chip registers directly controls current and refresh rate, while simultaneously synchronizing a version snapshot of the compensation coefficient matrix to a historical database in the cloud. An online learning microservice, based on a publish-subscribe model, acquires brightness decay feedback data from actual display panels, appends a compensation coefficient matrix version identifier, and triggers incremental model updates. A federated learning framework aggregates cross-device aging data, and a service mesh routing mechanism updates global dynamic aging model parameters, forming a closed loop for display parameter optimization. All functional modules maintain state consistency through the service mesh and version control middleware, preventing display anomalies caused by partial updates and achieving adaptive adjustment capabilities throughout the display system's lifecycle.

[0065] Specifically, the adaptive adjustment method for active matrix organic light-emitting displays described in this invention includes the following data acquisition process for the distributed photosensitive sensor nodes:

[0066] Each sensor node independently collects ambient light parameters through a microservice architecture and uploads the ambient light parameters to the central data bus via a lightweight API protocol;

[0067] After receiving multi-source data including ambient light parameters, current stress parameter fluctuations and temperature distribution data, the central data bus calls the edge computing unit to perform abnormal pulse preprocessing on the current stress parameter fluctuations and generate filtered current stress time series data.

[0068] The temperature sensor network is based on a topology-aware dynamic orchestration strategy. It establishes a heat conduction path model based on preprocessed current stress time series data and corrects the spatial resolution deviation of the heat distribution data.

[0069] The data acquisition process of distributed photosensitive sensor nodes achieves high-precision synchronous acquisition and preprocessing of environmental parameters through a multi-level collaborative architecture. Each sensor node is deployed using containerized microservices, independently running an ambient light parameter acquisition program. It converts the light intensity signal into a digital quantity using a preset photoelectric conversion calibration coefficient and encapsulates it into a standardized data packet. This data packet is transmitted to the central data bus via a lightweight API protocol. A message queue mechanism is used to control data flow during concurrent uploads from multiple nodes, avoiding transmission delays caused by bus congestion.

[0070] After receiving a multi-source heterogeneous data stream containing ambient light parameters, current stress parameter fluctuations, and temperature distribution data, the central data bus triggers the preprocessing process of the edge computing unit. The edge computing unit deploys a digital filtering algorithm to detect abnormal pulses in the current stress parameter fluctuations, and uses a sliding window mean filter to eliminate high-frequency noise interference, generating filtered current stress time-series data. This time-series data, along with the raw thermal distribution data uploaded from the temperature sensor network, is input to the data fusion module, where a timestamp alignment algorithm matches the acquisition timing of the multi-source data.

[0071] The temperature sensor network reconstructs the thermal distribution data processing link based on a topology-aware dynamic orchestration strategy, and establishes a heat conduction path model based on the filtered current stress time-series data. This model analyzes the spatial correlation between current stress parameters and temperature distribution to identify the heat conduction path offset of the organic light-emitting display panel, and uses an interpolation algorithm to correct the spatial resolution deviation caused by the sparse layout of sensor nodes. The corrected thermal distribution data and the filtered current stress data together serve as input parameters for the dynamic aging model microservice cluster, forming a closed-loop logical link between sensor data acquisition, preprocessing, and model input.

[0072] Specifically, the adaptive adjustment method for active matrix organic light emission displays described in this invention includes the construction of a dynamic aging model microservice cluster comprising:

[0073] The current stress accumulation calculation service, the temperature bias nonlinear function solver, and the material relaxation coefficient matrix generator are separated into independent microservices.

[0074] The data pipeline coupling between the current stress accumulation calculation service and the material relaxation coefficient matrix generator is achieved through an asynchronous message queue, wherein the filtered current stress time series data triggers the real-time correction of the material relaxation coefficient matrix.

[0075] The material relaxation coefficient matrix generator calls the pre-stored current density and mobility decay curves in the experimental database via RPC, and combines them with the corrected thermal distribution data to quantify the mobility decay error.

[0076] The construction of the dynamic aging model microservice cluster achieves parameter coupling optimization by decoupling the core computing module from the distributed service collaboration. The current stress accumulation calculation service is deployed as an independent microservice, which calculates the cumulative amount of filtered current stress time-series data based on a time integration algorithm, outputting a current stress accumulation distribution map. The temperature bias nonlinear function solver microservice receives the corrected thermal distribution data and uses a polynomial fitting algorithm to construct the mapping relationship between the temperature gradient field and the current stress distribution, generating a temperature compensation coefficient matrix.

[0077] A data pipeline is established between the current stress accumulation calculation service and the material relaxation coefficient matrix generator through an asynchronous message queue. When the filtered current stress time-series data reaches a preset time window threshold, the message queue is triggered to push a data update event to the material relaxation coefficient matrix generator. The material relaxation coefficient matrix generator initiates a remote procedure call protocol, accesses the pre-stored current density and mobility decay curve dataset in the experimental database, and calculates the time-varying decay rate of the organic material's carrier mobility by combining the gradient parameters of the current temperature compensation coefficient matrix.

