发表机构
Global Institute of Future Technology, Shanghai Jiao Tong University; School of Automation and Intelligent Sensing, Shanghai Jiao Tong University（上海交通大学全球未来技术学院; 上海交通大学自动化与智能感知学院）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
RECAST利用3D高斯泼溅从单个车辆观测生成视图完整参与者，实现超出日志回放的闭环驾驶仿真，显著提升无碰撞率和碰撞时间。
AI 中文摘要
闭环驾驶仿真要求当自车和周围参与者超出其记录轨迹移动时，渲染的观测仍然可靠，从而暴露源日志中不存在的视图。现有的数据驱动仿真器从稀疏观测中重建动态参与者，这在这些视角变化下可能导致渲染伪影。我们引入RECAST（重建可控参与者用于仿真与测试），这是一个3D高斯泼溅框架，能够从驾驶日志中的单个分割车辆观测生成视图完整的参与者，并将生成的参与者注册到重建场景中。RECAST支持在受控自车-参与者交互下的规划器在环渲染。为了将图像到3D先验适应于真实车辆，我们进一步引入RECAR，一个包含约2万真实车辆、60万张无背景RGBA图像的数据集，涵盖多样化的车辆颜色和类型。我们使用两阶段适应来改进从真实驾驶日志观测中的车辆生成。在参与者层面，RECAST将FD_incep从9.788降低至7.992，相对于未适应的TRELLIS。在场景层面，在参与者运动超出记录轨迹的情况下，RECAST将FD_incep从129.35降低至112.10，并将CLIP_margin（×1000）从0.14提升至3.47，相对于Street Gaussians。我们使用图像条件规划器GTRS-Dense展示了规划器在环仿真。与原生Street Gaussians参与者相比，RECAST将无碰撞（NC）率从22.2%（12/54）提升至63.0%（34/54），并将平均最小预测碰撞时间（TTC）从0.798秒提升至2.150秒。这些实验表明，RECAST支持在超出日志回放的控制自车-参与者交互下的闭环规划器评估。访问我们的项目页面：此https URL
英文摘要
Closed-loop driving simulation requires rendered observations to remain reliable as the ego vehicle and surrounding actors move beyond their recorded trajectories, exposing views absent from the source log. Existing data-driven simulators reconstruct dynamic actors from sparse observations, which can result in rendering artifacts under these viewpoint changes. We introduce RECAST (REconstructing Controllable Actors for Simulation and Testing), a 3D Gaussian Splatting framework that generates a view-complete actor from a single segmented vehicle observation in a driving log and registers the generated actor in the reconstructed scene. RECAST supports planner-in-the-loop rendering under controlled ego-actor interactions. To adapt an image-to-3D prior to real vehicles, we further introduce RECAR, a dataset of approximately 20K real vehicles with 600K background-free RGBA images spanning diverse vehicle colors and types. We use two-stage adaptation to improve vehicle generation from real driving-log observations. At the actor level, RECAST reduces $\mathrm{FD}_{\mathrm{incep}}$ from 9.788 to 7.992 relative to unadapted TRELLIS. At the scene level, under actor motion beyond logged trajectories, RECAST reduces $\mathrm{FD}_{\mathrm{incep}}$ from 129.35 to 112.10 and increases $\mathrm{CLIP}_{\mathrm{margin}}$ ($\times1000$) from 0.14 to 3.47 relative to Street Gaussians. We demonstrate planner-in-the-loop simulation with the image-conditioned planner GTRS-Dense. Compared with native Street Gaussians actors, RECAST increases the no-collision (NC) rate from 22.2% (12/54) to 63.0% (34/54) and the mean minimum predicted time-to-collision (TTC) from 0.798 s to 2.150 s. These experiments show that RECAST supports closed-loop planner evaluation under controlled ego-actor interactions beyond log replay. Visit our project page at https://zijunkr.github.io/RECAST/
Comments8 pages, 5 figures
