GAN Migration Model Optimizes Simulation Data Accuracy

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing simulation technologies face challenges in achieving accurate and efficient simulation data generation, particularly in complex systems, due to noise and unknown rules, which limits the accuracy of point cloud data and requires extensive resources for data screening and priori knowledge expansion.

Innovation Solution

The use of a generative adversarial network (GAN) with a migration model to optimize simulation data by refining it to be closer to real data, reducing the need for high-quality diverse samples and human resources, and improving the accuracy and reliability of simulation data while reducing simulator construction costs.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If data screening is used to improve point cloud accuracy, then measurement precision is improved, but device complexity increases due to high requirements for simulator and diverse samples

Engineering Contradiction:
Improvepoint cloud accuracyVSAvoidsimulator complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent replaces the traditional mechanical data screening process with a deep learning-based automatic correction system. The neural network model automatically identifies and corrects abnormal values in point cloud data, substituting the manual or rule-based screening process with an intelligent system that learns from training data, thereby reducing the complexity requirements of the simulator while maintaining or improving accuracy

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent introduces an intermediary correction model (neural network) that mediates between the raw simulation data and the final processed data. This intermediary layer automatically identifies and corrects abnormalities without requiring the simulator itself to be highly complex or to generate perfectly diverse samples, thus resolving the contradiction between accuracy and complexity

Inventive Principle:
Principle #24Intermediary (Mediator)

2Measurement precision

If priori knowledge is increased to improve simulation accuracy, then measurement precision is improved, but loss of time increases due to endless work for complex systems

Engineering Contradiction:
Improvesimulation accuracyVSAvoiddata collection time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent applies preliminary action by pre-training the neural network model on a dataset before deployment. The model learns from training data in advance, capturing patterns and relationships that would otherwise require extensive manual priori knowledge collection. This preliminary learning phase automates what would otherwise be an endless manual work process for complex systems

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system implements self-service through automatic correction, where the neural network model autonomously identifies and corrects abnormalities in point cloud data without requiring continuous human intervention or manual priori knowledge input. The model serves itself by learning from training data and then independently processing simulation data, eliminating the endless cycle of manual data collection and processing

Inventive Principle:
Principle #25Self-service

3Measurement precision

If high-quality diverse samples are required for data screening, then measurement precision is improved, but productivity decreases due to high requirements put on simulator

Engineering Contradiction:
Improvesample qualityVSAvoiddata generation efficiency
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The patent replaces the mechanical requirement for high-quality diverse samples with an intelligent correction system. Instead of relying on the simulator to generate perfect diverse samples, the neural network model learns from training data and automatically corrects abnormalities, thereby maintaining high measurement precision while reducing the burden on the simulator and improving overall productivity

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentEP3620988B1Method, device for optimizing simulation data, and computer-readable storage medium
Publication Date: 2024.08.21 BAIDU ONLINE NETWORK TECH (BEIJIBG) CO LTD
  • EP3620988B1 patent drawingFigure 1
  • EP3620988B1 patent drawingFigure 2~3
  • EP3620988B1 patent drawingFigure 4~6

AI summary

A method, device and a computer-readable storage medium for optimizing simulation data are provided. The method for optimizing simulation data includes: inputting the simulation data generated by a simulator to a first generative adversarial network comprising a migration model; and optimizing the simulation data generated by the simulator with the migration model to generate optimized simulation data. In an embodiment of the present application, the simulation data is optimized by the generative adversarial network to enable the simulation data closer to the real data in representation. Therefore, the quality and accuracy of the simulation data can be ensured, the validity and reliability of the simulation data can be improved to some extent, and the cost for constructing the simulator can also be reduced.