Slipping helps steering in a multilegged robot

Abstract

Multilegged robots offer a potentially useful locomotion platform for both man-made and uncontrolled environments, as well as functioning as models for the scientific study of biological locomotion. Simple models, computable in faster than real time, are needed for design optimization, model based control, and comparative biomechanics. Commonly, models assume a non-slip contact with the substrate. We present an experimental study of hexapedal “steering”: gaits that allow turning while moving forward. Foot slipping is crucial for understanding these gaits in our robot. Surprisingly, increased slipping produces an improved ability to steer. Turning radii nearly halve and non-dimensional turning rates (deg/cycle) grow by 50% as the distance feet slip is doubled. Yet comparison of the foot motions in the body frame suggests that the geometry of leg motion remains virtually unchanged. We conclude that substantial work is needed to better understand and model multi-legged slipping, and relate body configuration changes to motion in the world frame.