RoboKeeper: The world’s best penalty killer

| |
1 Star2 Stars3 Stars4 Stars5 Stars

It gives Messi, Neymar and the others a very tough time. With acceleration 20 times faster than a Formula 1 racing car and a response time of 300 ms, RoboKeeper saves penalties from the world’s best soccer players.

Once again, at the FIFA World Cup in Russia, many matches will be decided by penalty shootouts. Generally, it is up to the goalkeeper whether the player scores or not. He can only hope that the player kicks a bad penalty. But not RoboKeeper!

What is probably the fastest goalkeeper in the world saves almost every penalty shot. Even balls flying towards the goal low at speeds of 100 km/h or swerving Brazilian shots can’t get past RoboKeeper. At this speed, the pure flight time of the ball is about 360 ms. During this extremely short time, the artificial goalkeeper has to extrapolate the likely trajectory and the point in the goal where the ball will land and also complete all the acceleration and deceleration processes of its motor.

 To do this, the robot has an industrial Kontron ThinkIO Duo PC and a servomotor like those used in aviation. Consequently, the radial movement of the goalkeeper figure is 20 times faster than that of a Formula1 racing car. There is also a fast image processing system with two color area scan cameras from Allied Vision at the top of each side of the goal. With a resolution of 659 x 493 pixels, they deliver up to 90 images per second. The data is then evaluated with software from Fraunhofer IML on the dual core processor system from Kontron. The first data is sent to the motor controls after just three consecutive images in which ball movement towards the goal is recognized.

RoboKeeper with blobs

The algorithms investigate the image data on what are known as blobs (binary large objects) and classify the sections of the image according to size, color, and shape. If two suitable blobs are found on the ball, ThinkIO calculates the position of the ball in the space. The ball speed is then calculated by comparing two consecutive images. When the ball speed and position are known, the place where the ball “would like” to cross the goal line can be determined. Since even the trajectory is included in the calculation, RoboKeeper is also able to save swerving or bouncing balls. The values are sent via the CAN interface to the servomotor, which makes the goalkeeper figure “fly”. Consequently, RoboKeeper is also suitable as a reference project for robotic systems that use complex recognition algorithms to identify objects and to determine positions on rapidly moving conveyor belts and actuate servomotors.

RoboKeeper at the Red Square

In the RoboKeeper project, which was started by Fraunhofer IML back in 2005, the conditions differ slightly from real conditions. The scorer kicks from about 9 meters and the goal is just 2.0 x 4.0 meters. RoboKeeper has also been on the road worldwide commercially since 2007, such as at supporting programs for sports events. For example, during the 2018 FIFA World Cup, fans can try their luck on the Red Square in Moscow.


Lern more about image processing and pattern recognition at electronica 2018 in hall A3.



RoboKeeper (Image: Fraunhofer IML).

During the 2018 FIFA World Cup, fans can try their luck on the Red Square in Moscow. (Image: Fraunhofer IML).