Date of Award

4-2011

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Computer Science

First Advisor

Dr. Prithviraj Dasgupta

Second Advisor

Dr. Carl Nelson

Third Advisor

Dr. Jong-Hoon Youn

Abstract

In this thesis, we consider the problem of autonomous self-reconfiguration by modular self-reconfigurable robots (MSRs). MSRs are composed of small units or modules that can be dynamically configured to form different structures, such as a lattice or a chain. The main problem in maneuvering MSRs is to enable them to autonomously reconfigure their structure depending on the operational conditions in the environment. We first discuss limitations of previous approaches to solve the MSR self-reconfiguration problem. We will then present a novel framework that uses a layered architecture comprising a conventional gait table-based maneuver to move the robot in a fixed configuration, but using a more complex coalition game-based technique for autonomously reconfiguring the robot. We discuss the complexity of solving the reconfiguration problem within the coalition game-based framework and propose a stochastic planning and pruning based approach to solve the coalition-game based MSR reconfiguration problem. We tested our MSR self-reconfiguration algorithm using an accurately simulated model of an MSR called ModRED (Modular Robot for Exploration and Discovery) within the Webots robot simulator. Our results show that using our coalition formation algorithm, MSRs are able to reconfigure efficiently after encountering an obstacle. The average “reward” or efficiency obtained by an MSR also improves by 2-10% while using our coalition formation algorithm as compared to a previously existing multi-agent coalition formation algorithm. To the best of our knowledge, this work represents two novel contributions in the field of modular robots. First, ours is one of the first research techniques that has combined principles from human team formation techniques from the area of computational economics with dynamic self-reconfiguration in modular self-reconfigurable robots. Secondly, the modeling of uncertainty in coalition games using Markov Decision Processes is a novel and previously unexplored problem in the area of coalition formation. Overall, this thesis addresses a challenging research problem at the intersection of artificial intelligence, game theory and robotics and opens up several new directions for further research to improve the control and reconfiguration of modular robots.

Comments

A Thesis Presented to the Department of Computer Science and the Faculty of the Graduate College University of Nebraska In Partial Fulfillment of the Requirements for the Degree M.S in Computer Science University of Nebraska at Omaha. Copyright 2011 Zachary Ramaekers.

Download

COinS