reinforement-learning

Standardized environment infrastructure for Agentic AI development.

Learning to Design City-Scale Transit Routes

This project implements a RL based PID tuning system for motor position control using an ESP32 microcontroller. It leverages micro-ROS with ROS2 ecosystems, allowing real-time PID tuning and motor control through ROS2 action servers.

Enables you to convert a PettingZoo environment to a Gym environment while supporting multiple agents (MARL). Gym's default setup doesn't easily support multi-agent environments, but this wrapper resolves that by running each agent in its own process and sharing the environment across those processes.

A Reinforcement Learning (Q-Learning) project that optimizes traffic flow at a 4-way intersection. The agent learns to control traffic lights to minimize total queue length and average wait times.

Learning cartpole swing-up control using Soft-Actor-Critic method

Fetch and orchestrate cryptocurrency data. Train RL model to sim trading activities.

SFT overtraining collapses output entropy, squashing GRPO gradient signal; we derive the exact failure condition, show standard checkpoint selection inverts quality ranking, and propose a 2-stage diagnostic requiring no RL compute.