Heads Up! It's An Autonomous Airplane That Flies Indoors.

In 2009, MIT’s Robust Robotics Group won the Association for Unmanned Vehicle Systems International's (AUVSI) aerial-robotics competition when its autonomous mini-helicopter navigated its way through a simulated nuclear meltdown without access to GPS data.

For an encore, the group set a tougher goal: develop an autonomous micro-airplane that can handle the close quarters of indoor flying using only its on board sensors. As a bonus, they built their plane from scratch.

Traditional autonomous micro air-vehicles are usually limited to slow, deliberate flights. The MIT group's fixed-wing vehicle, which weighs a little more than four pounds, can fly and navigate obstacles at relatively high speeds.

In the MIT News video, the airplane is put through a series of tests in the parking garage below the Stata Center and successfully avoids obstacles like columns, cars, and a low-ceiling. The plane averaged 22 miles per hour and covered more than three miles.

[youtube http://www.youtube.com/watch?v=kYs215TgI7c&w=400&h=225]

In tight spaces, airplanes are more difficult to navigate than helicopters because they can achieve faster speeds but can’t make arbitrary motions like hovering or moving sideways. The team, which includes Professor Mark Drela and Associate Professor Nick Roy, created a slim plane with short, wide wings (about six and a half feet long) and the computational power of a netbook.

From PC World:

It needs all this processing power to run a state-estimation algorithm in conjunction with a set of lasers, accelerometers, and gyroscopes. With these combined technologies, the UAV is able to figure out its own orientation (i.e. pitch, roll, and yaw) and velocity, as well as 15 other in-flight factors without a GPS signal. At the same time, the UAV constantly runs an algorithm that it uses to avoid obstacles it comes across on the fly.

The MIT-designed airplane was uploaded with a digital map of its surroundings, something the helicopter did not have. Their next goal is to develop an algorithm that can map the plane’s environment on the fly.