ML Data Quality Evaluation for Autonomous Driving Maps and Sensors

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

In autonomous driving applications, imprecise and inaccurate sensor data can lead to erroneous observations, potentially causing accidents and illegal driving maneuvers due to the generation of ineffective navigation and mapping models.

Innovation Solution

A system and method using a machine learning model to evaluate the quality of sensor data and map data by calculating information scores and recursively splitting features, determining reliability based on pre-defined thresholds, and generating notification messages for accurate navigation assistance.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If sensor data is collected from consumer vehicles for navigation applications, then large amounts of data are available for autonomous driving, but the data quality becomes imprecise and inaccurate

Engineering Contradiction:

Improveamount of sensor dataVSAvoidquality of sensor data

Core Design Contradiction:

Quantity of substanceVSMeasurement precision

Solution Approach 1:

The patent introduces an intermediary quality evaluation system that acts as a mediator between raw sensor data collection and navigation model development. This system evaluates data quality using multiple metrics (completeness, consistency, accuracy) and filters/selects high-quality data before it is used for training navigation models, thereby resolving the contradiction between large data quantity and data precision

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent replaces traditional mechanical/manual data quality assessment methods with automated machine learning-based evaluation systems. The system uses algorithms to automatically assess sensor data quality, detect errors, and filter problematic data, substituting human judgment with computational analysis to maintain precision at scale

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Productivity

If imprecise sensor data is used to develop navigation algorithms, then model development can proceed with available data, but the generated models become ineffective and lead to erroneous observations

Engineering Contradiction:

Improvemodel development speedVSAvoideffectiveness of navigation models

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The patent implements preliminary data quality evaluation and filtering actions before the model development process begins. By pre-assessing and pre-filtering sensor data to ensure quality standards are met, the system prevents ineffective data from entering the training pipeline, thereby maintaining both development efficiency and model reliability

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent establishes feedback loops where navigation model performance is continuously monitored and used to refine data quality standards. When models produce erroneous observations, the feedback mechanism identifies problematic data patterns and adjusts quality thresholds, creating a self-improving system that maintains reliability while preserving productivity

Inventive Principle:

Principle #23Feedback

3Speed

If erroneous observations are utilized locally by vehicles for navigation functions, then immediate navigation decisions can be made, but the vehicle may incorrectly comprehend observations leading to accidents or illegal maneuvers

Engineering Contradiction:

Improvenavigation decision speedVSAvoidrisk of accidents and illegal maneuvers

Core Design Contradiction:

SpeedVSObject-affected harmful factors

Solution Approach 1:

The patent implements beforehand cushioning by layering multiple verification mechanisms between sensor observation and navigation decision-making. The system uses redundant sensing, cross-validation algorithms, and safety constraint checks that cushion against erroneous observations before they can lead to harmful actions, allowing fast decisions while mitigating risk

Inventive Principle:

Principle #11Beforehand cushioning (Prior cushioning)

Data Source

PatentUS11537944B2Method and system to generate machine learning model for evaluating quality of data

Publication Date: 2022.12.27 HERE GLOBAL BV

AI summary

A system, a method and a computer program product are provided for evaluating quality of data, such as sensor data and map data, using a machine learning model. The system may include at least one memory configured to store computer executable instructions and at least one processor configured to execute the computer executable instructions to obtain first sensor features of the first sensor data associated with a road object in a first geographic region, first map features of the first map data associated with the road object and ground truth data associated with the road object. The processor may be configured to generate the machine learning model by configuring the ground truth data and calculating first information scores for each of the first sensor features and the first map features by recursively splitting each of the first sensor features and the first map features.