[0078] The mobility decay rate and the cumulative current stress distribution spectrum are nonlinearly coupled, and the spatial distribution characteristics of the material relaxation effect are solved using a tensor decomposition algorithm to generate a dynamically updated relaxation coefficient matrix. This matrix is ​​synchronized to the temperature bias nonlinear function solver microservice through the service mesh, forming a cross-feedback mechanism of current-temperature-material properties to achieve closed-loop iterative optimization of aging model parameters. The data dependencies between microservices are maintained through version control middleware to ensure temporal consistency during parameter updates.

[0079] Specifically, the adaptive adjustment method for active matrix organic light emission displays described in this invention includes the following process for running the containerized simulation microservice:

[0080] The LSTM network is deployed as a containerized inference microservice. After receiving the filtered current stress time series data stream, online weight fine-tuning is performed to generate an adaptive time series correlation prediction model.

[0081] An incremental learning algorithm is used to inject organic material response delay data into the training set of the adaptive time-series correlation prediction model to update the weight parameters of the prediction model.

[0082] The compensation strategy is computed in parallel based on the adaptive temporal correlation prediction model through a distributed simulation microservice, and the spatiotemporal correlation of the compensation parameters is maintained by a consistent hashing algorithm.

[0083] The containerized simulation microservice operates by dynamically generating compensation strategies through time-series modeling and distributed computing technologies. An LSTM network is encapsulated as a containerized inference microservice, leveraging the Kubernetes orchestration platform for elastic scaling. Upon receiving the filtered current-stress time-series data stream, the online inference service is initiated. The LSTM network employs a sliding time window partitioning strategy to segment the input data, and an adaptive learning rate adjustment algorithm fine-tunes the network weights, generating an adaptive time-series correlation prediction model that reflects the time-varying degradation characteristics of organic materials.

[0084] During the operation of the adaptive time-series correlation prediction model, the incremental learning algorithm captures the response delay data of organic materials in real time and performs time series alignment and normalization processing through a data buffer queue. The preprocessed delay data is injected into the model training set, and the backpropagation algorithm is used to update the weight parameters of the hidden layer of the LSTM network, enabling the prediction model to dynamically adapt to the changing trend of the material aging rate. The updated weight parameters generate a difference snapshot through version control middleware, triggering parameter synchronization events in the distributed simulation microservice.

[0085] Based on the updated adaptive temporal correlation prediction model, the distributed simulation microservice decomposes the compensation strategy calculation task into three parallel subtasks: gamma correction weights, sub-pixel emission duration, and TFT gate voltage adjustment. A virtual node ring is constructed using a consistent hashing algorithm. Subtasks are mapped to specific computing nodes based on the spatiotemporal labels of the compensation parameters, and the spatiotemporal correlation of the compensation parameters is maintained through the redundant replication mechanism of the hash ring. The local compensation parameters generated by each node are aggregated and reconstructed through a service mesh, outputting a global compensation coefficient matrix that matches the physical layout of the display panel.

[0086] Specifically, the adaptive adjustment method for active matrix organic light-emitting displays according to the present invention includes the following process for correcting the material relaxation coefficient matrix:

[0087] The temperature bias parameters of the intrinsic relaxation effect of the material are updated based on the corrected heat distribution data output by the temperature sensor network.

[0088] The current density and mobility decay curves pre-stored in the experimental database are called, and the output results of the current stress accumulation calculation service are combined to generate the mobility decay compensation gradient.

[0089] The mobility decay compensation gradient is nonlinearly coupled with the current stress accumulation calculation result, and the corrected relaxation coefficient matrix is ​​output to the dynamic aging model microservice cluster.

[0090] The correction process for the material relaxation coefficient matrix achieves dynamic calibration of material properties through multi-source data fusion and nonlinear modeling. The corrected thermal distribution data output from the temperature sensor network is input to the material intrinsic relaxation effect analysis module. A polynomial interpolation algorithm is used to reconstruct the global temperature field of the display panel, and the temperature bias parameters are updated based on the thermodynamic relaxation equation. These temperature bias parameters are matched with activation energy thresholds in the organic material band structure database to generate temperature-dependent carrier mobility correction coefficients.

[0091] When accessing pre-stored current density and mobility decay curves in the experimental database, a remote data channel is established with the material property analysis microservice via the RPC protocol. A piecewise linear interpolation algorithm is used to map the current stress distribution map output by the current stress accumulation calculation service to the mobility decay curve. A three-dimensional gradient vector field is constructed using temperature bias parameters to generate a mobility decay compensation gradient. This compensation gradient characterizes the nonlinear decay rate of the organic material's mobility under the coupled effects of current stress and temperature.

[0092] The mobility decay compensation gradient and the current stress accumulation calculation results are subjected to tensor decomposition, and the coupling interference between different physical dimensions is eliminated by orthogonal projection algorithm. Nonlinear least squares method is used to optimize the weights of the decomposed eigenvectors, outputting a corrected relaxation coefficient matrix. This matrix is ​​synchronized to the asynchronous message queue of the dynamic aging model microservice cluster via the service mesh, triggering a parameter recalibration event in the current stress accumulation calculation service, forming a cross-iterative optimization loop of temperature-current stress-material properties. Version control middleware records the updated version identifier of the relaxation coefficient matrix, providing data backtracking for incremental model updates.

