VOXL^®

U.S.-Built Video and Autonomy for UAS

ModalAI accelerates UAS development with U.S.-built video and autonomy — components and complete aircraft that are mission-ready on delivery and open for developers to extend. Whether you are integrating a compute board into your own airframe or flying one of ours out of the box, the same stack, the same SDK and the same documented interfaces carry all the way through.

Components

VOXL handles the compute, video and GPS-denied navigation that autonomous flight depends on.

	Component	What it does
	VOXL 2	AI companion computer and flight controller in one 16-gram board. Qualcomm QRB5165, eight cores up to 3.091 GHz, 8 GB LPDDR5, 15 TOPS of on-board AI, and a PX4 or ArduPilot flight controller running on the sensors DSP.
	VOXL 2 Mini	The same stack in an 11-gram, 42 mm x 42 mm board with a 30.5 mm standard mounting pattern, for the smallest airframes and ground robots.
	VOXL SDK	GPS-denied navigation and visual inertial odometry (VIO), SLAM, obstacle avoidance and object recognition, with ROS 1 and ROS 2 and PX4 support.

Around those boards sits a supported ecosystem of image sensors, ESCs, radios and data links, and connectivity for beyond visual line of sight over 4G/LTE and 5G.

Aircraft

Our aircraft put that stack in the air, and stay open once they land.

	Aircraft	What it is
	Starling 2 Max	VOXL 2-powered development drone for GPS-denied flight, built for computer-vision dead reckoning with 500 g of payload capacity. See the Starling family.
	Stinger Vision FPV	Ready-to-fly FPV aircraft on VOXL 2 Mini, on the Blue UAS Cleared List, with GPS-denied autonomy and an 11 km encrypted video link. See our FPV components.

Open by Design

Source access, documented interfaces and a MOSA architecture mean a government lab or integrator can change flight behavior, add a payload or run their own perception stack without waiting on us. That is the difference between a drone you operate and a drone you can build on.

• Open source Linux kernel, cross-compilers, PX4, ROS and OpenCV, with full technical documentation
• Documented connector pinouts, flex adapters and expansion boards, so a new payload is an integration job rather than a support ticket
• Docker build environment for CPU, GPU (OpenCL) and DSP (Hexagon SDK) computer vision and deep learning, and Docker on target
• Long-standing participation in the open autonomy community, including Dronecode

Lineage: Snapdragon Flight to VOXL 2

This architecture has a long history, and the same people have carried it the whole way.

• 2015 — Snapdragon Flight. A breakthrough capability launched by the team that went on to found ModalAI: a full flight controller and Linux companion computer on a single small board, at a time when that combination did not exist. It later flew on NASA JPL's Ingenuity helicopter, the first aircraft to make a powered flight on another planet — that story is here.
• Qualcomm Flight Pro. The successor, adding processing, connectivity and machine vision to the same idea.
• VOXL 1. Built directly from the Qualcomm Flight Pro architecture by the engineers who created it, and the first board to carry the VOXL name. It is no longer available, along with VOXL Flight, VOXL Flight Deck and VOXL m500.
• VOXL 2 and VOXL 2 Mini. Where the line is today: an order of magnitude more compute in a smaller, lighter package, on the Blue UAS lists and supported by weekly software releases.

Also from this lineage: the Qualcomm Flight RB5 5G Platform.

Built in the U.S.A.

Everything here is designed and built in the U.S.A. and aligned with Blue UAS and NDAA '20 Section 848 requirements. There are two lists, and we are on both:

• Blue UAS Framework covers components — VOXL 2 and VOXL 2 Mini are on it, so an integrator can build on a cleared autopilot. See what that means for your program.
• Blue UAS Cleared covers complete, fully vetted unmanned aircraft — Stinger Vision FPV is on it.