Driving Log Extraction for ADAS Simulation Scenario Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional simulation tools for advanced driver assistance (ADAS) and autonomous driving are limited by the inability to effectively utilize large volumes of driving log data from on-road tests and often miss critical objects, leading to incomplete or unusable simulation scenarios.
Innovation Solution
A system and method that utilizes WGS84 GPS coordinates to combine driving log files with map data, generating whitelists and blacklists to create comprehensive simulation scenarios, using templates to ensure compatibility with virtual simulation tools, and iteratively updating blacklists for accurate and robust simulations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Duration of action of moving object
If conventional simulation tools are used to extract scenarios from driving log files, then the extraction process is simple, but the simulation duration is limited to only small periods (10-20 seconds) and critical objects may be missed
Solution Approach 1:
The system segments the driving log file processing into multiple passes: initial scenario extraction, object detection verification, and iterative refinement. By dividing the processing into discrete stages with specific focus on different object types and time periods, the system can extend simulation duration while maintaining detection accuracy through targeted analysis at each segment.
Solution Approach 2:
The system performs preliminary actions by pre-identifying and cataloging all objects in the driving log file before simulation extraction. This includes pre-processing sensor data to detect and track objects, creating a comprehensive object database that ensures no critical objects are missed during the extended simulation period, thereby enabling longer durations without sacrificing reliability.
2Adaptability or versatility
If conventional extraction methods are used, then the process is straightforward, but the extracted scenarios cannot be used in general (open-source) simulation tools
Solution Approach 1:
The system implements universality by designing a multi-functional extraction platform that can output scenarios in multiple simulation tool formats (CARLA, LGSVL, open-source tools). The extraction engine performs core processing once and generates compatible output for various simulation environments, allowing a single complex system to serve multiple simulation tool requirements without requiring separate extraction processes for each tool.
Solution Approach 2:
The system introduces an intermediary conversion layer that translates driving log file data into standardized scenario formats compatible with various simulation tools. This intermediary processing stage acts as a mediator between the raw sensor data and the specific requirements of different simulation environments, enabling broad compatibility while managing complexity through a unified intermediate representation.
3Reliability
If conventional simulation tools are used, then they work for their intended purpose, but moving objects near the host vehicle are missed without any reason
Solution Approach 1:
The system applies dynamics by implementing adaptive processing that adjusts analysis intensity based on object characteristics and proximity to the host vehicle. Critical objects near the host vehicle receive enhanced detection algorithms and extended tracking periods, while less critical objects use standard processing. This dynamic approach improves detection accuracy for important objects while maintaining overall processing efficiency through selective resource allocation.
Solution Approach 2:
The system incorporates feedback mechanisms that continuously monitor detection results and adjust processing parameters in real-time. When objects are detected or when detection confidence is low, the system automatically triggers re-analysis or extended tracking, ensuring no critical objects are missed. This feedback-driven approach maintains high reliability while optimizing productivity by focusing computational resources only when and where needed.
Data Source
AI summary
Vehicle advanced driver assistance (ADAS) and autonomous driving feature simulation and verification systems and method receive a driving log file and map data corresponding to an on-road driving scenario by a vehicle and overlay the driving log file and the map data to generate combined data that localizes a position of the vehicle. A whitelist comprising a set of included objects for an ADAS/autonomous driving feature simulation and (ii) a blacklist comprising a set of excluded objects for the ADAS/autonomous driving feature simulation and a template for the ADAS/autonomous driving feature simulation are then obtained. Finally, an ADAS/autonomous driving feature simulation scenario is generated in a desired format using the white and blacklists and the template and executed using a corresponding simulation tool and a result of the executed simulation of the ADAS/autonomous driving feature simulation scenario is verified.