Reproduce “Memory-State Critic for Asymmetric Actor-Critic with Application to Vision-Based Pursuit-Evasion” 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
Arthur Louette, Alejandro Sánchez Roncero, Gaspard Lambrechts, Pascal Leroy, Julien Hansen, Petter Ögren, Damien Ernst
The authors introduce a memory-state critic for recurrent agents under partial observability. During training, it combines privileged state with actor memory, avoiding a separate critic-side history encoder; value-loss gradients do not update the actor encoder. In simulated quadrotor pursuit, it reaches the history-state critic’s final return in fewer environment steps in both arenas, but evidence is limited to this task.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.