Implementation of Re-Simulation-Based Integrated Analysis System to Evaluate and Improve Autonomous Driving Algorithms

Autonomous driving technology requires rigorous testing and validation of perception, decision-making, and control algorithms to ensure safety and reliability. Although existing simulators and testing tools play critical roles in algorithm evaluation, they struggle to satisfy the demands of complex,...

Full description

Saved in:
Bibliographic Details
Main Authors: Soobin Jeon, Junehong Park, Dongmahn Seo
Format: Article
Language:English
Published: MDPI AG 2024-12-01
Series:Vehicles
Subjects:
Online Access:https://www.mdpi.com/2624-8921/6/4/108
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Autonomous driving technology requires rigorous testing and validation of perception, decision-making, and control algorithms to ensure safety and reliability. Although existing simulators and testing tools play critical roles in algorithm evaluation, they struggle to satisfy the demands of complex, real-time systems. This study proposes a re-simulation-based integrated analysis system designed to overcome these challenges by providing advanced visualization, algorithm-testing, re-simulation, and data-handling capabilities. The proposed system features a comprehensive visualization module for real-time analysis of diverse sensor data and ego vehicle information, offering intuitive insights to researchers. Additionally, it includes a flexible algorithm-testing framework that abstracts simulator-specific dependencies, enabling seamless integration and evaluation of algorithms in various scenarios. The system also introduces robust re-simulation capabilities, enhancing algorithm validation using iterative testing based on real-world or simulated sensor data. To address the computational demands of high-frequency sensor data, the system employs optimized data-handling mechanisms based on shared memory, thereby significantly reducing latency and improving scalability. The proposed system overcomes critical challenges faced by existing alternatives by providing a robust, efficient, and scalable solution for testing and validating autonomous-driving algorithms, ultimately accelerating the development of safe and reliable autonomous vehicles.
ISSN:2624-8921