NVIDIA Jetson for the Military and Aerospace Industries
NVIDIA® is well known for its graphics, GPU, and processing technologies that are trusted by not only gamers, but increasingly businesses, data centers, and defense system integrators to take their computing capabilities to the next level.
Embedded applications in aerospace and defense require technology that is rugged, has a long lifecycle, and is designed for size, weight, and power (SWaP)-constrained environments. NVIDIA’s Jetson™ family of system-on-modules (SoMs) serve as exceptionally power-efficient, high-performance building blocks for embedded solutions optimized for artificial intelligence (AI), machine learning, deep learning, and edge computing.
By integrating NVIDIA’s industrial Jetson technology into Curtiss-Wright’s hyper-rugged, scalable mission computers, today’s defense and aerospace system integrators are able to deploy Mil-grade, low-SWaP computing solutions at the edge of the battlefield.
Rugged Mission Computing with NVIDIA Jetson inside
NVIDIA’s Jetson family currently includes two rugged industrial-rated SoMs ─ the TX2i and AGX Xavier Industrial. These are integrated into the Parvus® small form factor mission computer product line in the DuraCOR® 312 and DuraCOR AGX-Xavier models, respectively. To expand system capabilities, Curtiss-Wright has designed the DuraCOR line with scalability, including a variety of add-on I/O card slots and SSD storage options.
Based on a multicore Arm® CPU architecture paired with an NVIDIA Pascal (312) or Volta (AGX) GPU, these mission computers offer low-power general-purpose or GPU-accelerated processing. As each leverages NVIDIA’s Linux-based JetPack software development environments and pre-trained machine learning models, software developers can port software running on a lower power device like the TX2i to a higher performance device in the AGX Xavier Industrial without requiring significant software development time.
Both the Jetson AGX Xavier Industrial and the Jetson TX2i SoMs are created with industrial uses in mind–with an impressive temperature range of -40°C to +85°C and a lifecycle of 10 years. These also have memory with error correction code (ECC) support and greater shock and vibration resistance than their commercial counterparts, making them ideal for military uses.
Parvus DuraCOR AGX-Xavier
The DuraCOR AGX-Xavier features the Jetson AGX Xavier Industrial SoM, which integrates not only a powerful 512-core GPU with tensor cores optimized for AI but also multiple dedicated deep learning and multimedia accelerator engines to boost machine learning performance. The Jetson AGX Xavier Industrial module boasts up to 22 TOPS of deep learning or 11 TFLOPS of CUDA computing performance. With reference libraries and trained models, the AGX Xavier is ideal for autonomous vehicles with AI needs.
The DuraCOR AGX-Xavier can deliver up to 20 times the deep learning performance of the DuraCOR 312 and features much higher speed interfaces, including 10 Gigabit Ethernet, USB 3.1 interfaces, and more display outputs, as well as removable high-speed NVMe SSD storage and optional high-speed GMSL camera interfaces and support for both mini-PCIe and PCIe104 add-on cards.
Parvus DuraCOR 312
Based on a Jetson TX2i SoM, the DuraCOR 312 is a lower-power option for very SWaP-constrained applications. Combining powerful NVIDIA Pascal/CUDA-core GPU processing with 64-bit Armv8 CPU for size, weight, power, and cost (SWaP-C) sensitive mobile, airborne, ground, manned, and unmanned vehicle and sensor platforms, the unit integrates a massively parallel, 256-core GPU architecture.
Thanks to its Pascal GPU architecture (supporting Max-Q and Max-P dynamic energy profiles), larger memory bandwidth, and support for CANbus, the DuraCOR 312 can deliver up to twice the performance or power efficiency of TX1-based systems together with native vehicle bus interfaces. The unit also delivers an unparalleled modular system design, boasting multiple Mini-PCIe I/O card slots, high-speed M.2 internal storage, removable SATA Flash SSD capabilities, and an aerospace-grade power supply supporting 50ms power hold-up in a fanless IP67-rated mechanical package designed for wide temperatures and harsh shock and vibration, along with ECC memory for mitigation against single event upsets at high altitudes.
Whether military or aerospace programs are seeking the AI and machine learning capabilities of the Jetson AGX Xavier Industrial or the lower power Jetson TX2i SoM, Curtiss-Wright has the most rugged COTS systems available based on these technologies with Mil-grade reliability to deliver enormous computing horsepower at the edge for embedded vehicle and aerospace applications.
How Can I Teach My Machine To Learn
Like humans, machines learn from experience. They make observations from inputs of images, text, or other data, and then look for patterns. After the machine runs through the mathematical layers, it learns to make better decisions based on the examples it was given.