abstract
In this project a multi-agent system consisting of two agents competing in tic-tac-toe was implemented. The system was developed in AgentSpeak using Jason. Two agent types were implemented; a random agent playing random moves and a strategic agent using the Mini-max algorithm to calculate it's moves. The strategic agent was tested against both a random agent and against a strategic agent using the same strategy. These results were compared to the work of Govind G Nair, who uses reinforcement learning to make strategic moves. The Mini-max agent performs optimally when playing against an equal opponent. Further, when the opponent is a random agent the reinforcement learning agent achieves a higher winrate than the Mini-max agent. However, the reinforcement learning agent sometimes looses to the random agent, while the Mini-max agent always wins or ties.
outcomes