arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2604.01129cs.CV
收藏

ReinDriveGen:强化学习后训练用于分布外驾驶场景生成

ReinDriveGen: Reinforcement Post-Training for Out-of-Distribution Driving Scene Generation

MMLab, CUHK(MMLab,香港中文大学);CASIA(中国科学院自动化研究所);SenseTime Research(商汤科技研究院)

Hao Zhang, Lue Fan, Weikang Bian, Zehuan Wu, Lewei Lu, Zhaoxiang Zhang, Hongsheng Li

AI总结:

ReinDriveGen通过动态点云场景生成与强化学习后训练,实现对驾驶场景的可控生成,提升安全关键场景的模拟质量。

摘要:

我们提出了ReinDriveGen,一个能够对动态驾驶场景实现完全可控的框架,允许用户自由编辑演员轨迹以模拟安全关键场景,如前车碰撞、漂移车辆、失控车辆、行人闯红灯和骑车人越线。我们的方法从多帧LiDAR数据构建动态3D点云场景,引入车辆补全模块从部分观测中重建完整的360度几何结构,并将编辑后的场景渲染成2D条件图像,引导视频扩散模型合成逼真的驾驶视频。由于此类编辑后的场景不可避免地超出训练分布,我们进一步提出基于强化学习的后训练策略,结合配对偏好模型和配对奖励机制,在无地面真实监督的情况下实现稳健的质量提升。大量实验表明,ReinDriveGen在编辑驾驶场景上优于现有方法,并在新颖的视角合成上取得最佳结果。

英文摘要:

We present ReinDriveGen, a framework that enables full controllability over dynamic driving scenes, allowing users to freely edit actor trajectories to simulate safety-critical corner cases such as front-vehicle collisions, drifting cars, vehicles spinning out of control, pedestrians jaywalking, and cyclists cutting across lanes. Our approach constructs a dynamic 3D point cloud scene from multi-frame LiDAR data, introduces a vehicle completion module to reconstruct full 360° geometry from partial observations, and renders the edited scene into 2D condition images that guide a video diffusion model to synthesize realistic driving videos. Since such edited scenarios inevitably fall outside the training distribution, we further propose an RL-based post-training strategy with a pairwise preference model and a pairwise reward mechanism, enabling robust quality improvement under out-of-distribution conditions without ground-truth supervision. Extensive experiments demonstrate that ReinDriveGen outperforms existing approaches on edited driving scenarios and achieves state-of-the-art results on novel ego viewpoint synthesis.

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。

↑