ConceptMarkov Game
Markov Game
The standard formal framework for multi-agent reinforcement learning, described by a tuple consisting of a set of agents, a joint state space, each agent's action space, state transitions under joint actions, each agent's reward function and a discount factor. Each agent's transitions and rewards depend on the joint actions of all agents.
Timeline (last 90 days)
Oct 18
The Markov Game is the standard formalization framework for MARL, usually described by the tuple (N, S, {A_i}, P, {R_i}, γ).
Unverified50%
Oct 18
In a Markov Game, each agent's transitions and rewards depend on the joint actions of all agents.
Unverified50%