The creators of robots or shift their management to artificial intelligence, or manage yourself with the help of controllers. Researchers from the Italian Institute of technology and the University of Bonn took a different path and taught the Centauro robot to repeat the movement for a person using computer vision.
The robotic centaur was submitted at the end of July 2018. It is equipped with four legs with swivel wheels, two hands and an array of cameras to perceive the environment. The robot is designed to perform rescue operations where precise movements, therefore informed to send commands to the exoskeleton was used. This method of manipulation required of the operator prior physical training, so the developers decided to get rid of the heavy structures.
Instead of an exoskeleton it was decided to use recognition technology and repetition of movements. To do this, engineers used a 3D camera, ASUS Xtion PRO and OpenPose algorithm to detect the position of the torso, arms and legs. To bring the robot to move, the algorithm examines the received frames from the camera and creates a moving model of the operator. Its key points provide guidelines for the scheduler movements, which makes the mechanisms of the robot to move.
Obeying the instructions of the operator, the Centauro was able to move and lift a cardboard box, and rotate the lever. The algorithm OpenPose also able to transmit the movement of the fingers, so when available, the robot can work with finer details.
In addition to interacting with objects, Centauro knows how to walk on four feet and overcome obstacles. The design supports six degrees of freedom, that is, except for the implementation of the movements on the sides and raise to a height, the robot is able to move at an angle.
What method of control robots you think is the best? An opinion can be expressed in our telegram chat.
Engineers have learned to control the robot Centauro with gestures
Ramis Ganiev