Reproduce “Camera equalisation of an 11.5x drop in environmental illuminance: four imitation-learning policies on an SO-101 arm” 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
Simon Cowling
The authors evaluate whether changing room illumination affects a robot’s camera input and pick-and-place success. Four imitation-learning policies attempt the same task on an SO-101 arm under about 520-lux and 45-lux conditions. Success is 20/40 versus 19/40, while camera-image brightness falls only 3% and 19%. Because exposure was automatic, this does not establish robustness to genuinely dark images.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.