Watch How the Robot Finds the Ball and Makes the Putt on Its Own

Now, for the first time, a robot dubbed Golfi can find a golf ball anywhere on a green, go to it independently, and make a putt.

Although robots capable of playing golf have been created in the past until now, they have required human assistance in both setting them up and programming them to perform the proper swing.

The most well-known is perhaps LDRIC, a robot who, in 2016, made a long hole-in-one at the TPC Scottsdale golf course in Scottsdale, Arizona.

Golfi, developed by Annika Junker and her team at Paderborn University in Germany, is able to locate golf balls and wheel itself into position using data from a 3D camera that surveys a green from above.

After a camera scans the green and an algorithm roughly estimates its shape, 3000 simulated golf swings are taken from random locations in the vicinity of the green and directed at the hole using physics-based equations that account for factors such as the speed and weight of the ball and the friction of the green.

It prepares a neural network to determine the optimal force and trajectory with which a robot should strike a ball.

Watch How the Robot Finds the Ball and Makes the Putt on Its Own
Watch How the Robot Finds the Ball and Makes the Putt on Its Own

This is similar to how professional golfers may practice their swing the day before a tournament, adds Junker, who presented the robot at the IEEE International Conference on Robotic Computing in Naples, Italy, in December.

Golfi and the ball may then be put anywhere on the green, and the robot will make its way to the ball and attempt to sink it.

Golfi made almost 60% of his putts on a perfectly level, 2-square-meter indoor green. Due to its insistence on being plugged in and having its 3D camera installed above the green, the robot isn’t practical for use on outdoor greens.

But the point of Golfi isn’t to win professional events. Team member Niklas Fittkau, also of Paderborn University, explains that the goal is to demonstrate how merging physics-based models with machine learning may simplify robotic application development.

Leave a Comment