Vertical jump optimization of robotic leg

Offline trajectory optimization of a robotic leg’s jump over an obstacle, using an iterative linear quadratic regulator (iLQR), is presented. The algorithm itself is modified by formulating the running and final cost as a single function of the system’s outputs while dynamics of the nonlinear system are modeled using the open-source physics engine MuJoCo. A redesign of the robotic leg with a parallel linkage configuration is also presented. The leg is based on the open-source Doggo project of Stanford University and intended for a quasi-direct-drive quadruped. The redesign of some of its mechanical parts focuses on increasing geometric accuracy and durability while, more importantly, reducing overall weight. An experimental setup for testing the vertical dynamics of the redesigned leg’s movement is consequently described.