[0093] Specifically, the adaptive adjustment method for active matrix organic light-emitting displays described in this invention includes the following process for constructing the virtual aging scene:

[0094] Based on the time-varying degradation parameters output by the microservice cluster of the dynamic aging model, a pixel-level aging compensation initial strategy is generated.

[0095] The pixel-level aging compensation initial strategy is divided into three parallel microservice tasks: gamma correction weight calculation, sub-pixel emissivity duration optimization, and TFT gate voltage adjustment gradient.

[0096] The three parallel microservice tasks are distributed to distributed computing nodes for collaborative execution via a service mesh, and the simulated state data of the virtual aging scenario is recorded using transaction logs.

[0097] The construction process of the virtual aging scenario achieves efficient verification of the compensation strategy through strategy decomposition and distributed collaborative computing. The time-varying degradation parameters output by the dynamic aging model microservice cluster are input to the compensation strategy generation module. Based on the pixel-level current stress distribution map and mobility decay gradient parameters, a weighted fusion algorithm is used to generate an initial aging compensation strategy. The initial strategy includes a set of compensation parameters in three dimensions: gamma voltage correction, sub-pixel emission time baseline, and TFT gate voltage offset.

[0098] When the initial aging compensation strategy is decomposed into parallel microservice tasks, the tasks are divided according to the data type of the compensation parameters and the characteristics of the computational load. The gamma correction weight calculation task calls the gamma lookup table data from the display panel and, combined with the mobility decay gradient in the time-varying degradation parameters, uses an interpolation algorithm to generate voltage compensation coefficients. The sub-pixel emission duration optimization task calculates the emission time baseline adjustment amount based on organic material response delay data and current stress accumulation distribution using a pulse width modulation model. The TFT gate voltage adjustment gradient solving task receives temperature bias parameters and mobility decay error data and uses a gradient descent algorithm to iteratively solve for the gate voltage compensation amount.

[0099] Three parallel microservice tasks are distributed to distributed computing nodes using a service mesh load balancing strategy, and a consistent hashing algorithm is used to maintain the mapping relationship between task allocation and physical display areas. When each node performs compensation parameter calculations, intermediate calculation results are synchronized through a message middleware, and transaction logs are used to record the version identifier, timestamp, and node status data of the compensation coefficients. This transaction log data is input to the status traceability module of the virtual aging scenario, providing historical data benchmarks for iterative optimization of the compensation strategy, forming a closed-loop logical link for compensation parameter generation, verification, and optimization.

[0100] Specifically, the adaptive adjustment method for active matrix organic light emission display described in this invention includes the following optimization process for the multi-objective loss function solver:

[0101] Receive preset ΔE chromatic difference threshold, ghosting suppression index and power consumption constraints, and construct Pareto front analysis model;

[0102] By backpropagating the bidirectional communication channel between the microservice and the dynamic aging model microservice cluster, the coupling weight parameters of the Pareto front analysis model and the dynamic aging model are updated synchronously.

[0103] Differential privacy technology is used to de-identify the gradient sharing data output by the backpropagation microservice.

[0104] The optimization process of the multi-objective loss function solver achieves global optimal solution search for display parameters through multi-constraint fusion and privacy protection mechanisms. After receiving preset ΔE color difference thresholds, ghosting suppression indices, and power consumption constraints, a multi-objective normalization algorithm is used to map constraint parameters of different dimensions to a unified evaluation space, constructing a Pareto front analysis model. The model generates a candidate solution set using a non-dominated sorting algorithm and calculates the color gamut coverage, ghosting index, and power consumption level evaluation values ​​corresponding to each solution based on the time-varying degradation parameters output by the dynamic aging model microservice cluster.

[0105] After establishing a bidirectional communication channel between the backpropagation microservice and the dynamic aging model microservice cluster, the weight update gradient of the Pareto front analysis model is transmitted to the dynamic aging model through a gradient sharing protocol. The temperature bias nonlinear function solver in the dynamic aging model microservice cluster receives the gradient data, adjusts the coupling weight parameters of the temperature compensation coefficient matrix using an adaptive momentum optimization algorithm, and feeds back the synchronously updated weight parameters to the Pareto front analysis model through version control middleware, forming a cross-model parameter collaborative update mechanism.

[0106] When performing privacy protection processing on the gradient sharing data output by the backpropagation microservice, differential privacy technology is used to inject Gaussian noise perturbation. The noise injection intensity is controlled by a privacy budget allocation algorithm, locally perturbing the feature dimensions of the gradient tensor to generate the anonymized gradient update. This anonymized gradient is transmitted to the dynamic aging model microservice cluster through a secure channel in the service mesh, maintaining model optimization accuracy while blocking sensitive data leakage paths. Noise parameters and perturbation strategies during the privacy processing are dynamically configured through independent microservices, forming a traceable privacy protection operation log.

