Preview Control-based Jumping and Spot-Jogging Trajectory Generation for Quadruped Robots.

This work presents a vertical center of mass (CoM) trajectory generation approach based on preview control for quadruped robots to accomplish agile locomotion tasks such as jumping and jogging. Conventionally, preview control is implemented to walking planning with constant CoM height, since this condition renders the pendulum equation linear. In this matter, we show that the same implementation can be done for varying CoM height, while the pendulum frequency is constant. To this end, we compactly derived the infinite horizon and finite horizon time-invariant linear quadratic regulator solutions for potential real-time applicability. The proposed trajectory generation approach was tested and compared with conventional approaches via a set of simulation experiments, where we used a realistic model of our quadruped robot Kara. As a result, we obtained dynamically-balanced, coherent, and continuous jumping and spot-jogging behavior, adequately confirming our approach.