Hierarchical Heuristic Search Techniques for Empire-based Games

Computer games have long been a fruitful and challenging area for the application of AI technologies. The empire-based strategy game is one genre that presents unique challenges to the implementation of an AI-based player due to the extremely large search space. Consequently, a majority of AI systems utilize ad hoc strategies such as hand-built finite state automata. While practical, this approach limits the computer player to the strategies designed by the programmer. A more flexible approach allows the computer to adaptively search for promising moves. This paper proposes a hierarchical architecture that breaks the problem space into manageable units. Through aggressive pruning strategies heuristic search may be employed across the hierarchical levels. The new search space examines approximately (nm) states for m moves, n pieces, and a search depth of d. In comparison, full search requires the examination of approximately (m) states. This technique was implemented on a simple empire-style game that generated expected results