Reproduce “Controlling the Humanoid Robot H1-2 With a Simulated Biologically Plausible Network” 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
Fatima Atwi, Constant Roux, Christian Bergaud, Olivier Stasse
The authors train a culture-inspired spiking network to balance a simulated humanoid using e-prop, which combines neuron-specific learning signals with local traces of incoming connections. Training uses dense expert action labels at each timestep. The network achieves 100.0% ± 0.0% survival over 10-second simulation episodes, but this measures balance, not command tracking; every physical robot trial ends in a fall.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.