VOXL 2 AI companion computer and flight controller

The VOXL 2 from ModalAI is an AI companion computer and flight controller on a single board for autonomous drones and ground robots. It combines onboard perception and autonomy computing with an embedded autopilot, so a vehicle can run navigation, obstacle avoidance and object recognition without a separate flight controller.

The board is built around a Qualcomm QRB5165 processor with eight cores running at up to 3.091 GHz, 8 GB of LPDDR5 memory and 128 GB of flash storage. An Adreno 650 GPU and a neural processing unit provide 15 TOPS of AI performance, which is the figure that determines how large a perception or detection model can run in real time on the vehicle. Video encoding reaches 8K at 30 frames per second in H.264 and H.265, and the board handles still images up to 64 MP. The complete board weighs 16 g and measures 70 mm x 36 mm, with power consumption between 0.1 W and 7 W depending on load.

The flight controller runs on the sensors DSP, with PX4 or ArduPilot software available, and uses two TDK ICM-42688-P inertial sensors and an ICP-10100 barometer on the board. The VOXL SDK supports GPS-denied navigation, visual inertial odometry, SLAM, obstacle avoidance and object recognition. ROS 1 and ROS 2, MAVROS, MAVSDK, OpenCV, QGroundControl and ATAK are listed, and the operating system is Ubuntu 18.04 with Linux kernel 4.19.

For integration, the board accepts up to six concurrent MIPI camera inputs, arranged as two stereo pairs and a tracking camera, and the maker lists pre-integrated image sensor modules. Thermal cameras can be connected: FLIR Boson over USB or MIPI, and FLIR Lepton over USB or SPI. UART and USB interfaces are available for ESCs, radios and peripherals. The board has no radio of its own; Wi-Fi, 4G/LTE, 5G and Microhard connectivity are added through modem add-ons. Companion boards such as the VOXL 2 I/O and VOXL Power Module v3 extend the interfaces and power supply.

The maker states that the board is NDAA Section 848 compliant, designed and assembled in the USA, and part of the Blue UAS Framework. Typical applications are autonomous inspection, indoor and GPS-denied flight, obstacle-avoidance and mapping drones, ISR platforms, delivery drones and robotics research.