Mobile robotics specialist Fast Sense Studio has launched an intriguing multi-board computer system that packs six fully-independent artificial intelligence (AI) accelerators into a single unit: the Fast Sense AI Edge Computer.
"We believe in the future where reliable physical agents become relevant for routine tasks that most of us perform today," claims Yuri Ruwmyancev, co-founder and chief technology officer of Fast Sense Studio. "Reliability is what we are working on today, creating the environment for robots to process a large amount of data on board, learning from this and making independent, safe, and reliable decisions."
"Bringing together hardware and robotics engineers, software developers and data scientists, we have developed a prototype computer for mobile robotics with six independent edge AI accelerators from Intel and Google. It is fully tested and already integrated into our wheeled robot. The launch of the Fast Sense AI Edge Computer indicates a new benchmark for end-to-end robotic systems to help people achieve a new level of productivity."
Designed as a board-level integration system, the Fast Sense AI Edge Computer packs up to six independent modular processors designed for AI inference workloads: Three Google Coral Edge TPU accelerators, two Intel Myriad X accelerators, and an Intel graphics processor built into the system's Atom x6425E quad-core general-purpose processor. All told, Fast Sense Studio claims that "in some cases" its platform offers up to eight times the performance of rival embedded AI systems like NVIDIA's Xavier AGX.
The system offers up to 16GB of LPDDR4 memory, two USB 3.0 ports, six USB 2.0 ports, six UARTs, I2C, 1-Wire, and DisplayPort video output, with up to 64GB of on-board UFS 2.0 flash storage. Despite this, it measures just 30x55x84mm (around 1.18x2.17x3.31in). Each unit comes with Lubuntu 18.04.5 LTS - approaching three years old, but still under Canonical's Long Term Support until 2023 and under Extended Security Maintenance through to 2028 - pre-installed and with ready-to-run Docker containers for use as an edge AI device or ROS node.
More information on the platform is available on the Fast Sense Studio website, but the company has not yet providing pricing information.