Drone technology has developed rapidly over the past few years, but they’re still pretty dim-witted – requiring an operator to be within eyesight to prevent collisions. However, that might be about to change follow research into artificial neural networks.
We’ve long had computers that can recognise objects and respond in real-time, but the tech and power demands have made them unsuitable for mounting on a lightweight drone. To date, research into getting drones to sense and avoid obstacles has focused on short-range technology like radar, but that often doesn’t give enough warning for a fast-moving drone to avoid a power line.
So instead, a US company called Bio Inspired Technologies is taking a different approach – mimicking a biological brain. Humans and animals are generally rubbish at traditional computing tasks like adding up numbers, but great at processing input from multiple senses and responding to it.
Bio Inspired has built a “brain” out of memristors – resistors that have a short memory. They’re hooked up like the neurons in a biological brain, giving the chip-sized system the ability to recognise and “remember” objects like clouds, birds, buildings, radio towers and other drones – as well as using a camera to estimate how far away they are.
“Objects like other aircraft can be catalogued in a vague sense, meaning ‘I see an aircraft’, or in an exact sense: ‘I see another drone’,” Terry Gafron, CEO of Bio Inspired told New Scientist. “Not only could the system fly autonomously, but it could conceivably tell the difference between a deer and a wolf from the air.” That information can then be used by the drone to plot a new flight path.
Applications for the technology include Amazon’s fleet of delivery drones, precision agriculture, industrial caretaking and of course the military. Gafron says it could even fly around town looking for empty parking spaces and relaying that information to you in real-time.
Bio Inspired is working on a prototype of a drone equipped with the technology, which the company believes will be ready for its first flight later this year.
Reprinted from techradar, the original link: techradar