Modular Robots Morphology Transformation And Task Execution

Loading...
Thumbnail Image
Degree type
Doctor of Philosophy (PhD)
Graduate group
Mechanical Engineering & Applied Mechanics
Discipline
Subject
Control
Modular Robots
Motion Planning
Reconfiguration
Robotics
Funder
Grant number
License
Copyright date
2022-09-09T20:21:00-07:00
Distributor
Related resources
Author
Liu, Chao
Contributor
Abstract

Self-reconfigurable modular robots are composed of a small set of modules with uniform docking interfaces. Different from conventional robots that are custom-built and optimized for specific tasks, modular robots are able to adapt to many different activities and handle hardware and software failures by rearranging their components. This reconfiguration capability allows these systems to exist in a variety of morphologies, and the introduced flexibility enables self-reconfigurable modular robots to handle a much wider range of tasks, but also complicates the design, control, and planning. This thesis considers a hierarchy framework to deploy modular robots in the real world: the robot first identifies its current morphology, then reconfigures itself into a new morphology if needed, and finally executes either manipulation or locomotion tasks. A reliable system architecture is necessary to handle a large number of modules. The number of possible morphologies constructed by modules increases exponentially as the number of modules grows, and these morphologies usually have many degrees of freedom with complex constraints. In this thesis, hardware platforms and several control methods and planning algorithms are developed to build this hierarchy framework leading to the system-level deployment of modular robots, including a hybrid modular robot (SMORES-EP) and a modular truss robot (VTT). Graph representations of modular robots are introduced as well as several algorithms for morphology identification. Efficient mobile-stylereconfiguration strategies are explored for hybrid modular robots, and a real-time planner based on optimal control is developed to perform dexterous manipulation tasks. For modular truss robots, configuration space is studied and a hybrid planning framework (sampling-based and search-based) is presented to handle reconfiguration activities. A non-impact rolling locomotion planner is then developed to drive an arbitrary truss robot in an environment.

Advisor
Mark Yim
Date of degree
2021-01-01
Date Range for Data Collection (Start Date)
Date Range for Data Collection (End Date)
Digital Object Identifier
Series name and number
Volume number
Issue number
Publisher
Publisher DOI
Journal Issue
Comments
Recommended citation