A team of Swiss scientists have developed and trained an AI-driven robot that can play Badminton against humans.
A video shared on 31 May shows the four-legged robot playing Badminton against a human opponent with precise coordination and accuracy.
Learning by trial and error
Named ANYmal-D, the robot was developed by scientists at ETH Zurich and detailed in Science Robotics.
Through reinforcement learning — a type of AI that improves decision-making through repeated attempts — it is able to make better decisions by repetition.
It combines visual and tactic control to be able to track the shuttlecock and hit it accurately, making it a huge new breakthrough.
The robot is also supported by a perception noise model which was trained on real-world camera data to help it maintain consistent performance.
A member of the team that developed the Badminton robot, Yuntao Ma commented:
“We introduced a perception noise model that maps the robot’s motion to perception quality. And this allowed the reinforcement learning algorithm to automatically balance between the robot’s agile motion and a reliable perception”.
This is the first of its kind, but the idea can be applied to other robotic use cases. According to Ma:
“The control algorithm also generalizes to other robot platforms such as humanoids and also other tasks such as search and rescue, and home services.”
He added that the creation of the Badminton robots reveals the great potential for advancements in autonomous, intelligent systems, including humanoids.
Robots in sports
Robot designers have started giving more attention to sports recently, with more robots being designed with capacity to participate in sporting activities.
In China for example, robots have been enrolled in one of its key sporting events, the Beijing Half-Marathon.
With this trend, we may be seeing more robots designed specifically for sports and competitions against human opponents.
ReadNext
