Multi-agent

Description:

• All agents are rational and optimal
• Can be:
  • deterministic or Stochastic Process
  • Zero Sum Game or general game
    • we assume there is a game value where 1 agent wants to minimize and the other maximize
• Perfect vs imperfect information
• A common formalization:
  • State S, starting at $s_0$
  • Players: $P=\{1,2,...,n\}$, typically take turns
  • Actions: $A$
  • Transition Function
  • Terminal Test: $S\to \{t,f\}$
  • Terminal utilities: $S\times P\to R$
• A solution for a player is a Policy: $S\to A$

Solving Zero-sum Game for Multi-agent