Behavior Generation for Mobile Robot Coordinating Simple Behavior
Yasuhisa Hasegawa* and Toshio Fukuda**
*Dept. of Micro System Engineering, Nagoya University Furo-cho, Chikusa-ku, Nagoya 464-8603, Japan
**Center for Cooperative Research in Advanced Science & Technology, Nagoya University Furo-cho, Chikusa-ku, Nagoya 464-8601, Japan
Complex behavior is rarely obtained using unsupervised learning because of the enormous search space. To reduce search space, a hierarchical behavior structure is effective. We proposed the hierarchical behavior controller, which consists of three types of modules – a behavior coordinator, behavior controller, and feedback controller. We also propose a learning algorithm to for a behavior coordinator and behavior controller consisting of subcoordinators and subcontrollers, This algorithm selects a deficient by evaluating each subcoordinator or subcontroller using multiple regression analysis based on previously obtained evaluation values. This reduces the search area and learning time by avoiding trying to tune good subcoordinators or subcontrollers. The hierarchical behavior controller is applied to controlling a seven-link brachiation robot, which moves dynamically from branch to branch swinging its body like a gibbon.
This article is published under a Creative Commons Attribution-NoDerivatives 4.0 International License.