ITSC 2024 Paper Abstract

Paper FrAT16.6

Tian, Bin (Chinese Academy of Sciences Institute of Automation), Yao, Tingting (Institute of Automation of ，Chinese Academy of Sciences), Lv, Yisheng (Institute of Automation, Chinese Academy of Sciences), Chen, shichao (Institute of Automation， Chinese Academy of Sciences), Sun, Yang (Hebei University of Engineering), Song, Ruiqi (institute of automation,chinese academy of sciences)

Parallel Data and Foundation Model Driven Closed-Loop of Autonomous Driving

Scheduled for presentation during the Poster Session "Operation and navigation of automated vehicles" (FrAT16), Friday, September 27, 2024, 10:30−12:30, Foyer

2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC), September 24- 27, 2024, Edmonton, Canada

This information is tentative and subject to change. Compiled on December 26, 2024

Keywords Sensing, Vision, and Perception, Multi-autonomous Vehicle Studies, Models, Techniques and Simulations

Abstract

Data closed loop plays a crucial role in autonomous driving for application in real world. The research for data closed loop on autonomous driving in urban scene have been conducted in the past few decades. But there is no unified framework for data closed loop related to autonomous driving in surface mine. The scenes in surface mine, which are unstructured, complex, changeable, and the objects in surface mine like rockfalls, which are differ in thousands of ways, not only put forward high generalization requirements for our perception system, but also bring many unpredictable risk for autonomous driving. In this work, we proposed a uniform framework of data closed loop driven by large-scale foundation model for autonomous driving in surface mine. Corner cases are predicted through hard scenes in simulation system, which is a parallel system with real scene. In addition, high quality data selection and deployment model distillation are conducted by method based on foundation model. This framework can not only enhance the generalization ability of the perception model, but also improve the efficiency for corner case mining, and has achieved good application results for autonomous driving in surface mine.