Co-Design of a Miniature Swarm Robotics Platform: Custom Embedded Hardware, Distributed Sensing and Communication Architecture, and Edge-Assisted Perception

Abstract

This thesis investigates the co-design of a miniature swarm robotics platform that integrates custom embedded hardware, distributed wireless sensing and communication architecture, and edge-assisted perception. The work focuses on the design of a robot node built around a compact microcontroller-based printed circuit board (PCB) with support for camera-based perception, inertial sensing, motor actuation, and power regulation, alongside a Jetson Orin Nano edge compute platform for localization, perception, and coordination. The platform targets miniature metal-climbing robots derived from the Rovable project, originally developed at the MIT Media Lab, and adapted for vertical ferromagnetic surface operation in the Self-Organizing Swarms and Robotics (SSR) Lab at Princeton University. A key contribution is the integration of micro-ROS on the ESP32-S3 microcontroller, enabling native ROS 2-compatible communication between the resource-constrained robot nodes and the Jetson edge platform via XRCE-DDS over Wi-Fi UDP. The edge platform runs monocular-inertial ORB-SLAM3 on its multi-core ARM CPU for simultaneous localization and mapping (SLAM), enabling robots to build spatial maps of their operating environment without external motion capture infrastructure. The Jetson’s GPU represents an untapped resource for future acceleration of perception pipelines. Through subsystem design, quantitative analysis of power budgets, data rates, and end-to-end latency, this work develops a practical architectural framework for future perception-enabled miniature swarm robots with dynamic role assignment based on battery health and mission state.