Directed Graph Traffic Situation Classification for Autonomous Driving

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for analyzing and categorizing driving scenario data sets are inadequate for effectively identifying and generalizing critical traffic situations, particularly in autonomous driving systems, as they rely on expert knowledge and struggle with data coverage and scenario optimization.

Innovation Solution

A computer-implemented method using a directed graph to classify traffic situations by segmenting movement behavior data into nodes and edges, allowing for the recognition of relevant scenarios and improved data selection, enabling the automatic identification of critical situations and their simulation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If expert knowledge-based methods are used to determine test scenarios, then scenario relevance is improved, but adaptability to new traffic situations deteriorates

Engineering Contradiction:
Improvescenario relevanceVSAvoidadaptability to new traffic situations
Core Design Contradiction:
ReliabilityVSAdaptability or versatility

Solution Approach 1:

The patent creates abstract templates that copy the essential structure and characteristics of critical traffic situations from real-world data. These templates serve as reusable models that can be instantiated with different parameters to generate varied test scenarios, maintaining reliability through proven patterns while enabling adaptability through parameter variation.

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The patent systematically varies parameters within the abstract templates (such as vehicle speeds, distances, timing intervals, and environmental conditions) to generate diverse test scenarios from a single template. This allows the system to adapt to new traffic situations by adjusting parameters while maintaining the core scenario structure that ensures reliability.

Inventive Principle:
Principle #35Parameter changes

2Productivity

If machine learning algorithms are used for data analysis, then data processing capability is improved, but generalization to different traffic situations deteriorates

Engineering Contradiction:
Improvedata processing capabilityVSAvoidgeneralization to different traffic situations
Core Design Contradiction:
ProductivityVSAdaptability or versatility

Solution Approach 1:

The patent extracts essential characteristics and patterns from large volumes of real-world sensor data to create simplified abstract templates. By taking out only the critical elements needed to define traffic situations (such as relative positions, velocities, and temporal relationships), the system achieves efficient data processing while maintaining generalization capability through the abstract nature of the templates.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent segments complex traffic situations into discrete, manageable components represented by nodes in the directed graph. Each node captures a specific aspect of the situation (e.g., vehicle detection, distance measurement, speed calculation), allowing the system to process data efficiently through modular operations while maintaining the ability to generalize by recombining these segments in different configurations.

Inventive Principle:
Principle #1Segmentation

3Reliability

If comprehensive test scenarios are created to cover all traffic situations, then validation coverage is improved, but device complexity deteriorates

Engineering Contradiction:
Improvevalidation coverageVSAvoidscenario management complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent creates abstract templates that serve multiple functions: they represent critical traffic situations, define test scenario structures, and enable scenario generation through parameter instantiation. This universal approach allows a single template to cover multiple specific situations by varying parameters, reducing the number of separate scenarios needed while maintaining comprehensive validation coverage.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent performs preliminary analysis of real-world data to identify and formalize critical traffic situation patterns before actual testing begins. By pre-defining abstract templates based on observed critical situations, the system prepares a reusable framework that simplifies subsequent scenario generation and testing, reducing complexity while ensuring comprehensive coverage of important cases.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP4475011A1Computer-implemented method and system for classifying a traffic situation
Publication Date: 2024.12.11 DSPACE SE & CO KG
  • EP4475011A1 patent drawingFigure 1a
  • EP4475011A1 patent drawingFigure 1b~2
  • EP4475011A1 patent drawingFigure 3~4

AI summary

The invention relates to a computer-implemented method and system (1) for classifying a given traffic situation encompassed by a data set of environmental data of a motor vehicle, comprising applying (S2) a directed graph (G) to the first data set (DS1), wherein nodes (11) of the directed graph (G) segment the first data set (DS1) into at least one segment of the movement behavior of the ego vehicle (10) and/or the fellow vehicle (12) relative to a vehicle environment, according to a first condition (14) fulfilled within a time interval, wherein edges (15) of the directed graph (G) symbolize connections between the respective nodes (11); and classifying (S3) the given traffic situation if all of the determined segments fulfill a second condition (16) of the given traffic situation.and an output (S4) of a class (K) representing the specified traffic situation and/or a second data set (DS2) containing the respective start and end time of the segment representing the specified traffic situation.;