Latest UK News - Breaking News & Headlines

Drone Chip Imec More Accurate Than Expected

22,810

The chip that imec presented last year to make AI decisions in an energy-efficient way is being expanded. At the same time, a study shows that AI is more accurate than other neural networks.

 

The chip, announced last year, uses a spiking recurrent neural network (SNN) to mimic the brain’s neural network. This provides chip technology that can make decisions quickly but also energy-efficiently.

A year later, imec says it is further scaling up its chip on a 40-nanometer architecture to allow for more IoT and autonomous robotics cases.

‘SNN technology will find its way into a wide range of applications: from smart, self-learning Internet of Things (IoT) devices – such as wearables – to autonomous drones and robots. But each of those cases has its specific requirements, says Ilja Ocket, program mamanger neuromorphic sensing at imec.

‘For example, spiking neural networks for IoT applications must mainly excel in energy efficiency, while autonomous drones primarily need a limited signal delay – so that they can quickly avoid obstacles.’

In addition to expanding to more cases, imec will also continue to use the chip for drones. It wants to move to an end-to-end spiking approach here. This means that the spiking recurrent neural network is used much more widely in processing data from sensors and images to make decisions by AI.

Imec says it is the first to do so and says it is also still looking for drone companies to experiment with the technology.

Leave A Reply