@misc{Tsukada2025f,

title = {Adaptive V2X Collaboration for End-to-End Autonomous Driving},

author = {Manabu Tsukada},

url = {https://drivex-workshop.github.io/itsc2025},

year  = {2025},

date = {2025-11-18},

abstract = {End-to-end autonomous driving requires more than single-vehicle perception and decision making—it depends on timely, reliable, and context-aware collaboration among multiple agents. This talk presents an adaptive V2X collaboration framework designed to enhance robustness and situational understanding in complex urban environments. We introduce three key components: Co3SOP, a collaborative semantic occupancy representation that enables feature-level information sharing beyond object messages; PHCP, an adaptive cooperative perception mechanism that handles heterogeneous sensors and model architectures during inference; and PrefDrive, a preference-guided decision module that improves cooperation and interaction comfort among vehicles and pedestrians. Integrated with open-source Autoware systems and large-scale digital twin environments, these components form a practical pipeline for realizing cooperative end-to-end driving. The talk highlights challenges in multi-agent alignment, reliability, and human-centered behavior, and outlines future directions toward scalable V2X-enabled cooperation.},

howpublished = {Keynote at DriveX (3rd Edition) Workshop on Foundation Models for V2X-Based Cooperative Autonomous Driving In conjunction with ITSC 2025, Nov 18, Gold Coast, Australia},

keywords = {},

pubstate = {published},

tppubtype = {presentation}

}