Title: Long-duration robot autonomy: from control algorithms to robot design

Gennaro Notomista

Robotics Ph.D. Candidate

School of Mechanical Engineering

Georgia Institute of Technology

Date: Monday, July 20, 2020

Time: 1:00 PM - 3:00 PM (EST)

Location: BlueJeans meeting https://bluejeans.com/286726877

Committee:

Dr. Magnus Egerstedt, School of Electrical and Computer Engineering, Georgia Institute of Technology (advisor)

Dr. Wayne Book, School of Mechanical Engineering, Georgia Institute of Technology

Dr. Samuel Coogan, School of Electrical and Computer Engineering, Georgia Institute of Technology

Dr. Seth Hutchinson, School of Interactive Computing, Georgia Institute of Technology

Dr. Anirban Mazumdar, School of Mechanical Engineering, Georgia Institute of Technology

Dr. Mac Schwager, Department of Aeronautics and Astronautics, Stanford University

Abstract:

The transition that robots are experiencing from controlled and often static working environments to unstructured and dynamic settings is unveiling the potential fragility of the design and control techniques employed to build and program them, respectively. A paramount of example of a discipline that, by construction, deals with robots operating under unknown and ever-changing conditions is long-duration robot autonomy. In fact, during long-term deployments, robots will find themselves in environmental scenarios which were not planned and accounted for during the design phase. These operating conditions offer a variety of challenges which are not encountered in any other discipline of robotics.

This thesis presents control-theoretic techniques and mechanical design principles to be employed while conceiving, building, and programming robotic systems meant to remain operational over sustained amounts of time. Long-duration autonomy is studied and analyzed from two different, yet complementary, perspectives: control algorithms and robot design. In the context of the former, the persistification of robotic tasks is presented. This consists of an optimization-based control framework which allows robots to remain operational over time horizons that are much longer than the ones which would be allowed by the limited resources of energy with which they can ever be equipped.

As regards the mechanical design aspect of long-duration robot autonomy, in the second part of this thesis, the SlothBot, a slow-paced solar-powered wire-traversing robot, is presented. This robot embodies the design principles required by an autonomous robotic system in order to remain functional for truly long periods of time, including energy efficiency, design simplicity, and fail-safeness.