A group of MIT researchers have developed an autonomous robotic plane that is capable of guiding itself around indoor environments without the use of GPS.
MIT's Robust Robotics Group recently created an algorithm that allows self-guided planes to maneuver their way around objects inside buildings without requiring GPS to do so. They did this by bringing different state-estimation algorithms together, including a particle filter, which is both accurate and time-consuming, and a Kalman filter, which is not quite as accurate but is more efficient.
The resulting algorithm was placed within the electronics on a robotic plane that the MIT team built themselves. The plane has short, broad wings to allow it to make tight turns and fly at higher speeds for better maneuverability.
The plane was equipped with an accurate digital map of its environment, accelerometers, gyroscopes and other tools to help it calculate its multi-dimensional surroundings. It also has to calculate other factors, such as the right amount of acceleration, the correct velocity and its orientation when flying indoors. This is where the algorithm is helpful -- the plane has to determine its current state using all of these properties' different dimensions, which equates to calculating 15 different values at a time. This is a pretty heavy workload for the plane.
The algorithms work like this: the particle filter was used for variables that required it, and the results were translated back into the Kalman filter's language. The idea was to put the different algorithms, maps, etc. on a single platform.
MIT has some more work ahead of it, such as creating an algorithm that can create a map of the plane's environment while it's traveling instead of in advance, but the team has managed to complete a series of flight tests in a MIT parking garage, where it successfully dodged obstacles like pillars.
