Weak actuators generate adaptive animal gaits without a brain

In this study, illustrative experiments were performed to explain the physical aspect of the gait adjustment mechanism for quadrupeds in nature. The major contribution of this work was to show that active walking with autonomous gait adjustment is possible without any explicit controller. Our testbed is capable of generating typical quadruped gaits and selects one of them automatically despite being composed of only a purely physical mechanism with no sensor, brain, or controller. The key idea of this approach is to exploit the weakness of a low-torque DC motor. This passivity takes the function of an oscillator model and delays its own phase. Some fundamental properties of the robot and its gaits with different input voltage were demonstrated in the experiments.