[0107] Specifically, the adaptive adjustment method for active matrix organic light-emitting displays described in this invention includes the following execution process of the edge-cloud collaborative architecture:

[0108] Deploy a display driver chip register mapping microservice at the edge to execute the optimized display driver parameter adjustment instructions output by the multi-objective loss function solver in real time;

[0109] The version snapshot of the compensation coefficient matrix is ​​synchronized to the cloud historical database to establish a timestamp-based aging parameter version control link;

[0110] The system obtains actual display panel brightness decay feedback data through a subscription and publish model, triggering incremental model update operations for the online learning microservice.

[0111] The edge-cloud collaborative architecture achieves dynamic optimization of the display system through layered computing and data closure. A display driver chip register mapping microservice is deployed at the edge, directly interacting with the optimized display driver parameters output by the multi-objective loss function solver. These parameters include gamma-corrected voltage values, sub-pixel emission duration references, and TFT gate voltage offsets. The microservice converts these parameters into a binary instruction set recognizable by the driver chip registers through a hardware abstraction layer interface, updating the display panel's current control module and refresh rate control unit in real time.

[0112] When synchronizing the version snapshot of the compensation coefficient matrix to the cloud historical database, an event-triggered timestamp generation mechanism is used. After each compensation strategy update, a version identifier and generation time node of the compensation matrix are appended. The historical database constructs a version control link, recording the parameter changes of adjacent version compensation matrices through a difference comparison algorithm, forming a traceable aging parameter evolution map. The data structure of the version control link adopts a key-value storage model, using timestamps as indexes to establish a mapping relationship between compensation parameters and the simulated state of the virtual aging scenario.

[0113] When obtaining brightness decay feedback data of the actual display panel through a publish-subscribe model, a photoelectric sensor array is embedded in the display driver circuit to periodically collect the brightness decay of sub-pixels and encapsulate it into a standardized message. This message is broadcast to the subscription queue of the online learning microservice via a message middleware. Before triggering the incremental model update operation, data integrity verification and time window alignment are performed. After the update operation is initiated, the service mesh routes to the corresponding dynamic aging model microservice node based on the version identifier of the compensation coefficient matrix, completing the closed-loop iterative optimization of aging parameters and compensation strategies.

[0114] Specifically, the adaptive adjustment method for active matrix organic light-emitting displays described in this invention includes an incremental model update process comprising:

[0115] After receiving the brightness attenuation feedback data, the online learning microservice generates a historical database update instruction and attaches the version identifier of the compensation coefficient matrix;

[0116] The update command is routed to the dynamic aging model microservice cluster via the service mesh, triggering parameter recalibration of the current stress accumulation calculation service and the material relaxation coefficient matrix generator.

[0117] The federated learning framework is used to aggregate the brightness decay feedback data and historical aging parameters across devices, and to update the weight coefficients of the global dynamic aging model.

[0118] The incremental model update process achieves continuous optimization of the aging model through data closure and distributed learning. After receiving brightness decay feedback data collected by the display panel's photoelectric sensor array, the online learning microservice performs a data preprocessing procedure, including outlier filtering and time window alignment, to generate a standardized decay feature vector. This feature vector is appended with a version identifier of the compensation coefficient matrix and pushed to the historical database update instruction generation module via a message queue. The version identifier includes a compensation strategy generation timestamp and a hash checksum.

[0119] When routing update commands to the dynamic aging model microservice cluster via the service mesh, a content addressing routing strategy based on version identifiers is employed to accurately match the computation node corresponding to the currently effective compensation coefficient matrix version. When triggering the parameter recalibration process of the current stress accumulation calculation service, the time integral weighting coefficient of the current stress accumulation calculation is adjusted using a sliding window algorithm, taking into account the spatial distribution information in the brightness attenuation feature vector. Simultaneously, the recalibration operation of the material relaxation coefficient matrix generator is initiated, and the interpolation reference point of the mobility attenuation curve is corrected based on the difference between the attenuation feature vector and historical compensation parameters.

[0120] When using a federated learning framework for cross-device data aggregation, a local differential privacy module is deployed at the edge to inject noise and desensitize features in the brightness decay feedback data and historical aging parameters. The processed data is uploaded to the federated learning server via a secure multi-party computation protocol. A weighted average algorithm is used to fuse the feature parameter distributions of multiple devices, generating a weight update gradient for the global dynamic aging model. This gradient is distributed to the dynamic aging model microservice cluster on each device via a service mesh. Combined with the parameter evolution graph recorded by the version control middleware, incremental iterative updates of the aging model parameters are completed, forming a dynamic adaptation closed loop between display panel degradation characteristics and compensation strategies.

[0121] Specifically, the adaptive adjustment method for active matrix organic light-emitting display of the present invention further includes:

[0122] The distributed sensor microservice and edge preprocessing unit reduce the computational load on the central node through asynchronous data processing, and support the dynamic expansion of ultra-high resolution screens.

[0123] The containerized simulation microservice dynamically allocates computing resources based on an elastic resource scheduling strategy, thereby accelerating the verification efficiency of the compensation strategy for the virtual aging scenario.

[0124] The incremental learning and federated learning microservices synchronously optimize the time-varying adaptive parameters and cross-device generalization parameters of the dynamic aging model through a closed-loop data link.

