Deep Q-learning (DQN) for Multi-agent Reinforcement Learning (RL)

RL projects including implementation of DQN/DDPG/MADDPG/BicNet on StarCraft II multi-agent learning environment SMAC

Deep Q Network for Multi-agent RL

DQN-based RL controller for multi-agent traffic simulation in SUMO

This project explores Multi-Agent Reinforcement Learning (MARL) in two environments: a custom Drone Delivery system with no-fly zones and package coordination, and Simple Spread v3 (PettingZoo). We implement tabular (Q-Learning, SARSA, Double Q) and deep RL (DQN, Double DQN, Dueling DQN, QMIX) methods, comparing performance and strategies.

No description available

Hybrid AI agents (Behavior Trees + DQN, PPO) for Lux AI Season 3 competition. Multi-agent RL solution for strategic space exploration and combat.

Explored reinforcement learning in-depth, covering key concepts like MDP, TD, Q-learning, SARSA, and MCTS. Advanced into Deep RL with DQN, Policy Gradient, and actor-critic methods (REINFORCE, DDPG, PPO). Explored Partially RL, Multi-agent RL, and Representation Learning. Applied RL in Robotics, engaging in coding exercises throughout the course.

Full-stack RL simulation where a Dueling Double DQN agent learns to schedule workforce tasks in real time, benchmarked against 5 classical heuristics across multi-day simulations. Features Prioritized Experience Replay, shaped reward functions, and a live React dashboard streamed via Socket.IO.

Production-ready Multi-Agent RL Trading System using PyTorch, Stable-Baselines3, Gymnasium & Streamlit. Simulates trading, trains PPO/A2C/DQN agents, tracks ROI, cash, trades, learns multi-stock policies. Ideal for quant prototyping, strategy research, and RL engineering demos!

This project explores the application of reinforcement learning to Ludo, a stochastic multi-agent board game. I implement and compare several RL algorithms including tabular Q-learning, deep Q-networks (DQN), Dueling DQN, and rule-based heuristics.

Udemy course companion for RL mastery: Q-learning, DQN, Policy Gradients, multi-agent systems. Hands-on OpenAI Gym implementations with PyTorch/TensorFlow. Projects include gridworlds, reward shaping, and AI decision-making labs. 🧠🤖

Deep Q-learning (DQN) for Multi-agent Reinforcement Learning (RL)

Decentralized multi-agent RL traffic signal control using Dueling DQN + GAT (CTDE) on a 16-intersection network, reducing avg wait time by 76% and improving travel time by 17%; Dockerized and deployed on AWS.

Advanced multi-agent RL system with DQN+PPO - 39.35-point improvement over 1000 episodes

Multi-agent DQN framework where RL agents proactively pre-install SDN flow rules to reduce control-plane latency

Multi-Agent RL simulation of adaptive traffic signal control and autonomous vehicle coordination using Pygame and DQN.

Multi-Agent Hide and Seek in a 2D grid with DQN-based RL. Agents learn emergent strategies involving door toggling, locking, and hiding.

In this folder, I will add different examples of basic RL algorithms (Q-learning, DQN, Actor-Critic, multi-agent, etc)

RL framework for multi-agent coverage & collision avoidance in 2D gridworlds. PPO/DQN train 1–4 agents; swarm covers 3x faster. Full visuals, GIFs, logs. Python // Gymnasium // SB3

A browser-based multi-agent environment that demonstrates hierarchical tasks, synergy requirements, day/night cycles, and a commander RL agent—powered by TensorFlow.js Dueling DQN. Inspired by “Ontoverse”-style knowledge graph concepts and advanced multi-agent synergy.

A multi-agent reinforcement learning simulator for the Traveling Officer Problem (TOP). Features PettingZoo-compatible environment, greedy/ACO baselines, RL/MARL algorithms (DQN, PPO, QMIX, VDN), multi-objective rewards, and W&B integration.

Application of reinforcement learning methods to optimization of state migration and economic policy. Implementation of Solow Growth Model as multi-agent environment with countries as agents trained via RL-methods (DQN, Replay Buffer, Target Network).

Research-grade Reinforcement Learning framework for single-agent and multi-agent warehouse navigation using Deep Q-Networks (DQN), PyTorch, replay buffer, target networks, logging, and full test suite. Built for PhD-level RL and autonomous systems research.

Multi-agent RL framework for warehouse logistics. Built a custom Gymnasium environment solving pathfinding & collision avoidance for 4 agents. Implemented Tabular Q-Learning & Double DQN with a 15D observation space and specialized reward functions to optimize throughput & safety.

A PyTorch implementation of the Rainbow algorithm for playing Atari games using deep reinforcement learning. This repository contains a Rainbow DQN agent that combines several improvements on the original DQN, including Double Q-learning, Prioritized Experience Replay, Dueling Networks, Multi-step Learning, Distributional RL, and Noisy Nets.

City Guard AI is a multi-agent reinforcement learning game where two AI agents, Shaktiman (DQN) and Heman (A2C), compete to eliminate the maximum number of enemies in a maze-like environment. The project involves building the game environment, training the agents using RL algorithms, and evaluating their performance.

Sliding-window protocol (SACK, fast retransmit, dynamic sizing); RTT estimation via RFC-6298 EWMA; optional DQN RL agent for window prediction; zero-copy I/O (mmap + circular buffers); parallel multi-stream transfers; integrity checks (CRC32 chunks, SHA-256 file); token-bucket rate limiting.