JavaScript is disabled for your browser. Some features of this site may not work without it.
Active Perception and Planning for Modular Self-Reconfigurable Robots

Author
Daudelin, Jonathan
Abstract
Modular robots have the unique ability to reconfigure their shape and capabilities to adapt to various challenges in the environment. In order to perform tasks autonomously in unknown environments, active perception and planning algorithms are required that can leverage their adaptive capabilities. This work presents several such perception and planning tools. An novel, probabilistic object reconstruction algorithm is presented that allows a generic mobile robot (such as a modular robot) intelligently position a 3D sensor to explore unknown objects in its environment. Then, it presents fully autonomous, perception-informed systems for modular self-reconfigurable robots (MSRRs) that enable them to explore, dynamically adapt to their environment, and even augment their environment to perform high-level tasks. Finally, it presents an end-to-end path planning framework for MSRR systems that enables them to reconfigure between multiple morphologies and use multiple gaits in order to traverse and plan optimal paths over challenging terrain.
Date Issued
2018-08-30Subject
machine learning; Autonomous Systems; Modular Robots; Computer science; Robotics; Path planning
Committee Chair
Campbell, Mark
Committee Member
Kress Gazit, Hadas; Ferrari, Silvia
Degree Discipline
Mechanical Engineering
Degree Name
Ph. D., Mechanical Engineering
Degree Level
Doctor of Philosophy
Type
dissertation or thesis