2022 Thayer Investiture

All Thayer Events

PhD Thesis Defense: Joshua Elliott

Feb

22

Tuesday
11:00am - 1:00pm EST

Videoconference

For info on how to attend this video conference, please email joshua.j.elliott.TH@dartmouth.edu.

"Dynamics Modeling and Control of Multi-Segment Passively-Articulated Autonomous Wheeled Vehicles"

Abstract

The purpose of this thesis is to describe a class of multi-segment, passively-articulated autonomous wheeled vehicles designed for long distance ground transportation across snow surfaces in the polar regions. Uncrewed ground vehicles (UGVs) are currently used in the polar regions for data collection over terrain that would be dangerous for crewed vehicles. The range of snow conditions that may be encountered in the polar regions is vast and hard to fully predict. Of the conditions that can be expected, traditional wheeled mobility is not always sufficient, even with sophisticated methods for immobilization detection and prevention. The obstacles to mobility in the polar regions are unique in that they often give no visual indications of their presence and can only be measured proprioceptively. Therefore, avoiding hazards is less of a focus, and UGVs must instead rely on robust mobility systems that can overcome or prevent imminent immobilizations. The risk of immobilization is one factor preventing greater use of UGVs in long distance applications. This thesis presents one solution to the polar mobility challenge in the form of the SnoWorm UGV. The SnoWorm is a high mobility UGV made up of multiple four-wheeled segments that are passively articulated relative to one another. SnoWorm's long length and other features give it superior mobility on snow surfaces when compared to similar wheeled vehicles. SnoWorm is simulated within the Gazebo robotics simulator using a custom terrain model that provides net force versus slip characteristics comparable to more sophisticated terramechanics models for snow. Data collected from a physical UGV operating on snow is used to ground-truth the terrain model for both steady state driving and immobilizations in poorly-bonded snow. Multiple control algorithms are explored, one allowing the SnoWorm to drive in a linear manner, where segments follow one another, and another allowing lateral "crab" movement that minimizes repeated passes over the same terrain. Measuring the forces between segments gives a better understanding of vehicle dynamics and allows for control schemes to vary the mobility required from individual segments.

While increasing vehicle length and the number of wheels does improve mobility over a variety of snow surfaces, it is not enough to decrease the risk of immobilization to the point where long distance uses of UGVs become feasible. Tracks instead of wheels help, but their much lower efficiency compared to wheels is especially relevant for UGVs covering long distances. Therefore, "inchworming," an alternative mode of locomotion for the SnoWorm, is evaluated in this thesis. Inchworming relies on powered prismatic joints between segments to enable them to push and pull one another. This inchworm technique vastly increases the total range of conditions over which the vehicle is able to make forward progress, with minimal added complexity.

Thesis Committee

  • Laura Ray, PhD (Chair)
  • Mary Albert, PhD
  • Douglas Van Citters, PhD
  • Hari Nayar, ScD


Contact

For more information, contact Theresa Fuller at theresa.d.fuller@dartmouth.edu.