Whether it’s in science fiction movies or according to the reported sightings of members of the general public, one repeated claim about so-called flying saucers is that they possess an extraordinary degree of maneuverability. One moment they could be hovering, the next moving rapidly vertically and, the next, speeding horizontally like a jet plane. It’s a movement that screams “alien presence” because, frankly, no earthbound vehicle is capable of pulling off such feats.
Of course, that’s exactly the kind of thing that sounds like a challenge to the researchers at MIT’s renowned Computer Science & Artificial Intelligence Laboratory (CSAIL). They have designed a new type of drone which can turn on a dime from hovering like an ordinary quadcopter to swooping and gliding like a fixed-wing airplane. In doing so, they may just have solved solve some of the biggest challenges which exist with modern drones.
“We’ve developed a way for people to design and create their own custom drones that have rotors, but also have wings that let them fly like planes,” MIT CSAIL grad student Jie Xu, who took the lead on the project, told Digital Trends. “This lets them take off and land vertically like traditional multi-copter drones, but fly faster and potentially be able to carry more weight while flying.”
MIT’s CSAIL group isn’t the first time a challenge like this has been taken on. Since drones first swept onto the scene, other researchers have attempted to create similar “hybrid” drones, although none have managed to solve this to everyone’s satisfaction. It turns out that taking two immensely complex flight dynamics — one involving rotors and the other wings — and getting them to work together in harmony is pretty darn tough. While designing such drones is possible, controlling them is far from easy.
Engineers trying to achieve this have therefore typically built their experimental drones with two switchable flight systems. One is used for controlling hovering, while the other is for horizontal gliding like a plane. Each one is controlled separately. This not only makes flying the finished drone difficult; it also makes designing them expensive, time-consuming, and challenging to easily translate to other drone designs and sizes. (Hence why the overwhelming majority of drones which are commercially available remain divided into distinct fixed-wing and multi-rotor categories.)
MIT’s approach represents a promising path forward. It makes it possible to design drones of different sizes and shapes easily able to switch between hovering and gliding — and all by using a single controller.
“Our system doesn’t have to store any particular modes for hovering or gliding, and can switch between the two actions by simply updating the drone’s target velocity,” Jie Xu continued. “We’ve done this by using neural networks to be able to develop the controller design for virtually any drone you want to design. Importantly, it works for real models without any additional parameter-tuning process, which helps close the gap between drones that work in virtual simulation and those that can actually be fabricated to work in actual real-world environments.”
To create their smart drone, the researchers used reinforcement learning, a type of machine learning in which A.I. agents learn to take actions in an environment that will maximize a particular reward. Being able to intelligently respond to metrics like target velocity means that MIT’s drone can adapt to different situations without the user having to manually switch between modes of flight. As the researchers write in a paper describing the work, “Our controller does not need to differentiate between the copter and flight modes or explicitly deal with the transition between modes. For example, the controller will automatically orient a tail-sitter hybrid UAV purely based on the input velocity – it will set it to a copter orientation for lower velocities and a plane orientation for higher velocities.”
As impressive as all of this is, however, what are the chances that it results in an increase in the number of consumer drones that can offer this kind of functionality? Quite good, actually. The MIT CSAIL system is not only “mode free,” but also what the team describes as “model agnostic.” That means that the same neural network and learning algorithm proves efficient in vastly different drone configurations.
As a result, it would not be necessary to redesign control systems for each different drone developed. As part of the work, the team integrated their technology into a CAD program that allows users to select and match different drone parts to develop custom drones.
“Our hope is that this is something that would make hybrid drones accessible to anyone who wants to use one, from everyday people to companies that design consumer drones,” Xu continued. “Even someone who is not an expert in design would be able to figure out a design for a drone, wait a couple of hours for the system to compute its controller, and then be able to walk away with a ready-to-fly drone.”
- This wild new display puts a gargantuan 120-inch virtual monitor on your desk
- Parrot’s 4G-connected Anafi Ai drone is the Google Maps car of the skies
- Part Terminator, part Tremors: This robotic worm can swim through sand
- Here’s what a trend-analyzing A.I. thinks will be the next big thing in tech
- Who needs headphones? Holoplot can beam audio directly to your ears from afar