Video
Enhance your home page with a video that relates to your process, products, or story.
VOXL® 2
↪ M0054-2 and M0154-2 have identical performance characteristics and capabilities to M0054-1 and M0154-1, except the -2 assemblies only support up to 4 cameras concurrently.
VOXL 2
Product Description
VOXL 2 is a Blue UAS Framework AI companion computer and flight controller in one 16-gram board — the edge compute and autopilot for an autonomous drone or ground robot in a single part. Powered by the Qualcomm QRB5165, it is the smallest single-board AI autopilot we build: eight cores up to 3.091 GHz, 8 GB LPDDR5, 15 TOPS of on-board AI, a PX4 or ArduPilot flight controller running on the sensors DSP, and 5G connectivity through add-on modems.
VOXL 2 is NDAA '20 Section 848 compliant and designed and assembled in the USA. It is supported by a large ecosystem of pre-integrated image sensor modules, modems and expansion boards, and by the VOXL SDK for GPS-denied navigation, visual inertial odometry (VIO), SLAM, obstacle avoidance and object recognition.
Built to Adapt
Because VOXL 2 carries a high-performance computer, a drone built on it is not frozen the day it ships: new AI models, sensors, autonomy behaviors and countermeasures can be deployed as software. We call this software maneuverability. Software maneuverability is the ability to change what a drone can do, as fast as the threat changes, through software.
The VOXL SDK is available in source code form at code.modalai.com, so you can make those changes yourself rather than wait on a vendor roadmap. More than 1,000 partners have built on VOXL, which is why we call it an open foundation for America’s drones.
For how VOXL 2 compares with other options, see Top 5 Companion Computers for UAVs.
Product Specifications
| Feature | Details |
|---|---|
| CPU | Qualcomm QRB5165 — 8 cores up to 3.091 GHz |
| Memory | 8 GB LPDDR5 / 128 GB Flash |
| GPU | Adreno 650, 1024 ALU |
| NPU | 15 TOPS |
| Operating System | Ubuntu 18.04, Linux kernel v4.19 |
| Flight Controller | Embedded — PX4 or ArduPilot on the sensors DSP |
| Integrated Sensors | 2x TDK ICM-42688-P IMU, 1x ICP-10100 barometer |
| Image Sensors | 6 concurrent MIPI — 2x stereo pair plus 1x tracking |
| Video Encoding | 8K30 H.264 / H.265, 64 MP still images |
| Add-on Connectivity | Wi-Fi, 5G, 4G/LTE, Microhard |
| Thermal Cameras | FLIR Boson (USB, MIPI), FLIR Lepton (USB, SPI) |
| Power Consumption | 0.1 - 7 W |
| Dimensions | 70 mm x 36 mm |
| Weight | 16 g |
| Compliance | NDAA '20 Section 848 compliant, assembled in the USA |
Software Features
- PX4, ArduPilot, ROS 1 and ROS 2, Debian 9 (stretch), OpenCV 4.4, MAVROS, MAVSDK
- VOXL SDK: GPS-denied navigation, SLAM, visual obstacle avoidance, object recognition
- QGroundControl and ATAK support
- Open-source Linux kernel and cross-compilers
- Docker build environment for CPU, GPU (OpenCL) and DSP (Hexagon SDK) heterogeneous computer vision and deep learning
Technical Details
Feature Matrix · Connectors · Mechanical Drawings · Camera Configs
Product Brief + System Architecture (PDF)
Please see our handling instructions for best practices on handling your device.
