Some of the material in is restricted to members of the community. By logging in, you may be able to gain additional access to certain collections or items. If you have questions about access or logging in, please use the form on the Contact Page.
Animals have demonstrated the capacity to traverse many complex unstructured terrains at high speeds by utilizing effective locomotion regimes. Motion in difficult and uncertain environments have only seen partial success on traditional wheeled or track-based robots and is limited to slow deliberative maneuvers on legged robots, which are focused on maintaining continuous stability through proper foothold selection. While legged robots have demonstrated successful navigation across many complex surfaces, motion planning algorithms currently fail to consider the unique mobility characteristics that honor the natural self-stabilizing dynamics of gait-based locomotion such as running and climbing. This dissertation outlines some of the specific motion planning challenges faced when attempting to plan for legged systems with dynamic gaits, with specific instances of these demonstrated by four robots, the dynamic running platforms: XRL, LLAMA, Minitaur and the dynamic climbing platform TAILS. Using a unique implementation of Sampling Based Model Predictive Optimization (SBMPO) designed expressly for dynamic legged robots, we demonstrate the ability to learn kinodynamic models, motion plan through obstacles on varied terrains and demonstrate navigation on vertical walls. This research has pioneered the technique which allows dynamic legged robots to navigate while honoring the natural dynamics of robot gait. Further, this document will describe to the reader the methods and algorithms that enabled Florida State University to be the first in the world to demonstrate motion planning on a dynamic climbing robot. This work is demonstrated in simulation and verified through hardware experiments on canonical motion planning scenarios, controlled laboratory settings and in unstructured terrains. Finally, this work has opened the field of dynamic legged robot intelligence for future researchers by enabling fundamental navigation and planning, efficient real-time algorithms for onboard computing, and the development of techniques to account for complex constrained motions unique to individual robots and terrains.
A Dissertation submitted to the Department of Scientific Computing in partial fulfillment of the requirements for the degree of Doctor of Philosophy.
Includes bibliographical references.
Gordon Erlebacher, Professor Co-Directing Dissertation; Emmanuel Collins, Professor Co-Directing Dissertation; Paul Beaumont, University Representative; Jonathan Clark, Committee Member; Sachin Shanbhag, Committee Member; Anke Meyer-Baese, Committee Member.
Florida State University
Harper, M. Y. (2018). Learning and Motion Planning for Gait-Based Legged Robots. Retrieved from http://purl.flvc.org/fsu/fd/2018_Fall_Harper_fsu_0071E_14735