[0125] The service mesh and version control middleware maintain state consistency among microservices based on simulated state data from transaction logs, preventing display parameter jumps caused by local microservice updates.

[0126] The distributed sensor microservices and edge preprocessing unit employ an asynchronous message queue mechanism to dynamically allocate computational load. Each sensor node fragments the raw data it collects and pushes it to the local message queue of the edge computing node. The edge preprocessing unit deploys a streaming processing engine to perform parallel filtering and feature extraction operations on the data fragments, generating standardized preprocessed data packets. A distributed caching mechanism temporarily stores the preprocessing results, uploading only key feature data to the central node. This reduces the transmission overhead of processing massive amounts of data on high-resolution screens and supports linear scaling of computational resources when screen resolution dynamically expands.

[0127] The containerized simulation microservice dynamically manages computing resources based on the Kubernetes elastic scheduler, assessing the computational task load according to the complexity of the compensation strategy for the virtual aging scenario. The scheduler automatically triggers horizontal scaling operations for container instances by monitoring container resource utilization metrics, increasing the number of simulation computing nodes during peak periods. Priority queues are used to manage verification tasks with different compensation strategies, and a time-constraint-aware algorithm is combined to allocate computing resources, shortening the verification cycle of high-priority strategies and improving the simulation efficiency of complex aging scenarios.

[0128] Incremental learning microservices and federated learning microservices construct a dual-channel data closed-loop link. The incremental learning microservice generates local model update gradients through real-time feedback data from the display panel, while the federated learning microservice aggregates encrypted gradient data from multiple devices. The dual channels employ a time window synchronization mechanism to coordinate the update rhythm. After the local gradient reaches the convergence threshold, a federated aggregation operation is triggered. A weighted average algorithm is used to fuse cross-device feature distributions, synchronously updating the time-varying adaptive parameters and device generalization parameters of the dynamic aging model, achieving a balance between model personalization and generality.

[0129] The service mesh and version control middleware build a state consistency maintenance system based on transaction logs. The service mesh appends transaction identifiers and operation sequence numbers during inter-microservice communication, while the version control middleware records state change events for each microservice in a distributed transaction log. When a local microservice version update is detected, state consistency is verified through a log replay mechanism, and a two-phase commit protocol is used to coordinate cross-service parameter updates. For detected risks of display parameter jumps, a version rollback mechanism is triggered to restore the system to the most recent stable state. Combined with a canary release strategy, updates are gradually rolled out to maintain a smooth transition during display parameter adjustments.

[0130] This invention achieves dynamic adaptive adjustment of an organic light-emitting display system through a multi-level technical architecture. Distributed photosensitive sensor nodes are deployed around the display panel. Each node operates independently based on a microservice architecture, collecting ambient light intensity and color temperature data via a photoelectric conversion module and transmitting standardized data packets to the central data bus using a lightweight API protocol. A programmable current sampling module captures fluctuations in current stress parameters in the drive circuit using an event-driven mode. This data is combined with heat distribution data generated by a temperature sensor network based on a topology-aware dynamic orchestration strategy, forming a multi-source heterogeneous data stream containing ambient light, current stress, and temperature. An edge computing unit performs sliding window mean filtering and abnormal pulse detection on the current stress data, generating filtered current stress time-series data. The temperature sensor network establishes a heat conduction path model based on the current stress time-series distribution and corrects the spatial resolution deviation of the heat distribution caused by the sparse layout of the sensor nodes using an interpolation algorithm.

[0131] The dynamic aging model microservice cluster consists of a current stress accumulation calculation service, a temperature bias nonlinear function solver, and a material relaxation coefficient matrix generator. Each microservice is coupled via an asynchronous message queue. Filtered current stress time-series data triggers a remote procedure call to the material relaxation coefficient matrix generator, accessing pre-stored current density-mobility decay curves in the experimental database and calculating the mobility decay error using corrected thermal distribution data. This error is nonlinearly coupled to the current stress accumulation distribution map using a tensor decomposition algorithm, outputting time-varying degradation characteristic parameters that reflect the synergistic effect of driving current stress, temperature gradient, and material relaxation effects in real time. The containerized simulation microservice constructs a virtual aging scenario based on these parameters, deploying an LSTM network as a containerized inference unit. After receiving the current stress time-series data stream, it performs online weight fine-tuning to generate an adaptive time-series correlation prediction model. The incremental learning algorithm injects the response delay data of organic materials in real time, updates the weights of the hidden layer of the model to adapt to the changes in aging rate, and executes the tasks of gamma correction weight, sub-pixel emission duration and TFT gate voltage adjustment in parallel through distributed computing nodes. The consistent hashing algorithm is used to maintain the spatiotemporal correlation of the compensation parameters and generate a global compensation coefficient matrix.

