Reproduce “TRACE: Target-State Risk-Aligned Conditional Execution for Embodied Tracking” 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
Mingyang Lyu, Gongce Cao, Yiyang Jia, Lu Jia, Erliang Lin, Yuhao Zhang, Yinqian Sun, Pin Lyu and 2 more
TRACE selectively scales a robot’s movement at execution time when a risk score is high, while leaving its planner unchanged. In simulated distracted tracking, it improves success rate by 2.03 percentage points and following rate by 2.14 points, while reducing collision rate by 3.96 points versus planner-head PPO. These paired results cover 1,380 episodes and do not establish physical-robot safety.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.