S.S. Tabrizi and S. Bagheri (Iran)
Fuzzy logic, Robotics, Biped robot, Expertise extraction,Linguistic control
Controlling dynamic biped robots is a complex challenge.
Almost all researches in this field assume that the internal
structure and dynamics of the robot is known to the
controller. But in real-world problems we don't have
exact model of the robot or this model is very complicated
and is not useful in real-time control. As a matter of fact,
we use many devices without any knowledge about their
internal structure or analytical models. Some researchers
used learning algorithms to control bipeds, but their
performance is not comparable to human's learning.
Fuzzy controllers can control the bipeds with lower cost,
but they need sufficient rules. In the previous works we
proposed a linguistic controller. In order to design the
controller a human with no robotics skills tried to control
a biped using a joystick. Visual information of robot
posture is all the available information to the operator.
Recognizing that the operator became enabled to control
the robot successfully, we extracted his knowledge as 23
fuzzy rules. These rules fed to robot and the robot could
walk autonomously. In this paper we show the robustness
of this controller against rough terrain.