Correlated Equilibrium

Correlated equilibrium (CE) is a solution concept in game theory that generalizes Nash equilibrium by allowing players to coordinate their actions through a shared, publicly known correlation device. Current research focuses on developing efficient algorithms for finding CEs in various game settings, including Markov games and extensive-form games, often employing techniques like optimistic-follow-the-regularized-leader (OFTRL), online mirror descent, and neural network-based approaches. These advancements are significant because CEs can lead to improved social welfare compared to Nash equilibria, and efficient algorithms are crucial for applying CE to real-world multi-agent systems such as traffic control and multi-agent reinforcement learning. Furthermore, research explores robust versions of CE to handle uncertainties and incorporates additional constraints like safety and fairness into the equilibrium selection process.