Foosbar: Autonomous table soccer robot shoots almost unstoppably
Can a robot beat a human at table soccer? A young developer has developed such a robot and tried it out.
The Foosbar robot plays table soccer autonomously against a human.
(Image: From Scratch (Screenshot))
The young developer Xander Naumenko has built a table soccer robot called Foosbar that can pass, block and shoot faster than a human player. Naumenko essentially uses computer vision with infrared cameras to better follow a specially prepared ball.
More than 500 hours of tinkering went into the project. Naumenko's aim was to build a table soccer robot that can handle the ball as precisely and sensitively as possible in order to better outplay the human player. The developer was not primarily interested in speed when passing, blocking and shooting. This was more or less a matter of course.
Naumenko built the Krökel table on a Tornado kicker T-3000. It provides the necessary high accuracy of the playing figures and the playing field. He automated one player side with a two-motor system that can move the goalkeeper, defender, midfield and striker rods horizontally as well as rotate them. Both work independently of each other and simultaneously. Naumenko used two Tecnic Clearpath stepper motors: a 2331P-RLNA for rotation and a 2310S-RLNA for moving the bar. The advantage of these motors is that they are responsive, powerful and relatively easy to control. The rod is moved by means of a drive belt that moves a carriage on which the motor for the rod rotation is mounted.
The motors are controlled via a conventional desktop PC using USB controller boards. The motors receive their movement commands via an API. This allows precise and fast positioning of the rod and therefore the playing figures.
Ball tracking via computer vision
To be able to track the ball at all times, Naurenko relies on an array of infrared cameras that can observe the ball from different angles. Previously, the developer tried a single camera from above and through a transparent playing field from below. However, both methods turned out to be unreliable, as the ball could not be tracked at all times. For better tracking, the ball was therefore prepared with a special reflective paint. As long as at least two infrared cameras have the ball in view, the exact position of the ball on the pitch can be determined.
In order to be able to play the ball in every game situation, the table soccer robot must first be taught certain game tactics. To do this, the robot must know whether to attack or defend depending on the position of the ball. In attack mode, for example, it must be able to pass back and forth between the individual poles. The striker bar is then mainly used to take a shot on goal. The focus in attack mode is on positioning the players as accurately as possible, rather than on speed.
Empfohlener redaktioneller Inhalt
Mit Ihrer Zustimmmung wird hier ein externes YouTube-Video (Google Ireland Limited) geladen.
Ich bin damit einverstanden, dass mir externe Inhalte angezeigt werden. Damit können personenbezogene Daten an Drittplattformen (Google Ireland Limited) übermittelt werden. Mehr dazu in unserer Datenschutzerklärung.
However, the situation is different in defense. Here, too, the pieces have to be positioned precisely in order to block the attacker's shots, but high reaction speed plays an overriding role.
The high quality of the T-3000 foosball table makes precise ball control possible in the first place, says Naurenko in the video. He had previously tried out the system with a conventional Krökel table. However, the inaccuracies of this table did not always allow the robot to play a precise passing game.
Reaction speed also plays a major role when scoring goals: the robot moves the playing pieces on the bar so quickly that a human can hardly react.
Room for improvement
During an initial test game, some of the system's problems became apparent. Whenever the ball bounced chaotically around the pitch, the robot system was unable to react quickly enough and the human player was able to take advantage of this. The marking color of the ball also wore off in the course of the game, so that the computer vision system could not always track the ball correctly.
Overall, however, the Foosbar robot cuts a good figure when playing table football. However, it is probably no match for a professional player. Naumenko has made the Foosbar robot open source. Interested parties can find instructions on Github.
(olb)