This lecture presents an introduction to basic real-time
control
methods for bipedal humanoid robots. The lecture will contain three
parts: 1) modeling, 2) balance control, 3) walking control.
In the
first part the lecture will start with a discussion of modeling issues
and highlight the properties of system models of different levels of
abstraction.
Based on complete multi-body dynamics models, the
problem of humanoid balancing control will be discussed in the second
part. Different strategies for either position controlled or force
controlled robots will be discussed.
The third part of the lecture
will focus on walking control. For this problem the use of different
simplified control models will be analyzed.
The lecture will also
cover classical ZMP based walking control using pattern generators, but
the main focus will be put on gait generation based on a state feedback
control approach. Therefore, the capture point will be used as a system
coordinate and the walking control problem will be redefined as a
control problem for the capture point. It will be shown that the use of
the capture point simplifies not only the feedback control problem, but
also the motion planning problem.
For the algorithms presented in this talk, practical implementations on
the humanoid robot TORO will be shown.
Universität Bonn, Institute for Computer Science, Autonomous Intelligent Systems