Abstract

Designing an Intelligent Agent for Eurorails

Student: Robert Warton (University of Findlay)
Mentor: Sean McCulloch (OWU Department of Mathematics & Computer Science)

Eurorails is a modern turn based strategy board game where the players build railroad track by drawing with crayon on the board to connect mileposts, and then traverse it with their trains to pick up and drop off goods. We have created a digital representation of the game board as a network, as well as built the structure around it to enforce the rules of the game. Using this, we have created an agent to efficiently evaluate various possible plans for playing the game and select the best strategy.

Classic board games such as chess have been the main focus of Artificial Intelligence work, while comparatively little has been done for modern board games. These games differ in many ways from classic board games, specifically in that they often have more options in a given turn, giving them a high branching factor making traditional A.I. techniques infeasible. We have created an intelligent agent with these limitations in mind for Eurorails, a popular modern rail-building board game, in order to create an algorithm to efficiently come up with a viable strategy. The game consists of nodes that represent either a city or terrain type, which we represent as the vertices of a graph, and train track the player builds that connect these nodes, represented by connections on our graph. We created a system to enforce the rules of the game on top of this graph that represents the game board. Additionally, we came up with a heuristic to look at possible strategies in laying out our rail and analyze them to choose the best one. We also created a heuristic to intelligently prune out options that are likely to be bad in order to reduce the number of possible strategies we have to consider. This resulted in an efficient algorithm for finding a good strategy in a reasonable amount of time.