Initial investigation into bipedal turning: A trajectory optimization study

Abstract

Humans and animals leverage agility to negotiate the unpredictable environments we occupy. In order for legged robots to leave sterile lab environments, they need to be agile enough to negotiate our lives. Currently, human agility is far superior to the state-of-the-art robotic platforms. Replicating this on robotic platforms require a profound understanding of how contact events are leveraged to complete agile tasks. In line with this aim, this letter was an initial investigation into bipedal turning, to gain insight into how turning was achieved, and to identify any kinematic trends that emerged from the optimization results. This research was conducted on a simulated 10 DoF non-planar bipedal platform with point feet, and made use of a realistic friction cone, and not a linearized approximation. The mathematical model used was based on the bipedal robot currently under development. Two experiments were conducted: rapid turns with a fixed turn angle at varying speeds, and rapid turns with varying turn angles at a fixed speed. Results indicated that slip occurred 93.32% of the contact duration, and turn overshoot was present in all trajectories analyzed. Additionally, a long-time-horizon trajectory was presented to motivate the feasibility and stability of the turn trajectories studied.