Toward dynamically scalable open-source motion planning on the mobile edge and in the cloud

Gallegos, Emmanuel

Permalink

https://hdl.handle.net/2142/120126 Copy

Description

Title

Toward dynamically scalable open-source motion planning on the mobile edge and in the cloud

Author(s)

Gallegos, Emmanuel

Issue Date

2023-05-01

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

Amato, Nancy M

Department of Study

Computer Science

Discipline

Computer Science

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

M.S.

Degree Level

Thesis

Keyword(s)

• Robotics
• Motion Planning
• High Performance Computing
• Cloud Computing
• Edge Computing
• Parallel Algorithms
• Containers
• Docker
• Singularity

Language

eng

Abstract

This work presents a platform capable of deploying massively scalable and portable high performance software for robotics applications by utilizing containerization technologies. To this end, we employ a fully automated method for creating a swarm of containers among machines with homogeneous or heterogeneous computing power and operating systems that can work in tandem to perform large-scale distributed computation, including the ability to leverage robots as mobile edge devices. We demonstrate that this method can also be massively scaled on a large, shared HPC cluster, and evaluate the overhead associated with containerization by performing highly distributed motion planning calculations. For each of these platforms, we compare bare-metal and containerized runtime and scalability and show that our containerized platform is capable of fully utilizing the resources of the host machines while achieving the same nearly-linear scalability for parallel sampling-based motion planning algorithms that is possible on bare-metal implementations.

Graduation Semester

2023-05

Type of Resource

Text

Handle URL

https://hdl.handle.net/2142/120126

Copyright and License Information

Copyright 2023 Emmanuel Gallegos

Owning Collections