End-to-End Autonomous Driving through V2X Cooperation

End-to-End Autonomous Driving through V2X Cooperation

20 Apr 2024 | Haibao Yu1,2*, Wenxian Yang2*, Jiaru Zhong2,3*, Zhenwei Yang2,4, Siqi Fan2, Ping Luo1, and Zaiqing Nie2†
The paper introduces UniV2X, an innovative end-to-end framework for cooperative autonomous driving that integrates key driving modules across diverse views into a unified network. UniV2X addresses the challenges of limited communication bandwidth, latency, and data corruption by proposing a sparse-dense hybrid data transmission and fusion mechanism. This mechanism enhances agent perception, online mapping, and occupancy prediction, ultimately improving planning performance. The framework is evaluated on the DAIR-V2X dataset, demonstrating significant improvements in collision rate and planning performance while maintaining low transmission costs. The paper also includes a detailed analysis of the effectiveness of each fusion module and evaluates the reliability of UniV2X under various communication conditions.The paper introduces UniV2X, an innovative end-to-end framework for cooperative autonomous driving that integrates key driving modules across diverse views into a unified network. UniV2X addresses the challenges of limited communication bandwidth, latency, and data corruption by proposing a sparse-dense hybrid data transmission and fusion mechanism. This mechanism enhances agent perception, online mapping, and occupancy prediction, ultimately improving planning performance. The framework is evaluated on the DAIR-V2X dataset, demonstrating significant improvements in collision rate and planning performance while maintaining low transmission costs. The paper also includes a detailed analysis of the effectiveness of each fusion module and evaluates the reliability of UniV2X under various communication conditions.
Reach us at info@study.space
Understanding End-to-End Autonomous Driving through V2X Cooperation