Dissertation
Perceptive Locomotion on Bipedal Humanoid Robots for Traversing Unknown and Challenging Environments
University of West Florida Libraries
Doctor of Philosophy (PHD), University of West Florida
2024
Abstract
Humanoid robots possess one of the most favorable morphologies for performing complex tasks in various environments. Disaster response, space exploration, and manufacturing are only some of the tasks that humanoid robots can potentially accomplish. Bipedal humanoid robots rely on legged locomotion involving terrain perception, contact planning, and whole-body control. However, the underactuated dynamics, high dimensionality, and inherent instability render the simplest tasks extremely challenging for bipedal robots, especially in complex environments. Yet, these same complexities afford them the versatility needed to accomplish challenging tasks. In this work, various perception, planning, and locomotion approaches for humanoid robots operating in complex terrain types are designed, implemented, and evaluated. Firstly, novel perception algorithms for graphics processing units (GPUs) to extract planar regions and elevation maps are presented. Extended mapping algorithms for generating persistent models of the environment are then built upon such instantaneous perception algorithms, bringing benefits in footstep planning due to extended planning horizons. Next, a novel Monte-Carlo Footstep Planner (MCFP) that offers graph-reuse and GPU-acceleration capabilities is presented. Perception and planning algorithms are evaluated separately against state-of-the-art bipedal humanoid systems and later combined within continuous locomotion frameworks. Experimental results for the proposed perceptive-locomotion pipeline demonstrate significant speed improvements in traversing rough terrain compared to existing approaches and reach 0.5x human traversal speeds. A terrain complexity metric is also proposed to help benchmark the presented work against previous approaches. Lastly, high-level behaviors are built upon the locomotion system including a dynamic target-following behavior that enables humanoids to detect, track, and follow dynamic entities over rough terrain. Reaching human-level speeds for traversing rough terrain remains a grand challenge for bipedal robots as it spans multiple areas of active research and involves complex software and hardware systems. This thesis presents efforts to close this gap between human and humanoid capabilities.
Details
- Title
- Perceptive Locomotion on Bipedal Humanoid Robots for Traversing Unknown and Challenging Environments
- Resource Type
- Dissertation
- Contributors
- Hakki Erhan Sevil (Committee Co-Chair) - University of West Florida, Intelligent Systems and RoboticsRobert Griffin (Committee Co-Chair)Kristen B Venable (Committee Member) - University of West Florida, Intelligent Systems and RoboticsJerry E. Pratt (Committee Member)Frank Dellaert (Committee Member)
- Publisher
- University of West Florida Libraries
- Format
- pdf
- Number of pages
- 191
- Copyright
- Permission granted to the University of West Florida Libraries by the author to digitize and/or display this information for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires the permission of the copyright holder.
- Identifiers
- 99380569597106600
- Academic Unit
- Intelligent Systems and Robotics
- Language
- English
- Awarding Institution
- University of West Florida; Doctor of Philosophy (PHD)
- Theses and Dissertations
- Doctor of Philosophy (PHD), University of West Florida