Salvato in:
Dettagli Bibliografici
Autori principali: Huang, Jihao, Xia, Xi, Li, Zhiyuan, Liu, Tianle, Wang, Jingke, Chen, Junbo, Ye, Tengju
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:https://arxiv.org/abs/2512.15038
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866915683523100672
author Huang, Jihao
Xia, Xi
Li, Zhiyuan
Liu, Tianle
Wang, Jingke
Chen, Junbo
Ye, Tengju
author_facet Huang, Jihao
Xia, Xi
Li, Zhiyuan
Liu, Tianle
Wang, Jingke
Chen, Junbo
Ye, Tengju
contents End-to-end paradigms have demonstrated great potential for autonomous driving. Additionally, most existing methods are built upon Transformer architectures. However, transformers incur a quadratic attention cost, limiting their ability to model long spatial and temporal sequences-particularly on resource-constrained edge platforms. As autonomous driving inherently demands efficient temporal modeling, this challenge severely limits their deployment and real-time performance. Recently, linear attention mechanisms have gained increasing attention due to their superior spatiotemporal complexity. However, existing linear attention architectures are limited to self-attention, lacking support for cross-modal and cross-temporal interactions-both crucial for autonomous driving. In this work, we propose LADY, the first fully linear attention-based generative model for end-to-end autonomous driving. LADY enables fusion of long-range temporal context at inference with constant computational and memory costs, regardless of the history length of camera and LiDAR features. Additionally, we introduce a lightweight linear cross-attention mechanism that enables effective cross-modal information exchange. Experiments on the NAVSIM and Bench2Drive benchmarks demonstrate that LADY achieves state-of-the-art performance with constant-time and memory complexity, offering improved planning performance and significantly reduced computational cost. Additionally, the model has been deployed and validated on edge devices, demonstrating its practicality in resource-limited scenarios.
format Preprint
id arxiv_https___arxiv_org_abs_2512_15038
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle LADY: Linear Attention for Autonomous Driving Efficiency without Transformers
Huang, Jihao
Xia, Xi
Li, Zhiyuan
Liu, Tianle
Wang, Jingke
Chen, Junbo
Ye, Tengju
Artificial Intelligence
End-to-end paradigms have demonstrated great potential for autonomous driving. Additionally, most existing methods are built upon Transformer architectures. However, transformers incur a quadratic attention cost, limiting their ability to model long spatial and temporal sequences-particularly on resource-constrained edge platforms. As autonomous driving inherently demands efficient temporal modeling, this challenge severely limits their deployment and real-time performance. Recently, linear attention mechanisms have gained increasing attention due to their superior spatiotemporal complexity. However, existing linear attention architectures are limited to self-attention, lacking support for cross-modal and cross-temporal interactions-both crucial for autonomous driving. In this work, we propose LADY, the first fully linear attention-based generative model for end-to-end autonomous driving. LADY enables fusion of long-range temporal context at inference with constant computational and memory costs, regardless of the history length of camera and LiDAR features. Additionally, we introduce a lightweight linear cross-attention mechanism that enables effective cross-modal information exchange. Experiments on the NAVSIM and Bench2Drive benchmarks demonstrate that LADY achieves state-of-the-art performance with constant-time and memory complexity, offering improved planning performance and significantly reduced computational cost. Additionally, the model has been deployed and validated on edge devices, demonstrating its practicality in resource-limited scenarios.
title LADY: Linear Attention for Autonomous Driving Efficiency without Transformers
topic Artificial Intelligence
url https://arxiv.org/abs/2512.15038