Autonomous Behavior-Mode Switching in a Two-Mobile Robotic System. S-Temperature Based On-line Learning.
Tsutomu Sawada, Sumiaki ICHIKAWA, Fumio HARA · JSME International Journal Series C · 2002
This paper deals with an autonomous behavior-mode switching in a behavior-based learning of a mobile-robotic system, which is composed of two mobile robotic modules consisting of a crawler and one DOF arm. The task is to carry a long bar horizontally along a compound road with keeping the bar-support length constant. We propose an on-line learning method “STL (S-Temperature based on-line Learning)”. During the learning process, the robotic system can build up the state categorization in terms of two-dimensional sensory data (support length and support inclination). The robotic system can transit seamlessly between the learning behavior and the task-achieving behavior according to the internal state of the robotic system that is represented by S-temperature introduced into the STL algorithms. We show the effectiveness of STL algorithm by hardware experiment and the learning characteristics with respect to important learning parameters by computer simulation experiment.