Decentralized task allocation in communication contested environments

• Main Author: Johnson, Luke B
• Other Authors: Jonathan P. How.
• Format: Others
• Language: English
• Published: Massachusetts Institute of Technology 2016
• Subjects: Aeronautics and Astronautics.
• Online Access: http://hdl.handle.net/1721.1/105606

Thesis: Sc. D., Massachusetts Institute of Technology, Department of Aeronautics and Astronautics, 2016. === Cataloged from PDF version of thesis. === Includes bibliographical references (pages 157-162). === This thesis explores the topic of decentralized task allocation. Specific emphasis is placed on how and when decentralized task allocation should be applied as a decision making tool for autonomous multi-agent missions. Even though the focus is on the decentralized aspect of task allocation, care is taken to identify the environments that do not actually benefit from decentralization, and what task allocation solutions are more appropriate for these environments. Chapter 2 provides a brief overview of the precise problem formulation and surveys the large number of task allocation approaches available. The result of this is an understanding when different classes of decentralized task allocation algorithms should be used. Chapters 3 and 4 introduce new algorithms that address fundamental issues with the past approaches to decentralized task allocation, and provide analysis of why these new approaches work. Specifically, Chapter 3 identifies a class of algorithms that utilize local information consistency assumptions (LICA), and an algorithm called Bid Warped Consensus Based Bundle Building Algorithm (BW-CBBA) is introduced to improve the state of art performance for this class of algorithms. Chapter 4 introduces an algorithm called the Hybrid Information and Plan Consensus (HIPC) algorithm, which uses LICA and two domains of information consensus in order to improve algorithmic performance over algorithms that exclusively perform consensus in a single domain. Chapter 5 introduces hardware experiments that verify and demonstrate the challenges associated with decentralized planning that are described throughout the thesis. Among other things, these experiments involved running a large number of planning algorithms onboard remote agents, and analyzing their performance in different communication and mission environments. Chapter 6 concludes the thesis with a summary of the contributions which highlights promising directions for new research. === by Luke B. Johnson. === Sc. D.