In recent years, the field of aerial robotics has witnessed significant progress, finding applications in diverse domains, including post-disaster search and rescue operations. Despite these strides, the prohibitive acquisition costs associated with deploying physical multi-UAV systems have posed challenges, impeding their widespread utilization in research endeavors. To overcome these challenges, we present STAR (Swarm Technology for Aerial Robotics Research), a framework developed explicitly to improve the accessibility of aerial swarm research experiments. Our framework introduces a swarm architecture based on the Crazyflie, a low-cost, open-source, palm-sized aerial platform, well suited for experimental swarm algorithms. To augment cost-effectiveness and mitigate the limitations of employing low-cost robots in experiments, we propose a landmark-based localization module leveraging fiducial markers. This module, also serving as a target detection module, enhances the adaptability and versatility of the framework. Additionally, collision and obstacle avoidance are implemented through velocity obstacles. The presented work strives to bridge the gap between theoretical advances and tangible implementations, thus fostering progress in the field.
STAR is six modules that a lab can adopt piece by piece: a mission is a single YAML file, and the same configuration runs in the simulator and on the drones.
An STM32F405 flight controller, a Flow Deck v2 for optical flow and range, and an AI Deck 1.1 adding a GAP8 SoC, a monochrome camera and an ESP32 for streaming frames to the ground station. Low inertia keeps crashes cheap and safe.
Telemetry and broadcast commands — take-off, landing, setpoints — travel over 2.4 GHz radio, with request–response used on the ground for parameters such as controller gains. Camera frames go to the ground control station over Wi-Fi.
The central coordinator starts agents, tracks status, issues swarm primitives, fuses task distribution with collision avoidance, and executes instructions from external modules such as a search planner. Plans can target one UAV or the whole swarm.
Optimal Reciprocal Collision Avoidance picks the safe velocity closest to the one the agent wanted. Static obstacles enter as virtual agents placed along 2D polygon edges derived from the obstacle map; inter-UAV avoidance uses live positions.
Printed AprilTags with known global poses are encoded as landmark observations in a GTSAM factor graph alongside the EKF pose from IMU, optical flow and rangefinder. The same module doubles as target detection.
A lightweight simulator visualised in RViz2 runs the Crazyflie firmware bindings with numerical integration, drawing obstacles and landmark markers as wireframes so a configuration can be checked before anything leaves the ground.
Three trajectories, three landmark configurations, three laps each, measured against a VICON system used only as ground truth.
Marker processing, not the camera, sets the correction rate: frames arrive faster than the ground station can solve them.
@inproceedings{chiun2024star,
title={STAR: Swarm Technology for Aerial Robotics Research},
author={Chiun, Jimmy and Tan, Yan Rui and Cao, Yuhong and Tan, John and Sartoretti, Guillaume},
booktitle={International Conference on Control, Automation and Systems (ICCAS)},
year={2024}
pages={141-146},
keywords={Location awareness;System performance;Stars;Swarm robotics;Object detection;Motion capture;Peer-to-peer computing;Maintenance;Collision avoidance;Robots;Aerial Robotics;Micro UAV;Landmark Localization;Robotic Application},
doi={10.23919/ICCAS63016.2024.10773308}
}