Researchers in the Hybrid Robotics Group at UC Berkeley, Simon Fraser University and Georgia Institute of Technology recently created a reinforcement learning model that allows a four-legged robot to efficiently play soccer in the role of goalkeeper. The model, presented in a paper previously published on arXiv, improves robot skills over time, through a process of trial and error.
“By letting the quartet play footballWe can push the boundaries of artificial intelligence for legged sports robots,” Xiaoyu Huang, Zhongyu Li, Yanzhen Xiang, Yiming Ni, Yufeng Chi, Yunhao Li, Lizhi Yang, Xue Bin Peng, and Koushil Sreenath, the researchers who carried out the study, told TechXplore. Goalkeeping is an exciting but challenging task that requires Robot To respond to a fast-moving ball, sometimes flying in the air, and intercept it using dynamic maneuvers in a very short period of time (usually within 1 second). By solving this, we can also gain insight into how to create intelligent and dynamic robots. “
The main goal of recent work by Huang and his colleagues has been to create a four-legged robotic goalkeeper that can improve his skills while playing, just as a human goalkeeper does. To do this, the researchers developed a reinforcement learning model that trains the robot via a trial-and-error process, rather than through a fixed, human-designed strategy.
“The robot first learns different movement control policies to form distinct skills, such as swerving, diving and jumping, while tracking the random trajectories of the robot’s fingers,” the researchers explained. “Based on these control policies, the robot then learns a high-level planning policy to determine the optimal skill and motion to intercept the ball after examining the position of the detected ball and the states of the robot.”
The researchers trained their reinforcement learning model in a series of soccer game simulations. Then, they published the policies they learned on the Mini Cheetah, a four-legged robot developed at the Massachusetts Institute of Technology (MIT) and tested its performance in the real world.
The reinforcement learning framework created by Huang and colleagues has been found to improve the abilities of the Mini Cheetah robot as a soccer goalkeeper. in the team The real world In tests, the robot was able to save 87.5% of 40 random shots.
“I think the coolest aspect of our work is that using our proposed method, the four-legged Mini Cheetah robot is able to perform highly dynamic and agile locomotion skills, such as jumping and diving, as well as quick and precise manipulation skills, such as pushing the ball away using its swinging legs in a short period of time. very “. “This actually pushes the limits of leg movement, which shows that the leg can also be a manipulator, just as humans can be.”
In the future, the reinforcement learning model created by this team of researchers could be used to improve the performance of bots designed to participate in the RoboCup and other robotic football competitions. In addition, the Model They can be used to improve the agility and physical capabilities of quadrobots designed to handle very different tasks, such as search and rescue missions.
“We hope that we can enable four-legged robots to compete with human football players in the near future,” the researchers added. “Robots need to perform a greater range of dynamic movements, agility and achieve more intelligence in the game of football.”
Xiaoyu Huang et al, Creation of a dynamic four-legged robotic goalkeeper with reinforcement learning.
arXiv: 2210.04435v1 [cs.RO]And the arxiv.org/abs/2210.04435
© 2022 Science X Network
the quote: A four-legged learning-based robot goalkeeper (2022, Oct 25) Retrieved Oct 25, 2022 from https://techxplore.com/news/2022-10-learning-based-four-legged-robotic-goalkeeper.html
This document is subject to copyright. Notwithstanding any fair dealing for the purpose of private study or research, no part may be reproduced without written permission. The content is provided for informational purposes only.
#fourlegged #robotic #goalkeeper #based #learning