“…Several researchers have recently explored using QPs to control bipedal systems both in simulation [1,17,48,41,39,26] and in hardware [2,32,60,36]. As in our formulation, these optimizations typically employ low-dimensional dynamic quantities such as center of mass (COM) acceleration [1], rate of change of linear and angular momenta [48,32,39], zero moment point (ZMP) [60], or canonical walking functions [2], where the QP cost consists of a weighted distance from a reference value computed using a simple PD control rule [1,48,32].…”