RECAST: From Log Replay to Closed-Loop Driving Simulation with View-Complete Actors

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 $F{D}_{incep}$ from 9.788 to 7.992 relative to unadapted TRELLIS. At the scene level, under actor motion beyond logged trajectories, RECAST reduces $F{D}_{incep}$ from 129.35 to 112.10 and increases $CLI{P}_{margin}$ ($\times 1000$) 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/