Reproduce “KlaskTron: An Open-Source Platform for Physical Adversarial Multi-Agent RL” 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
Aswin Ramachandran, Jona Schulz, Maurus Derungs, Carlo Angelini, Tobias Meier, Raffaello D'Andrea
KlaskTron is an open, desktop-scale testbed for adversarial two-agent learning. Its magnetic gantries move strikers beneath a KLASK board, and simulated training uses a learned model of real actuator response. The model achieved a validation RMSE of 0.013 m/s and enabled zero-shot policy transfer. The PPO baseline won one of ten games against a recreational human, not establishing human-level play.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.