| Summary: | 碩士 === 國立雲林科技大學 === 資訊管理系碩士班 === 101 === This paper builds a hybrid genetic algorithm which combines the K-line theory and the Seek-Whence theory with genetic algorithm. Bioloid humanoid robots can use this mechanism to innovate or recall dances. Different from general dancing robots, this mechanism makes humanoid robots learn and recall to create more creative dance from existent dance step database.
This study uses the Bioloid robot developed by Robotis Company as a research platform. In many different kinds of situation, using hybrid genetic algorithm and pure genetic algorithm and compare which mechanism is better at innovating and recalling dances. First is deciding the way of creating target dances and similarity thresholds. There are two ways to create target, one is randomly generate dance steps to combine dance, and another is choosing from
dance step database to combine dance. Moreover, the similarity threshold is divided into four thresholds, including 10%, 30%, 50% and 70%, and using hybrid genetic algorithm and pure genetic algorithm to achieve the target dances and similarity thresholds. After achieving that, it computes the total average times that two algorithms waste in the process and the real similarity which two algorithms achieve. It uses total average time and real similarity to decide which mechanism is better at innovate or recall dances.
This study is using computer simulation experiment. From the experiment result, when the target dance is the same and the dance is combine with the dance steps which is chosen from the dance steps database, hybrid genetic algorithm is good at memorizing dance and it spend
short time in recalling when the similarity threshold is 70%. No matter in single or multiple target dance, when he dance is combine with the dance steps which is chosen from the dance steps database or randomly generate, the pure genetic algorithm’s recalling ability is the same as the hybrid genetic algorithm. But the pure genetic algorithm spend less time than the hybrid genetic algorithm.
|
|---|
