Limbless Locomotion Control in Unstructured Terrains

Abstract

We are interested in uncovering the fundamental elements that enable limbless locomotors to move in unstructured terrestrial environments. This is challenging because while it is possible to observe the changing shapes of biological limbless locomoters as they move, we know little about the underlying mechanisms that govern their motion. Motivated by this difficulty, this work uses observations of biological behaviors to make hypotheses about the dominant type of control strategies employed by different species of snake locomoting in unstructured environments. Based on these observations, we developed a locomotion control framework that provided the ability to freely change low-level parameters as well as sensing modalities of a model system, a physical snake-like robot. This freedom in selecting how to implement the robot’s control method gave us the ability to effectively hypothesize about and switch between different neurological control architectures. Furthermore, implementing the different architectures on the robot and comparing the results to data from the biological snakes grounded the hypotheses in the physical world. Initial results show that we are able to capture many of the salient kinematic features exhibited by the biological snakes using somewhat simple force control schemes, as well as that compliant force-based controllers in general far outperform stiffer position-based approaches. We thus believe that using a force-based control framework as the basis for generating and comparing further data from the model system to the biological systems will lead to a deeper understanding of how limbless systems move. The physical nature of this understanding will be more intuitive and ultimately more valuable than simply observing or using reduced models to represent biological systems alone.