Real-time strategy games pose several difficult questions to developers from an AI perspective. Arguably the most important aspect of any strategy game is navigation. Navigation may be split up into higher and lower level tasks. The higher level tasks include finding a path from the start to the destination, and a lower level task could handle how it should be traversed. While a unit is navigating on its path, it will most likely encounter other units along the way, and it should avoid collisions with them if possible.
This paper intends to address at least some of these problems. AI may be vast, but it requires simple, yet powerful tools to be developed and used throughout a game's development. In this paper I extended and implemented a robust velocity based solution for collision avoidance.
This solution was also used for a real-time strategy game I designed in a game engine. The game is designed to have a simple yet extensible architecture while encapsulating diverse behaviors like resource harvesting, building structures, and attacking. The AI is hierarchical with each unit having a controller, and an enemy bot player simulating a human.