Dynamic presentations by Sebastian Thrun and Nicholas Roy have alerted and informed CAFE Foundation’s Electric Aircraft Symposia attendees of highly sophisticated efforts to allow autonomous full-size automobiles and miniature helicopters to navigate through or over unfamiliar terrain.
Using clues from lasers, infrared sensors, inertial guidance systems and sometimes GPS coordinates, the vehicles use control algorithms to guide themselves around obstacle-strewn courses. As noted in Science Daily and the The Massachusetts Institute of Technology’s press office this week, “Dozens of research teams have competed in a series of autonomous-helicopter challenges posed by the Association for Unmanned Vehicle Systems International (AUVSI); progress has been so rapid that the last two challenges have involved indoor navigation without the use of GPS.”
Mini-copters have shown an amazing ability to not only navigate, but to perform complex tasks in swarms, such as building large architectural assemblies – all without human intervention (other than, one assumes, someone pushing a “go” button).
MIT’s Robust Robotics Group has taken a different tack, finding ways for “GPS-denied” fixed-wing aircraft to navigate indoors.
At the last two International Conferences on Robotics and Automation (ICRA), Robust Robotics team members presented papers on developing algorithms to calculate a plane’s trajectory and determine its “state” – its location, physical orientation, velocity and acceleration.
Nicholas Roy, associate professor of aeronautics and astronautics and head of the group, explains their motivation. “The reason that we switched from the helicopter to the fixed-wing vehicle is that the fixed-wing vehicle is a more complicated and interesting problem, but also that it has a much longer flight time. The helicopter is working very hard just to keep itself in the air, and we wanted to be able to fly longer distances for longer periods of time.”
He notes the added complexity. “It’s going much faster, and it can’t do arbitrary motions. [Fixed wing aircraft] can’t go sideways, they can’t hover, they have a stall speed.”
Researchers needed a slow, maneuverable airplane to fly in close quarters. It needed to be slow enough that on-board computing could keep up with the slow particle filter algorithms that calculate very precise “state” estimates, and faster Kalman filter algorithms that provide a less precise but quicker estimate of the airplane’s situation based on inertial information from on-board sensors including mini-gyros and accelerometers, and a view of the outside world coming from laser rangefinders. These algorithms have to determine the airplane’s orientation, velocity and acceleration, then make quick control decisions based on 15 different factors being input into the central control package.
Because the airplane could not stop in mid-flight to get its bearings, MIT researchers incorporated a digital “map” into their computations to help give a sense of boundaries to otherwise unknown locations. This map will later be discarded and the airplane will fly under the same “blind” conditions that helicopters are constrained by in these challenges.
The initial efficient motion-planning algorithms were developed by AeroAstro professor Emilio Frazzoli of the Aerospace Robotics and Embedded Systems (ARES) Laboratory, but were normally used in a stable platform. To ensure that mobile state estimates were reliable, Bry and Roy added an extra variable, increasing the dynamic computing complexity of the problem.
To create that all-important slow, maneuverable aircraft to test these mathematical guides, Adam Bry, lead author of the ICRA papers and a graduate student in the Department of Aeronautics and Astronautics, turned to professor Mark Drela about a design that would suit the low-speed maneuverability criteria. Drela was designer of Daedalus, the human-powered airplane that flew 72.4 miles between Crete and the Greek island of Santorini in 1988. Drela created XFoil, a much-used aerodynamic program, to assist design of that craft.
His airplane is a two-meter wingspan, pod-and-boom with tubular spars and a broad flat area for sensors and computers to size up the surrounding area, assess the airplane’s current state and guide the craft through new challenges. Since the second of two tests was in a 2.5-meter high parking garage, precise flight became a necessity.
According to MIT, “The researchers’ next step will be to develop algorithms that can build a map of the plane’s environment on the fly. Roy says that the addition of visual information to the rangefinder’s measurements and the inertial data could make the problem more tractable. ‘There are definitely significant challenges to be solved,’ Bry says. ‘But I think that it’s certainly possible.’”
Their work can be seen in their paper, “State Estimation for Aggressive Flight in GPS-Denied Environments Using Onboard Sensing,” by Adam Bry, Abraham Bachrach, and Nicholas Roy.
The videos are courtesy of Melanie Gonick, MIT News.