In this project, we leverage reinforcement learning techniques to replicate and extend the seminal experiment conducted by Keser in 1992. 

In the initial phase, we employ various reinforcement learning models to train autonomous agents for participation in the well-established multi-round duopoly pricing game, originally introduced by Selten in 1965. This game serves as a standard benchmark in game theory and economic research.

The second phase of our study involves a Double-Oracle framework, where the trained agents assume the roles of strategies in an evolutionary game. These agents engage in dynamic interactions, and new agents are trained to adapt and compete against the Nash equilibrium strategies identified within the game. This twofold investigation contributes to a deeper understanding of strategic behavior and learning dynamics in competitive settings.

In this project, we studied general non-cooperative 2x2x2 games, establishing an upper bound on the number of Nash equilibria for non-degenerate cases. 

Moreover, we introduced an algorithm capable of computing all equilibria for these games, not restricted to generic cases. I implemented this algorithm in Python, enabling the computation and visualization of best-response surfaces and Nash equilibria in 3D space. The resulting code is set to be integrated into a publicly accessible web-based software designed for automated equilibrium analysis.

This project is dedicated to modeling the Facility Location problem in a planar space for two competing firms, modeling it as a strategic game. The core objective involves implementing an algorithm to determine the Nash Equilibria – essentially, the ideal locations for the firms' shops. The key approach involves computing Weber points derived from partitions of the demand set. 

This algorithm is implemented in MATLAB.