ctde

Code for Backpropagation Through Agents (BPTA/BPPO, AAAI 2024) and AgentMixer: Multi-Agent Correlated Policy Factorization (AAAI 2025). Unified PyTorch implementations for cooperative multi-agent reinforcement learning.

Multi-agent RL for wastewater treatment plant control (BSM2/Simulink) — MSc Thesis

3D battle arena with 9 agents (3 teams x 3) trained with PPO + PyTorch, comparing six MARL information-flow architectures: CTE, DTE, CTDE, and three CTDE variants (value decomposition, centralized actor-critic, learned communication).

for: Scalable Collision Avoidance in Large LEO Constellations: A Shared-Policy Multi-Agent Reinforcement Learning Framework with Locality-Constrained Coordination

État de l'art du Multi-Agent RL avec RLlib (Ray 2.56, new API stack) : 13 pages en français, 6 exemples exécutés, 3 expériences mesurées.

Cooperative multi-agent controller for Pokémon Showdown multi battles with CTDE training, intent exchange, benchmarks, and replay analysis.

AeroRider: MAPPO/CTDE multi-agent RL for drone-and-rider delivery dispatch. Built-in FFD/BFD/GA/PSO baselines, 3 scale presets, collision avoidance, full pytest suite.

Reproducible multi-agent AI lab with matched-seed arenas, honest robustness audits, CTDE and PPO.

Multi-agent elevator dispatch with QR-DQN and CTDE (Centralised Training, Decentralised Execution) in PyTorch. Three cooperative elevators trained with dueling networks, prioritised replay, and a centralised critic — beating ETA-Dispatch by 27% wait and 9.6% energy across 3,500 episodes on a non-uniform 10-floor traffic environment.

Multi-Agent Reinforcement Learning for Adaptive TCP congestion control