[0132] A multi-objective loss function solver receives the ΔE color difference threshold, ghosting suppression index, and power consumption constraints to construct a Pareto front analysis model. Coupled weight parameters are synchronously updated via a bidirectional communication channel between the backpropagation microservice and the dynamic aging model. Differential privacy technology injects Gaussian noise perturbation into the gradient sharing data, generating anonymized gradient updates that are transmitted to the dynamic aging model microservice cluster. At the edge, a display driver chip register mapping microservice is deployed to convert optimized driver parameters into binary instruction sets to adjust pixel drive current and refresh rate in real time. Compensation coefficient matrix version snapshots are timestamped via an event-triggered mechanism and synchronized to a cloud historical database to construct an aging parameter version control link. An online learning microservice, based on a publish-subscribe model, acquires brightness attenuation feedback data collected by the display panel's photoelectric sensors. This triggers a federated learning framework to aggregate cross-device feature parameters and, combined with a service mesh routing strategy, updates the global dynamic aging model weights, forming a dynamic adaptation closed loop between the compensation strategy and the actual attenuation. The service mesh and version control middleware maintain the consistency of microservice state through transaction logs, and gradually promote parameter updates by adopting a canary release strategy to suppress display parameter jumps caused by local microservice version iterations, thereby achieving continuous optimization and balance of display brightness, color gamut coverage and power consumption.

[0133] The technical features of this invention are explained below:

[0134] Distributed photosensitive sensor nodes: These refer to multiple independent sensing units deployed around the display panel. Each node collects ambient light intensity and color temperature data in real time through a photoelectric conversion module and operates independently based on a microservice architecture. Nodes communicate with the central data bus via a lightweight API protocol, enabling concurrent acquisition and low-latency transmission of multi-source data, avoiding single points of failure, and adapting to the expansion needs of different screen sizes.

[0135] The dynamic aging model microservice cluster consists of a distributed computing module comprised of a current stress accumulation calculation service, a temperature bias nonlinear function solver, and a material relaxation coefficient matrix generator. Each microservice interacts with data via an asynchronous message queue. The current stress accumulation service calculates the time integral effect of the driving current, the temperature bias solver establishes a nonlinear mapping between heat distribution and current stress, and the material relaxation coefficient generator quantifies the material aging rate using an experimental database. Together, they construct time-varying degradation characteristic parameters, replacing the traditional static model.

[0136] Containerized simulation microservice: This microservice encapsulates the LSTM network inference module using containerization technologies (such as Docker / Kubernetes) to achieve elastic scaling of computing resources. It receives filtered current stress time-series data streams, and through sliding time window partitioning and adaptive learning rate adjustment, fine-tunes network weights online to generate an adaptive time-series prediction model, supporting dynamic simulation of virtual aging scenarios.

[0137] Multi-objective loss function solver: Integrates a multi-dimensional optimization module that includes ΔE color difference threshold, ghosting suppression index, and power consumption constraints. It evaluates the non-dominated ranking results of candidate solution sets using a Pareto front analysis model. Combined with bidirectional communication between the backpropagation microservice and the dynamic aging model, it synchronously updates the model coupling weights, balancing the conflicting objectives of display quality and energy consumption.

[0138] Edge and cloud collaborative architecture: At the edge, a microservice for mapping display driver chip registers is deployed to directly execute optimized driver parameter adjustment instructions; the cloud stores historical versions of the compensation coefficient matrix and aging parameter evolution data. Brightness attenuation feedback data is synchronized through a publish-subscribe model, forming a layered optimization system of local real-time control and global model updates.

[0139] Incremental learning and federated learning microservices: Incremental learning microservices update LSTM network weights using real-time brightness decay data to adapt to changes in material aging rates; the federated learning framework aggregates anonymized data from multiple devices and uses a weighted average algorithm to fuse cross-device feature distributions, improving the model's generalization ability. Both technologies coordinate update rhythms through a time window synchronization mechanism, resolving the conflict between data privacy and model generalization.

[0140] Service Mesh and Version Control Middleware: The service mesh manages communication routing and load balancing between microservices, and adds transaction identifiers to ensure the timing consistency of data transmission; the version control middleware records parameter change events to a distributed transaction log, supports state rollback and canary release, and prevents display parameter jumps caused by partial updates.

[0141] Asynchronous message queue: A message middleware (such as RabbitMQ / Kafka) is used to achieve decoupled communication between microservices. When the filtered current stress data reaches the time window threshold, the queue is triggered to push the event to the material relaxation coefficient generator, ensuring the real-time performance and reliability of the data pipeline and avoiding service blockage.

[0142] Consistent hashing algorithm: In a distributed simulation microservice, a virtual node ring is constructed, mapping computational tasks to specific nodes based on the spatiotemporal labels of compensation parameters (such as pixel coordinates and timestamps). A redundant replication mechanism maintains the spatiotemporal correlation of parameters, preventing data loss due to node failures.

[0143] Differential privacy technology: During multi-objective optimization, Gaussian noise perturbation is injected into the gradient data output by the backpropagation microservice, and the noise intensity is controlled by a privacy budget. Local desensitization is implemented on the feature dimension of the gradient tensor to block the leakage path of sensitive information (such as the aging trajectory of specific pixels) and meet privacy compliance requirements.

[0144] Explanation of the synergistic relationship of the technical features of this invention:

[0145] Data flow closed loop: Distributed sensors → edge preprocessing → dynamic aging model → simulation microservices → multi-objective optimization → driving parameter adjustment → online learning feedback, forming a complete closed loop from data acquisition to model iteration.

