BriefGPT.xyz
Jul, 2024
HACMan++:空间基础运动原语用于操作
HACMan++: Spatially-Grounded Motion Primitives for Manipulation
HTML
PDF
Bowen Jiang, Yilin Wu, Wenxuan Zhou, Chris Paxton, David Held
TL;DR
在我们的方法HACMan++中引入了具有空间基础的参数化运动基元,以改进策略的泛化能力,并实现了零样本从模拟到真实的迁移。
Abstract
Although
end-to-end robot learning
has shown some success for robot manipulation, the learned policies are often not sufficiently robust to variations in object pose or geometry. To improve the
policy generalization
→