ML Meteorological Data Compression for Low-Bandwidth Distribution

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for compressing and distributing meteorological data are inefficient, requiring significant storage and bandwidth due to large dataset sizes, and often result in information loss or high computational demands, especially in scenarios like aviation where resources are limited.

Innovation Solution

The use of machine learning encoding models, such as autoencoders and proper orthogonal decomposition, to compress meteorological data into a smaller, more manageable format that can be decoded with minimal information loss, allowing for efficient storage and distribution, particularly in systems with limited memory and bandwidth.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If conventional compression methods are used on raw meteorological data, then storage size is reduced, but information loss occurs and computational demand increases

Engineering Contradiction:
Improvestorage sizeVSAvoidinformation loss
Core Design Contradiction:
Quantity of substanceVSLoss of information

Solution Approach 1:

The patent transforms meteorological data from physical space to latent space using machine learning encoders, changing the parameter representation from raw measurements to compressed latent vectors that preserve essential information while reducing storage requirements by 90-95% without significant information loss

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent replaces conventional mechanical compression algorithms with machine learning-based encoder-decoder models that learn optimal compression representations, achieving better information preservation through neural network-based transformation rather than traditional compression mechanics

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

2Loss of information

If raw meteorological data is transmitted to ensure data accuracy, then information completeness is maintained, but bandwidth requirements increase significantly

Engineering Contradiction:
Improvedata accuracyVSAvoidbandwidth requirements
Core Design Contradiction:
Loss of informationVSQuantity of substance

Solution Approach 1:

The patent extracts only the essential meteorological information features through the encoder model, transmitting only the compressed latent representations rather than complete raw data, thereby maintaining data accuracy for critical parameters while reducing bandwidth consumption by 90-95%

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent transforms data from high-dimensional physical space to low-dimensional latent space, enabling efficient transmission by representing complex meteorological states in a compressed dimensional framework that preserves accuracy while minimizing bandwidth usage

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

3Quantity of substance

If machine learning encoding models are used to compress data, then storage and bandwidth requirements are reduced, but device complexity increases

Engineering Contradiction:
Improvestorage requirementsVSAvoiddevice complexity
Core Design Contradiction:
Quantity of substanceVSDevice complexity

Solution Approach 1:

The patent performs data compression and encoding in advance using trained machine learning models, so that when data needs to be stored or transmitted, the complex compression task has already been completed, reducing the computational burden on resource-constrained devices at the time of use

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent uses pre-trained encoder models that can be deployed as standalone compression components, allowing the complex machine learning infrastructure to be separated from the resource-constrained aviation systems that only need to store or transmit the already-compressed representations

Inventive Principle:
Principle #26Copying

4Quantity of substance

If small unmanned air vehicles use reduced storage systems, then device size and weight are reduced, but the ability to process weather information accurately deteriorates

Engineering Contradiction:
Improvestorage capacityVSAvoidweather information accuracy
Core Design Contradiction:
Quantity of substanceVSMeasurement precision

Solution Approach 1:

The patent transforms weather data into a compressed latent representation that preserves the essential physical relationships and patterns, allowing small vehicles with limited storage to maintain accurate weather information processing by working with the transformed parameter space rather than raw data

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP4304095A1Compression and distribution of meteorological data using machine learning
Publication Date: 2024.01.10 THE BOEING CO
  • EP4304095A1 patent drawingFigure 1
  • EP4304095A1 patent drawingFigure 2
  • EP4304095A1 patent drawingFigure 3

AI summary

Apparatuses, methods, systems, and program products are disclosed for compression and distribution of meteorological data using machine learning. An apparatus includes a processor and a memory that stores code executable by the processor to receive a raw meteorological data set for a time frame, the raw meteorological data set comprising a plurality of dimensions. The code is executable by the processor to compress the raw meteorological data set using a machine learning encoding model to create an encoded meteorological data set that has a storage size that is smaller than a storage size of the raw meteorological data set, wherein the encoded meteorological data set can be decoded to create a decoded meteorological data set that is substantially similar to the raw meteorological data set. The code is executable by the processor to make the encoded meteorological data set accessible to one or more end users.