SUMMARY

Jumping is an important behavior for many animals and robots. Unlike periodic gaits such as hopping or running, whereby energy generated in previous cycles can be leveraged to efficiently sustain motion, jumping relies almost purely on a transient burst of activity to produce take-off from rest. While bioinspired robots have utilized some jumping mechanisms revealed from numerous biological studies, there have been few systematic studies of the dynamics of these transient behaviors, particularly on complex media like sand. This dissertation presents a robophysics approach (the pursuit of principles of self generated motion) to systematically study the dynamics of jumping on both hard and deformable ground. For jumping on hard ground, the present work expands on the previous results from Aguilar et al. which characterized the dependence of jumping performance on the robot's hybrid air/ground dynamics, and analyses how relative jumping performance and power requirements of different actuation strategies change at different scales of mass, gravity, stiffness and forcing amplitude. To contrast with the dynamics of jumps on hard ground (in which the unyielding ground supplies the necessary normal force to counteract downward motion), we study a relatively simple deformable medium: dry granular media, which can exhibit both solid and fluid-like dynamics. Through the simultaneous analysis of both the robot and granular dynamics during jumping, our study reveals not only actuation principles crucial to jumping on complex media, but also new granular physics, like an added mass effect induced by a jammed granular cone beneath the robot's foot. Additionally, in collaboration with the Professor Aaron Ames' group at Georgia Tech, we incorporate these granular dynamics into a motion planning optimizer to produce optimal open loop controlled jumps.