Reproduce “VOKE: Visual Object Kinematics Estimator for Dexterous In-Hand Manipulation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Shengjia Zhu, Wenqiang Xu, Lin Shao
VOKE estimates an object's linear and angular velocity and acceleration from temporal features, then uses estimates to refine pose and inform control. Across DexYCB, OakInk, and simulation data, it lowers errors for all four quantities than smoothed pose differencing; on DexYCB, angular-acceleration MEE falls from 32.997 to 19.832 rad/s². Simulation results show policy gains, while physical deployment is qualitative.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.