arXiv cs.RO 机器人学· Siyuan Liu, Miao Li, Haibao Yu, Haohong Lin, Qing Zhou, Bingbing Nie, Ding Zhao·· 2 天前AI 评分56
ControlPed 生成可控照片级危险行人场景用于端到端驾驶安全评测
Controllable and Photorealistic Pedestrian Risky Motion Generation for End-to-End Driving Safety Evaluation
AI 导读
Siyuan Liu 等提出 ControlPed 框架，将轨迹级冲突合成与 3D Gaussian Splatting 结合，生成照片级、运动可控的车-人危险交互场景。
来源：arXiv cs.RO 机器人学 · arxiv.org