“AI and the cloud is old news – the future is in edge,” says Lattice Semiconductor’s Deepak Boppana.

image

Lattice Semiconductor is a Silicon Valley maker of programmable logic devices founded in 1983 (yep, that’s the “Halt and Catch Fire” era), that sees the future in edge computing.

“AI and the cloud is old news, it’s been implemented over the last few years,” says Deepak Boppana, senior director, product and segment marketing at Lattice, said at the recent edition of Embedded World held in Nuremberg, Germany. “The move of AI is towards the edge: that’s the future.”

To show how they’re pushing closer to that future, Boppana has been on the conference circuit with the company’s low-power, wireless field-programmable gate array (FPGA) devices.  And when they say low power, they mean low power: some of the solutions are in the range of milliwatts — as in your average laser pointer.

At the 2018 edition of the Consumer Electronics Show (CES), he showcased this demo of vision applications for collision avoidance — with the toy robot gracefully moving around cones without hitting any.

The autonomous robot glides without incident across the factory floor using under one watt of power, Boppana says, adding “that’s really key for battery powered applications at the edge.” A wireless HD link uploads the video of the robot to its human overlords. Some key specs:

 

Other solutions Lattice is working on include AI face-based tracking (it works for objects, too) that could be implemented with drones or surveillance cameras and facial recognition solution.  That low power iCE40 UltraPlus FPGA performs edge computing applications locally on a device, sending metadata to the cloud but does that despite consuming than a milliwatt of power. Lattice offerings also feature a host of city applications — cameras that could be used for traffic flow monitoring, driving violations, smart parking and toll collection.

Check out the interview from Embedded World 2018 (and take a look the robots!) below. And you can catch up on these IRL at the upcoming Embedded Vision Conference in May 2018.

Via Electronics Weekly

Superuser