ABSTRACT
This chapter looks at the AI used to implement characters in the game Project Highrise by SomaSim, a skyscraper construction management sim where players build, lease out, and manage a series of highrise buildings. The player's job in Project Highrise is to invest in the construction of highrise buildings, and then manage them successfully. Designers gave themselves a concrete performance goal: The game needed to be able to simulate and display 1000 NPCs, running at 60 FPS on a reasonably recent desktop-grade personal computer. Action selection is a fundamental decision process in character AI. The planning system was the first stab at action selection for NPCs. Many NPC AI systems are closed-loop feedback systems—they monitor the world while actions are performed, and adjust behavior appropriately, primarily so that they can handle failures intelligently. Some architectures like subsumption or teleoreactive trees accept constant resensing and recomputation as the cost of doing business—while various behavior tree implementations.
