Abstract
Today's humanoid robots are controlled typically by "highly precise control algorithms": their joint angle trajectories are determined accurately at any time with vast amount of computation. This approach, however, causes serious problems, particularly in terms of energy efficiency and adaptability. On the other hand, recently, another extreme approach has been gaining a lot of attention. A good instantiation is the passive dynamic walker. One of the crucial points is that this approach has shed light on the fact that control system is not everything. More specifically, it is of highly importance to exploit intrinsic dynamics of a robot's body, leading to "cheap" control algorithms. In light of these facts, this paper deals with the transition between quadruped walking and running, particularly focusing on the stiffness of joints.
