MIT and Ford create delivery robots that navigate without GPS

Recently, more and more companies are doingBet on the use of courier robots. In order for them to be able to deliver the parcel to the addressee, they, of course, need to somehow orient themselves. The solution is simple - use GPS coordinates. In most cases, this requires a preliminary mapping of the location so that the robot knows where to go. But this causes problems from the side of data security and confidentiality, because the signal can be intercepted. Is there any way to make robots navigate the terrain somehow differently? Engineers from MIT and Ford Motor Company have found such an opportunity.

Drones that don’t need GPS is something new

How to find a location without using GPS

Massachusetts Technological SpecialistsThe institutes have developed a technology that allows robots to navigate along "prompts", focusing on the terrain and without resorting to communication with GPS satellites.

See also: In MIT created ink that can change colors like a chameleon

In fact, this system works as well asour memory with you. When we go through a locality in which we have already been, but remember only some common features, we focus on what we remembered during our last visit best and thus find the way. For robots, prompts can be described in general terms, such as a “front door” or “garage”.

A robot can be trained to know that, for example,a sidewalk branch leads to a path that is likely to lead to the front door. Says Michael Everett, a graduate student at the Department of Mechanical Engineering at the Massachusetts Institute of Technology. Even if the robot delivers the package to an area in which it has never been, hints added to the navigation system in advance will help them orient themselves. In addition, some sites may be similar to those that he has already seen. As is the case with the track already mentioned. The robot remembers such patterns, which allows him to better navigate in the conditions of the city.

Already uses a new approach in his workexisting algorithms that evaluate and analyze incoming information that comes through cameras. As a result of this, everything that the robot sees breaks down into semantic clues such as “sidewalk”, “staircase”, “front door” and so on. Also, the distance to the objects is calculated, and the algorithm itself has a character recognition system. For example, house number or street name. As a result, it turns out that if, let’s say, to the robot that he needs house number 9 at the end of the street, and the parcel should be left near the stairs, he will figure out how to get to the target faster. And in order to read a lot of interesting news, you don’t need to calculate anything. It is enough to subscribe to us in Telegram.

Researchers have used the abovealgorithms for building maps, and allowed the robot to make decisions and determine the most effective way to the destination. As a result, the car delivered goods without using GPS directly to its destination. At the same time, the algorithm worked quite accurately and was slightly inferior to ordinary courier robots in delivery speed. The algorithm itself, of course, can be used not only for delivery, but also, for example, to create reconnaissance drones that will work in conditions with a poor level of GPS signal.