June is wanting brilliant thus far within the open supply drone group. With ICRA wrapping up on the finish of Could, we thought it will be good to focus on a number of the analysis popping out using PX4. Researchers have introduced important milestones in navigation, neural community integration, notion, and impediment avoidance, and underwater arm manipulation. Additionally, there’s one other group occasion arising that we’re excited to share, and an ideal video on ModalAI’s picture processing pipeline we’re positive you’ll get pleasure from.
Guadalajara meetup: A Mexico first
Coming off the again of its profitable occasions in San Diego and Philadelphia, The Dronecode Basis is internet hosting its first-ever joint meetup with the Zephyr Mission in Mexicoscheduled for July eighth on the NXP Places of work in Guadalajara. This occasion will carry collectively the worldwide PX4 group with embedded methods builders, creating new alternatives for collaboration and innovation.
We’re actively on the lookout for audio system — when you’ve been engaged on one thing cool with PX4, Pixhawk, Zephyr RTOS, or associated applied sciences, we wish to hear from you! Submit your proposal through the CFP kind. And whether or not you wish to current or just attend and community, make sure you register right here now as a result of house is restricted.
PX4 analysis highlights: Q2 2025
OmniNxt: Revolutionary Omnidirectional Notion
HKUST Aerial Robotics Group developed the OmniNxt platforma totally open supply aerial robotic that includes omnidirectional visible notion utilizing multi-fisheye cameras with a Jetson Orin NX. The system navigates cluttered indoor environments at speeds as much as 1.0 m/s whereas avoiding obstacles, with omnidirectional notion permitting flight with out yaw rotation for improved localization accuracy and vitality effectivity in comparison with conventional restricted field-of-view stereo methods. Learn the paper.
Analysis by: Peize Liu et al, HKUST Aerial Robotics Group, The Hong Kong College of Science and Know-how
Neural networks on embedded flight controllers
Norwegian College of Science and Know-how’s Autonomous Robots Lab created a neural flight controller for PX4, creating a customized module that allows TensorFlow Lite neural networks to run immediately on embedded microcontrollers just like the STM32H743. This breakthrough fully replaces conventional cascaded PID management with an end-to-end neural controller, demonstrating profitable simulation-to-reality switch whereas preserving PX4’s security options with out requiring a companion laptop. Learn the paper.
Analysis by: Sindre M. Hegre et al, Autonomous Robots Lab, Norwegian College of Science and Know-how
Superior fixed-wing navigation
New York College’s Agile Robotics and Notion Lab launched novel fixed-wing navigation utilizing differential flatness and Bernstein polynomials much like quadrotor “minimal snap” strategies. Offered at ICRA 2025, the implementation dynamically adapts to curved trajectory constraints whereas sustaining optimum paths, doubtlessly enabling extra aggressive flight maneuvers for fixed-wing platforms with wonderful sim-to-real switch capabilities. Learn the paper.
Analysis by: Luca Morando et al, The agile robotics and notion lab, New York College
Underwater innovation with BlueROV2
TU Hamburg showcased improved trajectory monitoring for underwater vehicle-manipulator methods (UVMSs) utilizing PX4 on the BlueROV2 platform. This analysis demonstrates easy underwater manipulation capabilities, making superior underwater robotics extra accessible via business, light-weight platforms fairly than costly analysis vessels.
Analysis by: Niklas Trekel et al, Institute of Mechanics and Ocean Engineering, TU Hamburg
Picture processing for Multi-Cam VIO: James Strawson PhD
This superior video options James Strawson, PhD, Director of Software program at ModalAIdemonstrating superior picture processing methods for GPS-denied drone navigation utilizing multi-camera Visible Inertial Odometry (VIO). His presentation options reside demonstrations of the Starling 2 improvement platformshowcasing ModalAI’s strategy to autonomous flight via customized {hardware} acceleration on Qualcomm chipsets. It’s a uncommon peek behind the curtains that you need to undoubtedly benefit from.
