Towards Next-Generation Autonomous Driving through Advances in 3D Perception and Computing

Abstract

Autonomous driving technology promises to revolutionize transportation systems, yet faces significant challenges in 3D perception, real-time computing, and decision-making in complex, dynamic environments. This dissertation addresses these critical issues by presenting five novel contributions that advance the state-of-the-art in autonomous driving systems. The research focuses on improving both accuracy and computational efficiency for deployment in embedded automotive systems, tackling the multifaceted challenges of perception, computing, and visualization in autonomous vehicles. Key contributions include: a graph-based optimization strategy for 3D Multiple Dynamic Target Motion Tracking (MDTMT), integrating deep learning with graph optimization for robust multi-object tracking; DP-PointNet++, an enhanced architecture incorporating deformable convolutions for improved 3D point cloud processing; GraphSFR, a graph optimization approach for accurate and efficient scene flow estimation from 3D point cloud sequences; DataKernelFusion, a comprehensive framework optimizing perception pipelines for embedded automotive platforms; and a parallel fluid simulation approach for real-time multimedia applications in connected vehicles. These contributions collectively demonstrate significant improvements over existing methods in terms of accuracy, efficiency, and robustness, pushing the boundaries of 3D perception and computing for autonomous vehicles while addressing practical considerations for real-world deployment.

Brief Biography

Hongxian Yi is a Ph.D. candidate in Electrical and Computer Engineering at King Abdullah University of Science and Technology (KAUST). His research focuses on computer vision, robotics, and autonomous driving systems, with particular emphasis on 3D perception and efficient computing for embedded platforms. Hongxian obtained his M.S. in Computer Science from KAUST in 2020 and his B.S. in Computer Science and Technology from Shandong University in 2018. He has also studied at Beijing Institute of Technology and Chung-Ang University as an exchange student. Hongxian's professional experience includes internships at Amazon Web Services, Hitachi R&D, and Shanghai Institute of Technical Physics of the Chinese Academy of Sciences.