Crowdsourced Neural Network Validation for Vehicle Sensor Failure Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Autonomous vehicles face diminished control due to inaccurate internal state parameter estimation from in-vehicle sensors, necessitating a method to detect and validate sensor failure states effectively.

Innovation Solution

A system and method utilizing crowdsourced data and a neural network to validate in-vehicle sensor data by training a neural network with both sensor and crowdsourced data, determining a check value, and assigning reputation scores to participating agents, thereby ensuring accurate state estimation and vehicle operation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If in-vehicle sensors are used to measure internal state parameters, then the vehicle can navigate based on these parameters, but the accuracy of state estimation deteriorates when sensors fail

Engineering Contradiction:

Improvevehicle navigation reliabilityVSAvoidinternal state parameter accuracy

Core Design Contradiction:

ReliabilityVSMeasurement precision

Solution Approach 1:

The patent introduces crowdsourced data from multiple external vehicles as an intermediary to validate sensor measurements. The neural network processes both in-vehicle sensor data and crowdsourced data from other vehicles to determine whether sensor failures have occurred, using the crowdsourced information as a mediator to cross-validate and verify the accuracy of internal state parameters.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Device complexity

If sensor data is used directly for state estimation, then the system remains simple, but the ability to detect sensor failures is insufficient

Engineering Contradiction:

Improvestate estimation system complexityVSAvoidsensor failure detection capability

Core Design Contradiction:

Device complexityVSReliability

Solution Approach 1:

The patent implements a feedback mechanism where the neural network continuously compares in-vehicle sensor data with crowdsourced data from multiple vehicles to generate validation results. This feedback loop enables the system to detect sensor failures by identifying discrepancies between direct sensor measurements and indirect crowdsourced measurements, thereby improving failure detection capability while maintaining manageable system complexity.

Inventive Principle:

Principle #23Feedback

3Measurement precision

If crowdsourced data is collected from multiple vehicles, then the validation accuracy improves, but the data processing complexity increases

Engineering Contradiction:

Improvestate parameter validation accuracyVSAvoiddata processing complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent replaces complex manual data processing mechanisms with a neural network-based automated system. The neural network efficiently processes crowdsourced data from multiple vehicles, performing pattern recognition and validation tasks that would be computationally intensive if handled by traditional algorithms, thereby achieving high validation accuracy while managing data processing complexity through intelligent automation.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS11532188B2Architecture and methodology for state estimation failure detection using crowdsourcing and deep learning

Publication Date: 2022.12.20 GM GLOBAL TECHNOLOGY OPERATIONS LLC

AI summary

A vehicle and a system and method for operating a vehicle. The system includes a state estimator and a processor. A detected value of a parameter of the vehicle is determined using sensor data obtained by in-vehicle detectors. The processor determines a check value of the parameter based on crowdsourced data, validates the detected value of the parameter based on the check value of the parameter, and operates the vehicle based on the validation.