Dynamic simulation of non-programmed gait generation of quadruped robot

This paper describes a bio-inspired gait generation mechanism for quadruped robots. The gait generation mechanism is an adaptation of one the authors previously developed. Our previous work focused on how a quadruped robot system can generate gaits using pulse-type hardware neuron models (P-HNMs), which have functions similar to the biological neurons. A microcontroller mounted on the robot changed the joints’ angle each time the P-HNMs output a pulse. The joints’ angular velocity changed by the pulse period’s variation according to the toes’ pressure. In this paper, the authors imported the previously developed robot’s body into a dynamic simulator and implemented the bio-inspired gait generation mechanism. The robot model’s mechanical properties are the same as the previously developed robot. The degrees of freedom are excluded except for the legs. Each leg has two joints and a force sensor at the end of the leg. The gait generation mechanism separately controls the legs using each toe’s normal force. Instead of using the P-HNMs’ pulse period variation, the authors defined a simple mathematical formula in which the joints’ angular velocity corresponds to the normal force. The authors confirmed that the robot model could actively generate animals’ gaits due to dynamic simulations. The robot model generated the walk gait and the trot gait according to the locomotion speed. These results show that we can generate gaits for quadruped robots by a straightforward method.