Built this as an experimental platform for large-scale multi-agent combat simulation and multi-agent RL. It uses PyTorch for vectorized simulation, per-agent neural policies, PPO training, checkpoint/resume, telemetry, and a Pygame viewer. The repo has a detailed README and technical docs. I’d especially value feedback on the simulation design, training setup, observability, and overall architecture.
Comments URL: https://news.ycombinator.com/item?id=47423351
Points: 1
# Comments: 0