Drone Races in the Sim-to-Real Gap: The Avatars of Artificial Intelligence to Beat Humans

The machines have beaten humans at games for many years, but they always trained in the same condition in which they compete. In chess for example, the board can be simulated exactly. Now though, researchers have demonstrated an AI that can beat humans in a place where simulation can only take you so far, the real world. The Swift AI system is able to race drones against champion-level humans, and beat them most of the time. The researchers are hoping this research will help improve the efficiency of drones.

Computers have been beating humans at their own games for quite a while now. In 1997, IBM’s Deep Blue bested Garry Kasparov at chess. In 2016 Google built a program using artificial intelligence that could beat world champion Lee Sedol at the game of Go. AI programs have also bested humans at poker and several video games.

But every one of these competitions has taken place on a board or at a desk. The computers haven’t been able to beat people. Kaufmann says that’s because it’s much harder to simulate real-world conditions if you’re flying a drone than if you’re playing a game on a board. “This is called the sim-to-real gap,” he says.

A variety of strategies were used to overcome the gap. Kaufmann taught the drone what racing gates looked like by hand-identifying the fabric gates in tens of thousands of images — a technique known as “supervised learning.” The team also used more conventional code to help the drone triangulate its position and orientation based on visual cues from its cameras.

“That means that all gates are in the correct sequence,” says Leonard Bauersfeld a PhD student at the University ofZURICH.

The Challenge of Running an Atomic Nucleus in a Virtual World: A Commentary on the Discovery of Oxygen 28, an Implication for Physicists

There are a lot of limitations with the drone. It only works for the specific course it’s been trained on and in a specific environment. Moving the course from inside to outdoors, for example, would throw the drone off due to changes in lighting. And the slightest things can send it spinning. Bauersfeld says if a rival bumps it it has no idea how to take care of it.

Bauersfeld says that lack of flexibility is one of the main reasons this technology can’t be easily made into a killer military drone.

In an accompanying commentary in Nature, Guido de Croon, a researcher at Delft University in the Netherlands says that the new technology has a way to go.

“To beat human pilots in a racing environment, the drones have to deal with external events such as the wind and changing light conditions, as well as with other racing drones and many other factors,” he writes.

Even if humans aren’t ready, the little drone still shows that the machine is prepared to jump from the virtual world into the real one.

Oxygen 28 is a type of Oxygen that has an atomic mass of 20 neutrons and eight protons. This strange isotope has long been sought after by physicists, as its proposed unusual properties would allow them to put their theories of how atomic nuclei work to the test. Physicists believe they have seen oxygen 28 in decades of experiments. The observations imply that more physicists aren’t aware of how atomic nuclei are held together.

This time, the Indian Space Research Organization’s successful moon landing, and the low level of support offered to researchers whose first language isn’t English by journals.