Abstract

Component-based design is a paradigm meant to aid in development of software applications by modularizing different functionalities of a system. This building-block approach is used extensively in cyber–physical and robotic systems. Common and established solutions to specific problems, such as perception nodes, state estimators, and motion planners, can be developed, verified, and reused as off-the-shelf modules. These modules may be integrated atop today’s heterogeneous hardware platforms, where an application can be distributed across GPUs, FPGAs, and CPU cores. When heterogeneous computational devices share the workload of a collection of components, the very act of integration may inject timing uncertainty. For example, data often must traverse different memory domains, incurring copies that may violate real-time budgets. Furthermore, the Robot Operating System (ROS)2, the de facto middleware for robotics, has a scheduler that departs from established real-time practice, and work has only recently begun on establishing timing guarantees for it. This dissertation builds a vertical solution that restores predictability without sacrificing modularity. The work is organized into 3 layers: memory management of hardware-accelerated devices, a real-time scheduler for ROS2, and a control/scheduling co-design study. Collectively, these contributions demonstrate that the promises of component-based design—modularity, compositional verification, and rapid development—can survive in the age of hardware acceleration, while still meeting cyber-physical constraints on timing and other properties end-to-end. By coupling memory-domain awareness with analyzable scheduling and timing-aware control, this work charts a practical path toward predictable, high-performance CPS and robotic systems.

Committee Chair

Chris Gill

Committee Members

Chris Gill; Roger Chamberlain; Ron Cytron; Tim York; Xuan Zhang

Degree

Doctor of Philosophy (PhD)

Author's Department

Computer Science & Engineering

Author's School

McKelvey School of Engineering

Document Type

Dissertation

Date of Award

8-18-2025

Language

English (en)

Download

Share

COinS