abstract
The goal of this project is to develop a Multi Agent implementation of the famous game called Werewolf. This game is based on the rivalry between two factions of agents (i.e. farmers and wolves) who challenge each other during the development of each game in order to eliminate as many agents as possible from the opposing faction. This game contains multiple components common to multi-agent systems: there is an advanced communication mechanism associated with each faction, which also contains different privacy policies for each of these. Furthermore, each agent is independent and simultaneously implements both the principle of cooperation with respect to the agents of his faction and the principle of competition with respect to the agents of the opposing faction. In addition to this, we have also implemented several elements of this game in the form of ”artifacts”, mainly represented by the ”perks” available to each agent. Finally, each agent has been provided with an ”intelligence” mechanism through the structuring of a Bayesian network capable of representing the probabilities associated with each game event throughout its continuation. The results presented were obtained from the simulation of multiple games based, in turn, on multiple different configurations relating both to the game and to the strategies associated with each agent. The results themselves highlighted the correctness of the completed implementation, highlighted the differences with respect to the possible game policies associated with each agent and, at the same time, brought to light the effects of the artifacts with respect to the game strategy of each agent.
outcomes
