The algorithms by which robots calculate the trajectory of their movement from point “A” to point “B” exist long ago. There is nothing surprising. However, the movement of the robot in place, with the animated movement of pedestrians, is a much more complicated task. After all, you need to move according to generally accepted rules and norms, trying not to get in the way of people. Engineers at the Massachusetts Institute of Technologydeveloped an algorithm just for such cases and successfully tested it.
The robot, which participated in the field trials, was equipped with several RGB-cameras, as well as a scanning sensor LiDAR. The software provided for the recognition of the faces of the surrounding people in order to respond to their appearance in the field of view of the robot and to track the trajectories of pedestrians moving relative to its current location. The robot neatly paves the route, trying to go around static obstacles and do not interfere with the people ahead. Its average speed is 1.2 meters per second.
In this case, the robot not only monitors obstacles in its path and calculates trajectories of pedestrians, but also tries to adhere to social norms adopted in a particular country. In the US, as in Russia, people are used to right-wing traffic. Therefore, the robot tries to cling to the right side of the openings and corridors that cross. Settings can be quickly changed to alternate, after which the robot will accept the normal movement on the left side.
Sponsor the project MIT undertook the American automobile company Ford. The results of the research will be announced at the forthcoming conference of the Institute of Electrical and Electronics Engineers dedicated to the development of robotics, which will be held next month.