Vehicle Sequence Inference for Driving Event Classification

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current Advanced Driver Assistance Systems (ADAS) and autonomous driving systems face challenges in detecting and classifying temporally extended driving events, leading to inefficiencies in monitoring driver behavior and attributing causes of unsafe situations.

Innovation Solution

A method and apparatus that combine first and second inferences from various sensors, such as camera and inertial sensors, into a data vector to determine a sequence inference, enabling improved detection and classification of driving events through a neural network-based system.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If current ADAS systems use simple sensor detection methods, then the system complexity is low, but the ability to detect and classify temporally extended driving events is insufficient

Engineering Contradiction:

Improvedetection and classification accuracy of driving eventsVSAvoidsystem complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the detection process into discrete temporal events with defined start and end conditions. Each driving event is broken down into detectable sensor signal patterns, allowing complex temporally extended events to be analyzed through sequential simpler detections rather than requiring the entire event to be captured simultaneously.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent adds a temporal dimension to traditional sensor detection by analyzing sequences of sensor readings over time. Instead of detecting events at a single moment, the system detects patterns across multiple time steps, transforming spatial-only detection into spatio-temporal detection to identify causally related events.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

2Reliability

If the system monitors more driving events and sequences, then the driver behavior monitoring capability improves, but the computational resources and processing time increase

Engineering Contradiction:

Improvedriver behavior monitoring reliabilityVSAvoidcomputational energy consumption

Core Design Contradiction:

ReliabilityVSUse of energy by moving object

Solution Approach 1:

The system pre-defines specific driving events and their corresponding sensor detection criteria before runtime. By establishing the sequence of events to monitor and their detection rules in advance, the system avoids computationally expensive real-time analysis of all possible driving scenarios, reducing energy consumption while maintaining monitoring reliability.

Inventive Principle:

Principle #10Preliminary action

3Measurement precision

If the system attributes causes to driving events using basic detection, then the processing is fast, but the accuracy of cause attribution is low

Engineering Contradiction:

Improvecause attribution accuracyVSAvoidevent processing time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent introduces an intermediary layer that maps sensor detections to predefined driving events and then to potential causes. This intermediary mapping structure allows the system to efficiently determine causes by checking against pre-established event-cause relationships rather than performing complex real-time causal analysis, balancing accuracy with processing speed.

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS11840239B2Multiple exposure event determination

Publication Date: 2023.12.12 NETRADYNE INC

AI summary

Systems, devices and methods provide, implement, and use vision-based methods of sequence inference for a device affixed to a vehicle.