| Summary: | 碩士 === 東海大學 === 資訊工程學系 === 98 === Abstract
Computer games have highly interactive ability and can integrate various media. Playing computer games have become people’s popular entertainment. Computer games have truly image and sound effect can give the player a rich game experence. But the technology of computer graphics alreay reach a bottleneck in the recent years. So many computer games developers have paid their attention to the AI of game characteristic. They hope the smart and various AI can make the computer game more interesting.
The most computer game today use the rule-base design approach because of the simple and easy to implement. But, if the player find the weak point of the computer character, nothing can stop the player to win the game. If the computer character can learn from mistake, there may be a solution of this problem. Some scholar try to implement some learning algorithm to the computer game. But we found it needs large computation and collecting the train data sometimes are difficult. And we found the team work is easily to be found in today’s computer games. So we try to give a new approach to the team play computer game’s AI.
For the application of computer game AI, the computation must be quick and stable. Multi-Objective Bees Algorithm（MOBA） is a new optimization and machine learning technology in Artificial Intelligence. MOBA is easy to emplement and there are few parameters to adjust. So we try to implement the MOBA as the learning algorithm of computer game characteristic. But according to MOBA, there is no coordination between each bees. So it can only create a powerful single character. So we propose a new learning strategy placing the emphasis on the team learning.
In summery, this paper proposes a novel method based on MOBA to help behavior design in computer games. Proposed method can create more efficient team. And there is no need of large computation and training date, which suit the application of computer game. This new mechanism can help AI developer adjust the behavioral parameters which can save the testing time of different combination of parameter. In the experimental results, the proposed mechanism was embedded to design the team bots that indeed presents more changeable and the stable learning characteristic in the Quake III single player mode : Free for Play.
Keywords: Artificial Intelligence, Multi-Objective Bees Algorithm, First-Person Shooter
|
|---|
