Robot dog playing soccer and dribbling in sand, gravel, mud and snow

Although robot dog models have been known in the industry for many years, the concept of a four-legged robotic system that controls a soccer ball, dribbles around, and recovers after falling to regain possession of the ball is entirely new. The innovation is described in a preprint of an article hosted on the platform ArXiv Monday (3) by scientists from MIT.

A statement from the Massachusetts Institute of Technology states, “although the bot doesn’t show a Lionel Messi-like skill level, It is an impressive dribbling system.

The big difference with the baler robot is a system that combines sensing and onboard computing to train the device on different terrain types such as sand, gravel, mud and snow and adapt the dynamics of the ball in each of these environments. Like any good dribbling player, DribbleBot gets up after falling and tries to retrieve the lost ball.

Robot dog playing football for what?

Programming robots to play football is a very active area of ​​research. The goal is not to register the boots in a championship, but to learn how to use their legs during a dribbling, for example. may success Allows exploration of difficult skills that can be used for movement over rough terrain.

Parallel simulations of the real robot dog and various terrains were created to speed up the research. Four thousand versions of the robot are simulated simultaneously. Initially, the robot still does not know how to dribble and receives positive reinforcement when successful and negative reinforcement when unsuccessful. discovers for himself the level of strength he needs to use in his legs.

But the researchers say this is just the “start”. Lessons learned by Dribble Bot will be adapted to object handling and manipulation tasks.