BEACON TRANSCRIPT – Drones have taken the market by storm, and they’re experiencing an accelerated development. Not so far in the future, companies – and nerds – will have their own autopilot drones that can avoid buildings trees, and other obstacles during flight.
A researcher at the Massachusetts Institute of Technology’s Computer Science and Artificial Intelligence Lab has taken things to the next level by crafting an autonomous drone that can fly itself and speed up to 30 miles per hour without guidance.
As he said in an interview published on MIT’s website, CSAIL PhD student Andrew Barry pointed to the fact that everyone and their brother is now focused on developing drones – and he’s not wrong. Just look at the drone projects released by social network Facebook, retail giant Amazon, and search giant Google.
They all have drone programs designed to improve their businesses; Google and Facebook plan on using them as tools for internet coverage, especially in remote areas, while Amazon hopes to generate more revenue by delivering goods to customers via drone. In spite of their advancements, no-one has figured out a way of stopping them from running into things.
And again, he’d be right. According to the U.S. Federal Aviation Administration, the main goal of regulating the widespread use of these flying computers is to find a way to protect people. Before allowing their public usage, regulators want safe drones that can avoid harm people or private property.
Barry was able to create a drone software that does just that – helps them fly more safely and duck in front of obstacles. Just like Google’s driverless car, Barry’s drone is equipped with a ‘self-driving’ feature, a sort of computer algorithm that can identify if there are any obstacles en route.
Barry’s software builds a map of the drone’s surroundings in real time, and the good news is that the MIT student has made the flight code available for all developers at Github. Unlike traditional drone algorithms that were using images captured by cameras and searching through the depth-field at multiple distances, Barry’s tech takes it a step further.
When he realized he doesn’t need step by step verification – seeing that the world simply doesn’t change much between frames – Barry designed his software to compute just a small subset of measurements, as far as 10 meters away. Barry’s next goal is to improve his invention by upgrading the drone’s algorithm so that it can detect obstacles at more than one depth.
Image Source: Fast Company