| This paper addresses a technique, which integrates four
balance control strategies and is used on Nao robot to
realize walking on uneven terrain that is not modelled in
advance. The most important two strategies are "Closed Loop
Gait Pattern Generator" and "Posture Control". The former
one uses the filtered robot state based on Kalman filter. It helps to improve joint tracking, which is important for
model based approaches. The latter one helps to make the
trunk vertical to the ground. This strategy is very
effective when walking on a slope. The other two strategies
are "CoG (Center of Gravity) Height Control" and "Ankle
Joint Control", which are used to resist relatively large
tilt and prevent potential falling over.
|