- 學位論文
Potential-Based Pattern Formation for Real-Time Strategy Game
Potential-Based Pattern Formation for Real-Time Strategy Game
並列摘要
Real-time strategy (RTS) game is a popular genre of computer games. It is a military simulation where players developing their armed force attempt to destroy each other's units and infrastructure (e.g., directed base, produced factories, and resource gathering buildings) in order to be the (last) surviving player. The units need to navigate their environment, surround the enemy, and attack targets in a battle. Pattern formation of the units plays an important role in the tactics of RTS games. In this paper, we propose two different methods: potential field and designed Lennard-Jones potential, which will generate the pattern formation with collision avoidance, where each unit considered as an independent robot has a limited communicating range, and therefore there is only local information available. The proposed methods control the movement of units and adjust their formation according to the number and types of targets. Once a robot detects the target, it will move in accordance with other robots within its communicating range in order to surround the targets in an attacking distance. The simulation results on StarCraft show the capability of the different proposed methods to generate pattern formation and adapt to the number and types of enemies. According to these results, we will compare potential field with designed Lennard-Jones potential and give some comments of difference between these methods.