Mathematical study of games with memory and many players
A study examines how agents interact in systems where memory effects influence dynamics.
This work studies linear-quadratic mean field games. In these games, the state of each agent follows a stochastic Volterra equation of convolution type.
These dynamics are non-Markovian. This means they account for memory effects through a completely monotone kernel.
The mean field interaction enters both the drift and diffusion coefficients. Researchers use a Markovian lifting of the Volterra dynamics to restore the Markov property.
This lifting moves the system to an infinite-dimensional state space. It allows for a characterization of the optimal response through Riccati equations in Banach spaces.
The study proves existence and uniqueness of the equilibrium on short horizons. For cases where control does not act on the diffusion coefficient, it holds on arbitrary horizons under a monotonicity condition.
The research also considers a finite-player game. It establishes a quantitative propagation of chaos result and an -Nash equilibrium for the -player game.
Key facts
- The dynamics are non-Markovian and include memory effects.
- The mean field interaction affects both drift and diffusion coefficients.
- The equilibrium is unique on short horizons.
- The feedback control yields an -Nash equilibrium as the number of agents tends to infinity.