Reproduce “Beyond Fixed Visual History: Learning Motion-Adaptive Temporal Receptive Fields for Complex-Terrain Quadruped Locomotion” 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
Hao Zhang, Yuping Yan, Yitian Hong, Luying Feng, Tianhong Wang, Fengqi Dai, Wei Chen, Yaochu Jin
TempoLoco combines four learned temporal views of depth history, weighting them using current body-state and terrain evidence; sampling strides remain fixed after training. In full-curriculum simulation, success rate—the share of episodes reaching the terrain goal within a ±1.0-meter lateral corridor—was 78.4 ± 2.3%, versus 58.2 ± 4.2% for Dense. This result covers one backbone and a fixed 0.35-meter-per-second command.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.