Vehicle Sensor Data Filtering Using ML Disagreement Signals

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Autonomous vehicles face challenges in managing and processing large amounts of sensor data, with much of it being irrelevant for analysis, leading to increased resource usage and storage needs.

Innovation Solution

Implementing uncertainty-based data filtering by acquiring sensor data, applying a first machine learning model, determining discrepancies with a second model or driver actions, and transmitting only data that causes disagreements to a server for analysis.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If all sensor data is stored and transmitted to server, then complete data analysis is possible, but storage and network resources are wasted on irrelevant data

Engineering Contradiction:

Improvedata analysis completenessVSAvoidstorage and network resource consumption

Core Design Contradiction:

ReliabilityVSLoss of energy

Solution Approach 1:

The patent applies preliminary action by using machine learning models to pre-analyze sensor data in the vehicle before transmission. The system processes sensor data through multiple ML models and identifies disagreements in results, filtering out data that doesn't require server analysis. This preliminary filtering action occurs at the edge device (vehicle) before data transmission, reducing unnecessary network and storage resource usage while maintaining analysis completeness for important data.

Inventive Principle:

Principle #10Preliminary action

2Loss of energy

If sensor data is filtered to reduce data size, then resource usage decreases, but valuable data for model refinement may be lost

Engineering Contradiction:

Improveprocessing and storage resource usageVSAvoidvaluable data for machine learning refinement

Core Design Contradiction:

Loss of energyVSLoss of information

Solution Approach 1:

The patent implements feedback by using multiple machine learning models to evaluate sensor data and comparing their results. When disagreements occur between model predictions, the system identifies this as valuable information indicating edge cases or uncertainties that require further analysis. This feedback mechanism ensures that data potentially useful for model refinement is preserved and transmitted, while filtering out routine, non-informative data.

Inventive Principle:

Principle #23Feedback

3Measurement precision

If multiple machine learning models are applied to filter data, then data quality for transmission improves, but processing time and computational resources increase

Engineering Contradiction:

Improvedata selection accuracyVSAvoiddata processing time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent applies partial action by using multiple machine learning models selectively rather than processing all data through all models. The system employs an ensemble of models but only transmits data where disagreements occur, representing a partial processing approach. This balances computational effort with data selection accuracy, avoiding excessive processing of clearly irrelevant data while maintaining precision through multi-model evaluation of borderline cases.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS11829896B2Uncertainty-based data filtering in a vehicle

Publication Date: 2023.11.28 APPLIED INTUITION INC

AI summary

Uncertainty-based data filtering in a vehicle may include acquiring sensor data from a plurality of sensors of the autonomous vehicle; applying a first machine learning model to the sensor data; determining that a result of the first machine learning model disagrees with another result; and transmitting, based on the result of the first machine learning model disagreeing with the other result, the sensor data to a server.