We are pleased to announce that Yun Li, a second-year doctoral student, and Yuze Jiang, a first-year doctoral student, represented Tsukada Lab at the IEEE ITSC 2024 conference, held in Edmonton, Canada, from September 24 to 27, 2024. The 27th IEEE International Conference on Intelligent Transportation Systems (IEEE ITSC 2024) serves as the flagship event of the IEEE Intelligent Transportation Systems Society (ITSS), gathering researchers from around the globe to share advancements in the field of Intelligent Transportation Systems (ITS).
Yun Li’s Presentation:
Yun Li presented a paper titled “Large Language Models for Human-like Autonomous Driving Decision Making: A Survey.” This research emphasizes the potential of Large Language Models (LLMs) in transforming the field of Autonomous Driving (AD). LLMs, known for their language comprehension and generation capabilities, offer a promising path toward more human-like AD systems. As AD evolves from rule-based approaches to learning-based techniques like deep reinforcement learning, the integration of knowledge-based AD empowered by LLMs represents a significant advancement. Yun’s survey critically reviews recent progress in applying LLMs within both modular AD pipelines and end-to-end AD systems, while addressing challenges in real-time inference, safety, and deployment costs. By identifying key advancements and proposing future research directions, Yun’s work aims to inspire further developments in creating safer, more intelligent, and human-centric AD technologies.
In addition to presenting his paper, Yun Li actively contributed to discussions in three workshops during ITSC 2024:
- The 1st FMAD Workshop – Foundation Models for Autonomous Driving
- The 2nd Workshop on Large Language and Vision Models for Autonomous Driving (LLVM-AD)
- Vision Language Model-based Human-Centered Autonomous Driving Workshop
Yuze Jiang’s Presentation:
Yuze Jiang presented his research on “Accurate Cooperative Localization Utilizing LiDAR-equipped Roadside Infrastructure for Autonomous Driving.” His work explores the enhanced accuracy and reliability of autonomous vehicle localization through the use of road side units (RSUs) in conjunction with vehicle-to-infrastructure (V2I) communications. With recent improvements in LiDAR technology, achieving centimeter-level localization accuracy has become more feasible. However, traditional vehicle self-localization techniques face challenges in environments lacking identifiable map features. Yuze’s approach leverages RSUs as stable reference points, enhancing localization precision in difficult environments. Evaluation results using an end-to-end autonomous driving simulator (AWSIM) showed that his method improves localization accuracy by up to 80% compared to traditional methods and demonstrates robust performance in network-delayed conditions. This research contributes to more reliable autonomous driving systems, especially in scenarios where conventional methods may fall short.
Publications
