A Computational Framework for Designing Skilled Legged-Wheeled Robots

摘要

Legged-wheeled robots promise versatile, fast and efficient mobile capabilities. To unleash their full potential, however, such hybrid robots need to be designed in a way that promotes the complex, full-body motions required for novel locomotion modes. This letter discusses the computational framework we have used to create a new type of legged robot which, when equipped with different types of end effectors, is capable of an array of interesting locomotory behaviors, including walking, roll-walking, roller-blading, and ice-skating. We show that this computational framework, which builds on a design system we recently introduced in the computer graphics community [1], can accurately predict the way in which different design decisions affect the robot's ability to move, thus serving as an important tool in engineering new types of mobile robots. We also propose a novel warm-starting method which leverages ideas from numerical continuation to drastically improve convergence rates for the trajectory optimization routine we employ to generate optimal motions.