One of the primary advantages of walking machines is their inherent ca
pacity for moving over different terrains. However it is important to
provide algorithms that modify the gait according to the terrain. Exis
ting articles about terrain-adaptive locomotion are based on intellige
nt foothold selection, and use periodic and/or aperiodic gaits. This a
rticle proposes a strategy to adapt walking robot locomotion to an irr
egular terrain in real time that is based on the variations of paramet
ers of a periodic gait affecting leg coordination. Its main feature is
that it does not require knowledge of footholds. In addition, this ad
aptive gait control can be incorporated into a system with external st
ereoceptive sensors to select footholds. As a working example, a parti
cular class of periodic gaits, called wave-crab gaits, are used in a q
uadruped robot. However, the proposed adaptive-gait method can be used
for any other periodic gait and for robots with a greater number of l
egs. Adaptive-gait control has been implemented in a quadruped robot n
amed RIMHO, demonstrating its ability to move over different terrains.
Static stability results from computer simulations and experimentally
obtained velocity results are also presented.