Rudi-NX embedded mission computer

The Rudi-NX embedded mission computer from Connect Tech is an enclosed embedded system built around the NVIDIA Jetson Orin NX module, a mission computer for sensor processing, machine vision and edge AI on uncrewed aircraft, robots and other vehicles. It is sold as an ordering code with either the 8 GB or the 16 GB version of the module, which sets the memory and the AI throughput.

The compute side follows the Jetson Orin NX module: a 1024-core NVIDIA Ampere GPU with 32 Tensor Cores and two NVDLA v2 deep-learning accelerators. The maker quotes 100 TOPS for the 16 GB version and 70 TOPS for the 8 GB version, with an 8-core Arm Cortex-A78AE CPU on the 16 GB module and a 6-core CPU on the 8 GB one. Memory is 128-bit LPDDR5. For video, the hardware H.265 encoder handles one 4K60 stream, three 4K30, six 1080p60 or twelve 1080p30 streams, which matters when several camera feeds must be compressed for a datalink or for recording at once.

Sensor connectivity is the main reason to choose this unit. It accepts up to 4 GMSL1/2 camera inputs that reach the Jetson over MIPI CSI-2, and also has four USB 3.0 Type-A ports, one USB OTG port and two 10/100/1000 Ethernet ports. For avionics and payload wiring there is one CAN 2.0b port, one UART, one RS-485 port, two I2C, two SPI, four GPIO and two PWM lines, plus 5 V and 3.3 V outputs. An HDMI 2.0 output is available for display and bench work.

For integration, the unit measures 135 mm x 50 mm x 105 mm (width x height x depth) and has a mass of 646 g with the module only, without NVMe storage, Wi-Fi/Bluetooth or LTE modules. Input power is 9 to 36 V DC, which covers common 12 V and 24 V vehicle supplies and multi-cell battery rails, and the operating temperature range is -20 to +80 degC. Storage is one M.2 M-key NVMe slot (PCIe x4, 2280); M.2 B-key and E-key slots take cellular or wireless cards. External mounting brackets are an ordered accessory and need no additional hardware. Software support comes as NVIDIA JetPack (Linux for Tegra) with board-support packages from Connect Tech, and the system is tagged ROS capable.

Typical uses are onboard vision processing, object detection and tracking, multi-camera perception and sensor fusion on drones, ground robots and industrial vehicles, where one computer must collect camera, CAN and serial data and run neural networks next to the sensors.