Controlling, modeling, and scaling underactuated, non-deterministic robot structures

Wasserman, Justin

Permalink

https://hdl.handle.net/2142/104041 Copy

Description

Title

Controlling, modeling, and scaling underactuated, non-deterministic robot structures

Author(s)

Wasserman, Justin

Contributor(s)

LaValle, Steven

Issue Date

2019-05

Keyword(s)

• distributed robotics
• minimal sensing/computing
• modeling and controlling multiagents
• physical interactions with multiagents
• cooperative manipulation

Abstract

This thesis studies the behavior of a class of underactuated, non-deterministic robots, the weaselball. In this study, weaselballs are placed in hubs that can attach to other hubs and form larger structures. These structures (of sizes ranging from a single weaselball to many) are studied. The behavior of a single weaselball structure is first investigated with the creation of a high-fidelity Gazebo simulation. This simulator is also highly compatible with Amazon Web Services and easily reconfigurable to study different weaselball structures. This allows the Gazebo simulation to be scalable. A Python toolbox is then introduced that allows for the behavior and trajectory of the simulated structures to be analyzed. After the behavior of a single weaselball structure is studied, the interaction between multiple weaselball structures is modeled. This behavior is modeled through the Kronecker product of the Markov transition matrices of the weaselball structures. This model blows up exponentially with the size of the states of the weaselball structures and the number of weaselballs. A library is thus introduced that can efficiently compute the Kronecker product matrix - vector dot product. From this library experimentation with the Kronecker product model of multiple weaselball structures can be explored.

Type of Resource

text

Language

en

Permalink

http://hdl.handle.net/2142/104041

Owning Collections