A Multithreaded Real-time Robot for Embedded Design Space Exploration

Yue Ma

This thesis introduces an autonomous robot platform for real-time scheduling exper- imentation and benchmark suite to evaluate real-time optimizations and apply modern task scheduling methods. It makes two contributions. First, it presents a reference hardware and software design for a line-following, obstacle-avoiding and maze-solving robot. This robot is based on a small commercially-available product. The software is structured as a multithreaded real- time system for use in evaluating scheduling approaches for cost-sensitive and resource- constrained applications. Second, it provides a detailed design space exploration showing the costs (processor speed and memory) of di erent scheduling approaches (static vs. dynamic and non-preemptive vs. preemptive). It also measures and analyzes each task's timing information and explores the mini- mum microcontroller clock speed under di erent scheduling approaches.