MountainAI
Войти
rlmulti-agentadvanced

Multi-agent reinforcement learning

MARL: independent learners, centralised training, QMIX, MAPPO — cooperation and competition among RL agents.

Уровни глубины

L0Intro~2ч

Knows "multiple agents, same environment" differs from single-agent RL.

L1Basics~10ч

Trains independent Q-learners on Pong 2-player; understands non-stationarity.

L2Working~20ч

Implements centralised-training / decentralised-execution (MADDPG, QMIX, MAPPO).

L3Advanced~30ч

Stochastic games; Nash / correlated equilibria; population-based training.

L4Research~60ч

Opponent modelling, emergent communication, large-scale MARL benchmarks (SMACv2).

Ресурсы