[0146] Dynamic adaptation mechanism: Asynchronous communication and containerized elastic scheduling of microservice clusters support real-time updates of aging model parameters; dual-channel optimization of federated learning and incremental learning balances the needs of model personalization and generality.

[0147] Risk-resistant design: The service mesh maintains state consistency and has a version rollback mechanism to ensure system stability during frequent updates; the combination of differential privacy and edge computing balances data utility and privacy security.

[0148] By combining the above-mentioned technical features, the present invention realizes real-time tracking and compensation of the nonlinear aging trajectory of organic light-emitting units under the coupling effect of current stress, temperature bias and material relaxation effect, solves the problem of cumulative deviation caused by traditional static models, and significantly improves the lifespan and visual consistency of display panels.

[0149] This invention addresses the problem that traditional static aging models cannot adapt to nonlinear time-varying degradation characteristics through dynamic data acquisition, real-time model iteration, and a closed-loop feedback mechanism. The specific technical solution is as follows:

[0150] This invention utilizes a distributed photosensitive sensor, a programmable current sampling module, and a temperature sensor network to collect ambient light, current stress, and thermal distribution data in real time. A central data bus calls edge computing units for multi-source data preprocessing. The filtered time-series current stress data and corrected thermal distribution data are input into a dynamic aging model microservice cluster. The calculation of accumulated current stress, the solution of the temperature bias nonlinear function, and the generation of material relaxation coefficients are separated into independent microservices. Asynchronous message queues are used to couple data between services. Combined with current density-mobility decay curves from an experimental database, the mobility decay error is quantified, and time-varying degradation characteristic parameters are constructed. These parameters reflect the coupling effect of current stress, temperature bias, and material relaxation effects in real time, replacing the traditional static parameter mapping relationship.

[0151] Based on time-varying degradation parameters, a containerized simulation microservice constructs a virtual aging scenario and utilizes an LSTM network for online fine-tuning to generate an adaptive temporal correlation prediction model. An incremental learning algorithm injects organic material response delay data in real time, updating model weights to adapt to changes in the aging rate. The distributed simulation microservice decomposes the compensation strategy into parallel tasks of gamma correction weights, sub-pixel emission duration, and TFT gate voltage adjustment, employing a consistent hashing algorithm to maintain the spatiotemporal correlation of the compensation parameters. Tasks are distributed to distributed nodes via a service mesh to generate a global compensation coefficient matrix, resolving the problem of accumulated spatiotemporal deviations in the static compensation strategy.

[0152] A multi-objective loss function solver, combining ΔE chromatic aberration, ghosting suppression, and power consumption constraints, constructs a Pareto front analysis model. Coupled weights are synchronously updated through bidirectional communication between the backpropagation microservice and the dynamic aging model. Edge-based optimization drives parameter adjustments, while a cloud-based historical database records compensation matrix version snapshots, forming an aging parameter evolution chain. An online learning microservice, based on a publish-subscribe model, acquires brightness decay feedback data, triggering a federated learning framework to aggregate cross-device features and update global dynamic aging model parameters. The service mesh and version control middleware maintain state consistency, preventing parameter jumps through canary releases and rollback mechanisms, achieving continuous dynamic adaptation of compensation coefficients to actual decay amounts.

Claims

1. An adaptive adjustment method for active matrix organic light-emitting displays, characterized in that, include: Ambient light intensity and color temperature data are collected by deploying distributed photosensitive sensor nodes, and the ambient light intensity and color temperature data are transmitted to the dynamic aging model microservice cluster through the central data bus. The fluctuation of current stress parameters in the drive circuit is captured by a programmable current sampling module, and the heat distribution data collected by the temperature sensor network is obtained. The fluctuation of current stress parameters and heat distribution data are input into the dynamic aging model microservice cluster for parameter coupling calculation, and the time-varying degradation characteristic parameters are output. Based on the time-varying degradation characteristic parameters output by the microservice cluster of the dynamic aging model, a virtual aging scenario is constructed through containerized simulation microservices to generate a compensation coefficient matrix. The compensation coefficient matrix is ​​input into a multi-objective loss function solver, and the parameters are optimized by combining the preset ΔE color difference threshold and power consumption constraints, and the optimized display driving parameters are output. The optimized display driving parameters are adjusted in real time through an edge and cloud collaborative architecture, and the compensation coefficient matrix is ​​synchronized to the online learning microservice. Incremental model updates are triggered based on the optimized display driving parameters and the compensation coefficient matrix.

2. The adaptive adjustment method for active matrix organic light-emitting display according to claim 1, characterized in that, The data acquisition process of the distributed photosensitive sensor node includes: Each sensor node independently collects ambient light parameters through a microservice architecture and uploads the ambient light parameters to the central data bus via a lightweight API protocol; After receiving multi-source data including ambient light parameters, current stress parameter fluctuations and temperature distribution data, the central data bus calls the edge computing unit to perform abnormal pulse preprocessing on the current stress parameter fluctuations and generate filtered current stress time series data. The temperature sensor network is based on a topology-aware dynamic orchestration strategy. It establishes a heat conduction path model based on preprocessed current stress time series data and corrects the spatial resolution deviation of the heat distribution data.

