June is looking bright so far in the open source drone community. With ICRA wrapping up at the end of May, we thought it would be nice to highlight some of the research coming out utilizing PX4. Researchers have brought significant milestones in navigation, neural network integration, perception, and obstacle avoidance, and underwater arm manipulation. Also, there’s another community event coming up that we’re excited to share, and a great video on ModalAI’s image processing pipeline we’re sure you’ll enjoy.
Guadalajara meetup: A Mexico first
Coming off the back of its successful events in San Diego and Philadelphia, The Dronecode Foundation is hosting its first-ever joint meetup with the Zephyr Project in Mexico, scheduled for July 8th at the NXP Offices in Guadalajara. This event will bring together the global PX4 community with embedded systems developers, creating new opportunities for collaboration and innovation.
We are actively looking for speakers — if you’ve been working on something cool with PX4, Pixhawk, Zephyr RTOS, or related technologies, we want to hear from you! Submit your proposal via the CFP form. And whether you want to present or simply attend and network, be sure to register here now because space is limited.
PX4 research highlights: Q2 2025
OmniNxt: Revolutionary Omnidirectional Perception
HKUST Aerial Robotics Group developed the OmniNxt platform, a fully open source aerial robot featuring omnidirectional visual perception using multi-fisheye cameras with a Jetson Orin NX. The system navigates cluttered indoor environments at speeds up to 1.0 m/s while avoiding obstacles, with omnidirectional perception allowing flight without yaw rotation for improved localization accuracy and energy efficiency compared to traditional limited field-of-view stereo systems. Read the paper.
Research by: Peize Liu et al, HKUST Aerial Robotics Group, The Hong Kong University of Science and Technology
Neural networks on embedded flight controllers
Norwegian University of Science and Technology’s Autonomous Robots Lab created a neural flight controller for PX4, developing a custom module that enables TensorFlow Lite neural networks to run directly on embedded microcontrollers like the STM32H743. This breakthrough completely replaces traditional cascaded PID control with an end-to-end neural controller, demonstrating successful simulation-to-reality transfer while preserving PX4’s safety features without requiring a companion computer. Read the paper.
Research by: Sindre M. Hegre et al, Autonomous Robots Lab, Norwegian University of Science and Technology
Advanced fixed-wing navigation
New York University’s Agile Robotics and Perception Lab introduced novel fixed-wing navigation using differential flatness and Bernstein polynomials similar to quadrotor “minimum snap” methods. Presented at ICRA 2025, the implementation dynamically adapts to curved trajectory constraints while maintaining optimal paths, potentially enabling more aggressive flight maneuvers for fixed-wing platforms with excellent sim-to-real transfer capabilities. Read the paper.
Research by: Luca Morando et al, The agile robotics and perception lab, New York University
Underwater innovation with BlueROV2
TU Hamburg showcased improved trajectory tracking for underwater vehicle-manipulator systems (UVMSs) using PX4 on the BlueROV2 platform. This research demonstrates smooth underwater manipulation capabilities, making advanced underwater robotics more accessible through commercial, lightweight platforms rather than expensive research vessels.
Research by: Niklas Trekel et al, Institute of Mechanics and Ocean Engineering, TU Hamburg
Image processing for Multi-Cam VIO: James Strawson PhD
This awesome video features James Strawson, PhD, Director of Software at ModalAI, demonstrating advanced image processing techniques for GPS-denied drone navigation using multi-camera Visual Inertial Odometry (VIO). His presentation features live demonstrations of the Starling 2 development platform, showcasing ModalAI’s approach to autonomous flight through custom hardware acceleration on Qualcomm chipsets. It’s a rare peek behind the curtains that you should definitely take advantage of.