What's Included
MDK-M0154-1-00 — VOXL 2 Dev Kit, Board Only
| MPN | Description |
|---|---|
| MCCA-M0154-1-T | QTY 1 CCA, VOXL 2, Tested |
MDK-M0154-1-01 — VOXL 2 Dev Kit, No Image Sensors
| MPN | Description |
|---|---|
| MCCA-M0154-1-T | QTY 1 CCA, VOXL 2, Tested |
| MDK-M0041-1-01 | QTY 1 VOXL Power Module v3 RevB Kit |
| MPS-00005-1 | QTY 1 Power Supply, 12V 3A, XT60 |
Additional Items Not Included
- Image sensor modules (sold separately — see image sensors)
- Modem add-on for Wi-Fi, LTE, 5G or Microhard connectivity (see modems)
- ESC and motors
Compatible Products
| Product | Compatible? |
|---|---|
| VOXL 2 I/O (M0065) | ✅ via MCBL-00061 UART cable |
| VOXL FPV ESC 4-in-1 (M0138) | ✅ 5 V version powers VOXL 2 directly; kit ships with VOXL 2 cabling |
| VOXL Mini 4-in-1 ESC (M0129) | ✅ use the 5 V version (M0129-5 / M0129-65) |
| Flight Core v2 (M0087) | ✅ external flight controller mode |
| MIPI image sensor modules (M0161, M0166, M0173, M0178, M0186) | ✅ up to 6 concurrent |
| M0189 8x8 ToF Module | ✅ requires additional cabling |
| FLIR Boson | ✅ USB and MIPI |
| FLIR Lepton | ✅ USB and SPI |
| Wi-Fi / 4G LTE / 5G / Microhard modem add-ons | ✅ |
| VOXL Power Module v3 (M0041) | ✅ |
| M0188 / M0195 VOXL 2 Mini carrier boards | ❌ VOXL 2 Mini only |
For the full accessory matrix see the VOXL 2 Feature Matrix.
VOXL 2 Reference Drones
Starling 2 · Starling 2 Max
Export Information
| Item | Value |
|---|---|
| ECCN | EAR99 |
| Harmonized Tariff Schedule (HTS) Code | 8542.31.0000 |
| Country of Origin | USA |
| Radio Frequencies Used | None on the board itself — RF is provided by optional add-on modems |
| RoHS Compliant | Yes |
Technical specifications
-
Autonomy
- 36 x 70mm SWaP-optimized for GPS-denied Navigation, Obstacle Avoidance
- Powered by Qualcomm QRB5165: 8 cores up to 3.091GHz, 8GB LPDDR5
- Seven simultaneous MIPI inputs such as EO, Thermal, TOF, and CV cameras
- Pre-integrated thermal camera support for FLIR Lepton® and Boson®
- Open-source software: Open CV, ROS 2, Docker, PX4 and Ardupilot integrated flight controller
- Integrated flight controller on DSP with TDK® ICM-42688 IMU and ICP-10111 Barometer
-
Artificial Intelligence
- The QRB5165’s Qualcomm AI Engine pushes 15 TOPS of AI performance with industry leading power consumption
- Onboard QRB5165: Hexagon Tensor Accelerator, GPU, DSP, NPU
- Perception models such as object segmentation and depth from mono
- NPU and GPU-accelerated TensorFlow Lite
-
Connectivity
- WiFi, 4G/5G, Doodle Labs, Microhard
Compare VOXL
CPU: QRB5165
OS: Debian Buster
GPU: Adreno 650 GPU - 1024 ALU
NPU: 15 TOPS
Built in Wifi: No
USB 3 Add-on Connectivity: WiFi, 5G, 4G/LTE, Microhard
MIPI-CSI2 Inputs: 4
Video Encoding: up to 8K30
Computer Vision Sensors: QTY 2 Stereo Pair, QTY 1 Tracking
Dimensions: 42mm x 42mm
Weight: 11g
VOXL SDK: Yes
ROS: ROS 2
NDAA Compliance: Assembled in USA, NDAA '20 Section 848, NDAA '21 Section 841
CPU: QRB5165
OS: Debian Buster
GPU: Adreno 650 GPU - 1024 ALU
NPU: 15 TOPS
Built in Wifi: No
B2B Add-on Connectivity: WiFi, 5G, 4G/LTE, Microhard, Doodle Labs
MIPI-CSI2 Inputs: 6
Video Encoding: up to 8K30
Computer Vision Sensors: QTY 2 Stereo Pair, QTY 1 Tracking
Dimensions: 70mm x 36mm
Weight: 16g
VOXL SDK: Yes
ROS: ROS 2
NDAA Compliance: Assembled in USA, NDAA '20 Section 848, NDAA '21 Section 841