AI has appeared that predicts the behavior of drivers, pedestrians and cyclists

According to the authors of the new work, their approach predicts the trajectories of all participants in the movement. AI First

identifies all objects, then predicts limit trajectories.

The researchers said that if robots were to drive vehicles, they should be able to predict what drivers, cyclists and pedestrians would do next.

The authors of the new work created a systembehavioral prediction that first guesses the relationship between two road users - which car, cyclist or pedestrian is allowed to pass and who needs to give way. The AI ​​then uses these relationships to predict future trajectories.

It is noted that these calculations were more accurate than thoseother machine learning models. The MIT method even outperformed a recently introduced model from autonomous vehicle company Waymo. The authors note that they split the task into several smaller ones, so the process consumes less memory.

The authors named their machine learning method M2I.It uses two types of input: the past trajectories of cars, cyclists, and pedestrians moving along an intersection or four-lane road, and a map with streets and lanes. Using this information, the AI ​​determines how, according to the rules of the road, objects should move. The prediction model then attempts to predict the trajectory for all road users.

M2I gives a prediction of how these objects will move over the next eight seconds.

Read more

It has been hunted for centuries: what do we know about the planet Vulcan next to the Sun

Physicists have experimentally confirmed a new fundamental law for liquids

Astronomers have found the source of mysterious radio bursts that come from space