3. The adaptive adjustment method for active matrix organic light-emitting display according to claim 2, characterized in that, The construction of the dynamic aging model microservice cluster includes: The current stress accumulation calculation service, the temperature bias nonlinear function solver, and the material relaxation coefficient matrix generator are separated into independent microservices. The data pipeline coupling between the current stress accumulation calculation service and the material relaxation coefficient matrix generator is achieved through an asynchronous message queue, wherein the filtered current stress time series data triggers the real-time correction of the material relaxation coefficient matrix generator. The material relaxation coefficient matrix generator calls the pre-stored current density and mobility decay curves in the experimental database via RPC, and combines them with the spatial resolution deviation of the corrected thermal distribution data to quantify the mobility decay error.

4. The adaptive adjustment method for active matrix organic light-emitting display according to claim 3, characterized in that, The operation process of the containerized simulation microservice includes: The LSTM network is deployed as a containerized inference microservice. After receiving the filtered current stress time series data stream, online weight fine-tuning is performed to generate an adaptive time series correlation prediction model. An incremental learning algorithm is used to inject the response delay data of organic materials into the training set of the adaptive time-series correlation prediction model, and to update the weight parameters of the adaptive time-series correlation prediction model. The compensation strategy is computed in parallel based on the adaptive temporal correlation prediction model through a distributed simulation microservice, and the spatiotemporal correlation of the compensation parameters is maintained by a consistent hashing algorithm.

5. The adaptive adjustment method for active matrix organic light-emitting display according to claim 4, characterized in that, The correction process for the material relaxation coefficient matrix includes: The temperature bias parameters of the intrinsic relaxation effect of the material are updated based on the corrected heat distribution data output by the temperature sensor network. The current density and mobility decay curves pre-stored in the experimental database are called, and the output results of the current stress accumulation calculation service are combined to generate the mobility decay compensation gradient. The mobility decay compensation gradient is nonlinearly coupled with the current stress accumulation calculation result, and the corrected relaxation coefficient matrix is ​​output to the dynamic aging model microservice cluster.

6. The adaptive adjustment method for active matrix organic light-emitting display according to claim 5, characterized in that, The process of constructing the virtual aging scenario includes: Based on the time-varying degradation parameters output by the microservice cluster of the dynamic aging model, a pixel-level aging compensation initial strategy is generated. The pixel-level aging compensation initial strategy is divided into three parallel microservice tasks: gamma correction weight calculation, sub-pixel emissivity duration optimization, and TFT gate voltage adjustment gradient. The three parallel microservice tasks are distributed to distributed computing nodes for collaborative execution via a service mesh, and the simulated state data of the virtual aging scenario is recorded using transaction logs.

7. The adaptive adjustment method for active matrix organic light-emitting display according to claim 6, characterized in that, The optimization process of the multi-objective loss function solver includes: Receive preset ΔE chromatic difference threshold, ghosting suppression index and power consumption constraints, and construct Pareto front analysis model; By backpropagating the bidirectional communication channel between the microservice and the dynamic aging model microservice cluster, the coupling weight parameters of the Pareto front analysis model and the dynamic aging model are updated synchronously. Differential privacy technology is used to de-identify the gradient sharing data output by the backpropagation microservice.

8. The adaptive adjustment method for active matrix organic light-emitting display according to claim 7, characterized in that, The execution process of the edge and cloud collaborative architecture includes: Deploy a display driver chip register mapping microservice at the edge to execute the optimized display driver parameter adjustment instructions output by the multi-objective loss function solver in real time; The version snapshot of the compensation coefficient matrix is ​​synchronized to the cloud historical database to establish a timestamp-based aging parameter version control link; The system obtains actual display panel brightness decay feedback data through a subscription and publish model, triggering incremental model update operations for the online learning microservice.

9. The adaptive adjustment method for active matrix organic light-emitting display according to claim 8, characterized in that, The incremental model update process includes: After receiving the brightness attenuation feedback data, the online learning microservice generates a historical database update instruction and attaches a version snapshot of the compensation coefficient matrix. The update command is routed to the dynamic aging model microservice cluster via the service mesh, triggering parameter recalibration of the current stress accumulation calculation service and the material relaxation coefficient matrix generator. The federated learning framework is used to aggregate the brightness decay feedback data and historical aging parameters across devices, and to update the weight coefficients of the dynamic aging model.

10. The adaptive adjustment method for an active matrix organic light-emitting display according to claim 9, characterized in that, Also includes: The distributed simulation microservice and edge preprocessing unit reduce the computational load on the central node through asynchronous data processing, which is used to support the dynamic expansion of ultra-high resolution screens. The containerized simulation microservice dynamically allocates computing resources based on an elastic resource scheduling strategy, thereby accelerating the verification efficiency of the compensation strategy in the virtual aging scenario. The online learning microservice and the federated learning framework synchronously optimize the time-varying adaptive parameters and cross-device generalization parameters of the dynamic aging model through a closed-loop data link. The service mesh and the preset version control middleware are based on simulated state data from transaction logs.

Images