StyleVLA: 驾驶风格感知的视觉语言动作模型用于自动驾驶
StyleVLA: Driving Style-Aware Vision Language Action Model for Autonomous Driving
AI总结：
StyleVLA通过引入物理约束和混合损失，生成多样且物理合理的驾驶行为，优于现有专有和先进模型。
摘要：
视觉语言模型（VLMs）将视觉感知与语言推理联系起来。在自动驾驶（AD）中，这种协同效应使视觉语言动作（VLA）模型成为可能，这些模型将高层次的多模态理解转化为驾驶行为，通常表示为未来轨迹。然而，现有VLA模型主要生成通用的无碰撞轨迹。除了避障之外，适应多样化的驾驶风格（例如，激进、舒适）对于个性化驾驶至关重要。此外，许多方法将轨迹生成视为简单的token预测，这可能导致运动学不可行的动作。为了解决这些限制，我们提出了StyleVLA，一个具有物理约束的VLA框架，用于生成多样且物理上合理的驾驶行为。我们引入了一种混合损失，结合运动学一致性约束和连续回归头，以提高轨迹可行性。为了训练StyleVLA，基于Qwen3-VL-4B，我们构建了一个大规模的指令数据集，包含超过1200个场景、76000个鸟瞰视图（BEV）样本和42000个第一人称视图（FPV）样本，其中包含五种驾驶风格的地面真实轨迹和自然语言指令。实验表明，我们的4B参数StyleVLA显著优于专有模型（例如，Gemini-3-Pro）和最先进的VLA模型。使用一个综合驾驶评分，测量成功率、物理可行性和风格适应性，StyleVLA在BEV上达到0.55，在FPV上达到0.51，而Gemini-3-Pro分别为0.32和0.35。这些结果表明，一个专门化、具有物理约束、轻量级的模型可以在领域特定任务上超越专有模型。
英文摘要：
Vision Language Models (VLMs) bridge visual perception and linguistic reasoning. In Autonomous Driving (AD), this synergy has enabled Vision Language Action (VLA) models, which translate high-level multimodal understanding into driving behaviors, typically represented as future trajectories. However, existing VLA models mainly generate generic collision-free trajectories. Beyond collision avoidance, adapting to diverse driving styles (e.g., sporty, comfortable) is essential for personalized driving. Moreover, many methods treat trajectory generation as naive token prediction, which can produce kinematically infeasible actions. To address these limitations, we present StyleVLA, a physics-informed VLA framework for generating diverse and physically plausible driving behaviors. We introduce a hybrid loss that combines a kinematic consistency constraint with a continuous regression head to improve trajectory feasibility. To train StyleVLA, built on Qwen3-VL-4B, we construct a large-scale instruction dataset with over 1.2k scenarios, 76k Bird's Eye View (BEV) samples, and 42k First Person View (FPV) samples, with ground-truth trajectories for five driving styles and natural-language instructions. Experiments show that our 4B-parameter StyleVLA significantly outperforms proprietary models (e.g., Gemini-3-Pro) and state-of-the-art VLA models. Using a composite driving score measuring success rate, physical feasibility, and style adherence, StyleVLA achieves 0.55 on BEV and 0.51 on FPV, versus 0.32 and 0.35 for Gemini-3-Pro. These results show that a specialized, physics-informed, lightweight model can surpass closed-source models on domain-specific tasks.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
