Comparison of different properties of multiagent systems in a simulated environment

The thesis presents the simulation environment GridLand, where we simulate two multi-agent systems, that are competing for resources. We have developed an agent that can operate in a multi-agent system and has various properties. These properties have an affect on its actions. We examined and implemented two search algorithms - Dijkstra and A*, which the agent can use for discovering resources and navigating in the environment. The agent's activity was tested in the simulation environment, we observed how specific properties of the agent effected the final simulation result. We were interested in whether the communication has an influence on the result and how this influence is reflected by changing the radius of communication. We also measured the influence of the number of agents in the system, the search method and the attacking strategy. We examined the simulation results and evaluated the importance of each property on the basis of the quantity of gathered resources. We have confirmed the hypothesis that increasing the number of agents and the radius of communication has a positive influence on an agent's performance