Researchers University of California, Berkeley (UC Berkeley), Université de Montréal and Mila developed a hierarchical reinforcement learning framework to improve the precision of quadrupedal robots in soccer shooting, Tech Xplore reported.
Zhongyu Li, one of the researchers who conducted the study, told Tech Xplore that their goal is to enable four-legged robots to achieve the ability to play soccer, just like humans. Their inspiration came from a league in the robotics community called RoboCup (Robot World Cup), which gives researchers around the world the opportunity to train their robots to play soccer games
Recent advancements have enabled the creation of more reliable hardware and advanced control algorithms for robots. As a result they could potentially tackle more complex tasks, including playing soccer alongside humans.
The new framework has two key components: a motion control policy and a motion planning policy. The motion control component allows the robot to track an arbitrary trajectory for the toe on its kicking leg. The motion planning policy, on the other hand, selects an optimal toe trajectory to shoot a nearby soccer ball (detected by an external camera) to a target location (e.g., the goalpost).
The researchers’ goal is to first focus on learning the control policy that works on the hardware. Li explains that in order to shoot the ball accurately at real-world targets, the planner is trained using real-world data when the robot shoots a real soccer ball.
The researchers also say that there are peculiarities in this sport, as the soccer ball brings more challenges because the robot has to deal not only with the hard-to-model soft contact with the deformable ball, but also with the uncertainty of the rolling friction between it and the ground. Thanks to their developed methodology, robots will be able to handle such problems as well as tasks where they need to interact with soft objects such as balls, ropes, straps, clothes, etc.
The scientists’ long-term goal is to create four-legged robot soccer players that could one day compete with humans. They hope to start a fully autonomous soccer game using these robots in the future.
In the future, the framework created by this team of researchers could be used to improve the performance of robots in soccer tournaments, particularly Robocup. Meanwhile, Li and his colleagues plan to devise other frameworks and machine learning models to improve the performance of robots in other elements of soccer playing.