MIT Professor Jonathan How’s research interests span the gamut of autonomous vehicles — from airplanes and spacecraft to unpiloted aerial vehicles (UAVs, or drones) and cars. He is especially focused on the design and implementation of distributed robust planning algorithms to coordinate multiple autonomous vehicles able to navigating in dynamic environments.
For the past yr or so, the Richard Cockburn Maclaurin Professor of Aeronautics and Astronautics and a team of researchers from the Aerospace Controls Laboratory at MIT have been developing a trajectory planning system that enables a fleet of drones to operate in the identical airspace without colliding with one another. Put one other way, it’s a multi-vehicle collision avoidance project, and it has real-world implications around cost savings and efficiency for quite a lot of industries including agriculture and defense.
The test facility for the project is the Kresa Center for Autonomous Systems, an 80-by-40-foot space with 25-foot ceilings, customized for MIT’s work with autonomous vehicles — including How’s swarm of UAVs usually buzzing across the center’s high bay. To avoid collision, each UAV must compute its path-planning trajectory onboard and share it with the remaining of the machines using a wireless communication network.
But, in line with How, certainly one of the important thing challenges in multi-vehicle work involves communication delays related to the exchange of knowledge. On this case, to deal with the problem, How and his researchers embedded a “perception aware” function of their system that enables a vehicle to make use of the onboard sensors to collect latest information concerning the other vehicles after which alter its own planned trajectory. In testing, their algorithmic fix resulted in a one hundred pc success rate, guaranteeing collision-free flights amongst their group of drones. The following step, says How, is to scale up the algorithms, test in larger spaces, and eventually fly outside.
Born in England, Jonathan How’s fascination with airplanes began at a young age, due to ample time spent at airbases along with his father, who, for a few years, served within the Royal Air Force. Nevertheless, as How recalls, while other children desired to be astronauts, his curiosity had more to do with the engineering and mechanics of flight. Years later, as an undergraduate on the University of Toronto, he developed an interest in applied mathematics and multi-vehicle research because it applied to aeronautical and astronautical engineering. He went on to do his graduate and postdoctoral work at MIT, where he contributed to a NASA-funded experiment on advanced control techniques for high-precision pointing and vibration control on spacecraft. And, after working on distributed space telescopes as a junior faculty member at Stanford University, he returned to Cambridge, Massachusetts, to hitch the school at MIT in 2000.
“Considered one of the important thing challenges for any autonomous vehicle is the right way to address what else is within the environment around it,” he says. For autonomous cars meaning, amongst other things, identifying and tracking pedestrians. Which is why How and his team have been collecting real-time data from autonomous cars equipped with sensors designed to trace pedestrians, after which they use that information to generate models to know their behavior — at an intersection, for instance — which enables the autonomous vehicle to make short-term predictions and higher decisions about the right way to proceed. “It’s a really noisy prediction process, given the uncertainty of the world,” How admits. “The true goal is to enhance knowledge. You are never going to get perfect predictions. You are just trying to know the uncertainty and reduce it as much as you possibly can.”
On one other project, How is pushing the boundaries of real-time decision-making for aircraft. In these scenarios, the vehicles must determine where they’re positioned within the environment, what else is around them, after which plan an optimal path forward. Moreover, to make sure sufficient agility, it is usually crucial to give you the option to regenerate these solutions at about 10-50 times per second, and as soon as latest information from the sensors on the aircraft becomes available. Powerful computers exist, but their cost, size, weight, and power requirements make their deployment on small, agile, aircraft impractical. So how do you quickly perform all of the crucial computation — without sacrificing performance — on computers that easily fit on an agile flying vehicle?
How’s solution is to employ, on board the aircraft, fast-to-query neural networks which can be trained to “imitate” the response of the computationally expensive optimizers. Training is performed during an offline (pre-mission) phase, where he and his researchers run an optimizer repeatedly (hundreds of times) that “demonstrates” the right way to solve a task, after which they embed that knowledge right into a neural network. Once the network has been trained, they run it (as a substitute of the optimizer) on the aircraft. In flight, the neural network makes the identical decisions that the optimizer would have made, but much faster, significantly reducing the time required to make latest decisions. The approach has proven to achieve success with UAVs of all sizes, and it may well even be used to generate neural networks which can be able to directly processing noisy sensory signals (called end-to-end learning), corresponding to the pictures from an onboard camera, enabling the aircraft to quickly locate its position or to avoid an obstacle. The exciting innovations listed here are in the brand new techniques developed to enable the flying agents to be trained very efficiently – often using only a single task demonstration. Considered one of the essential next steps on this project are to make sure that these learned controllers could be certified as being secure.
Over time, How has worked closely with corporations like Boeing, Lockheed Martin, Northrop Grumman, Ford, and Amazon. He says working with industry helps focus his research on solving real-world problems. “We take industry’s hard problems, condense them right down to the core issues, create solutions to specific points of the issue, exhibit those algorithms in our experimental facilities, after which transition them back to the industry. It tends to be a really natural and synergistic feedback loop,” says How.