Minimalist multi-agent filtering and guidance

Bobadilla Molina, Jaime

Permalink

https://hdl.handle.net/2142/45299 Copy

Description

Title

Minimalist multi-agent filtering and guidance

Author(s)

Bobadilla Molina, Jaime

Issue Date

2013-08-22T16:35:18Z

Director of Research (if dissertation) or Advisor (if thesis)

LaValle, Steven M.

Doctoral Committee Chair(s)

King, Samuel T.

Committee Member(s)

• LaValle, Steven M.
• Abdelzaher, Tarek F.
• Shell, Dylan A.

Department of Study

Computer Science

Discipline

Computer Science

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

• Robotics minimalism
• navigation
• patrolling
• tracking
• counting
• filtering.

Abstract

Advances in technology have allowed robots to be equipped with powerful sensors, complex actuators, computers with high processing capabilities, and high bandwidth communication links. This trend has enabled the development of sophisticated algorithms and systems to solve tasks such as navigation, patrolling, coverage, tracking, and counting. However, these systems have to deal with issues such as dynamical system identification, sensor calibration, and computation of powerful filters for state-feedback policies. This thesis presents novel techniques for tackling the above mentioned tasks. Our methods differ from traditional approaches since they do not require system identification, geometric map building, or state estimation. Instead, we follow a minimalist approach that takes advantage of the wild motions of bodies in their environment. The bodies move within regions connected by gates that enforce specific flows or provide simple sensor feedback. More specifically, five types of gates are proposed: 1) static gates, in which the flow direction of bodies cannot be changed during execution; 2) pliant gates, whose flow directions can be changed by gate-body collisions; 3) controllable gates, whose flow directions can be changed by powered actuators and sensor feedback; 4) virtual gates, in which the flow is affected by robot sensing and do not represent a physical obstruction; and 5) directional detection gates that do not change the flow of bodies, but simply detect bodies’ transitions from region to region. We show that the proposed methods lead to low-cost and easily deployable systems. More- over, they use efficient and simple algorithms that in most cases do not depend on the number of bodies. In order to demonstrate the practical feasibility of our ideas, we experimentally evaluate all five gates in various setups. First, we used static, pliant, and controllable gates in a series of experiments that manipulate the flow of Weasel Balls (without the weasels) and Hexbug Nano vibrating bugs. Second, we used inexpensive mobile robots and virtual gates to solve complex tasks, such as patrolling, disentanglement, and navigation. Finally, a low-cost hardware implementation of directional detection gates was developed for tracking and counting mobile bodies.

Graduation Semester

2013-08

Permalink

http://hdl.handle.net/2142/45299

Copyright and License Information

Copyright 2013 Jaime Leonardo Bobadilla Molina

Owning Collections