bhargav141223/warehouse-multi-agent-rl-using-mappo

A codebase for multi-agent warehouse navigation that trains cooperative policies with Multi-Agent Proximal Policy Optimization (MAPPO) enhanced by LLM-based reward shaping and a Retrieval-Augmented Generation (RAG) memory layer. Intended for researchers and engineers working on multi-agent RL, warehouse automation, and human-in-the-loop reward design. Key features: - Simulated multi-agent warehouse environment and task scenarios - MAPPO training and evaluation pipelines for synchronized cooperative policies - LLM-in-the-loop reward shaping to incorporate high-level guidance into scalar rewards - RAG memory to store and retrieve episodic/contextual information for agents - Utilities for checkpointing, evaluation, and visualizing agent behavior