近日，DeepMind公司的一项新研究取得突破，他们开发出一个乒乓球机器人智能体，其技能水平已达到业余选手级别。该机器人不仅掌握了正手和反手等基本技能，还能应对多种对手的打法和不同旋转的发球。在29场与人类选手的比赛中，机器人取得了45%的胜率，显示出其接近人类业余选手的实力。
尽管该机器人目前还无法战胜乒乓球公园里的资深选手，但它的出现无疑为人工智能领域的研究提供了新的方向。DeepMind首席科学家表示，乒乓球机器人有助于解决高速控制和感知问题，为机器人的技术提升提供了有价值的基准。
该智能体由低级技能库和高级控制器组成，通过模拟环境和真实世界的数据收集，结合强化学习技术，实现了技能的自动提升。研究还提出了一种分层和模块化的策略架构，帮助智能体在比赛中实时适应未见过的对手。
未来，随着技术的不断进步，我们有望看到机器人乒乓球技能的进一步提升，甚至可能在未来挑战专业选手。这一研究不仅推动了人工智能在体育领域的应用，也为机器人技术的发展提供了新的动力。
英语如下：
Title: “DeepMind’s Robot Ping Pong Breakthrough: Struggles Against Advanced Players Despite Victory Over Beginners”
Keywords: DeepMind, Table Tennis, Robot
News Content:
Title: DeepMind’s Robot Table Tennis Skills Reach Amateur Level, Paving Way for Enhanced Robotic Abilities
Recent research by DeepMind has achieved a breakthrough with the development of a table tennis-playing robot that has reached an amateur level of skill. The robot has mastered basic techniques such as forehand and backhand shots and can adapt to various playing styles and different types of spin serves. In 29 matches against human players, the robot achieved a winning rate of 45%, showcasing its strength approaching that of amateur human players.
While the robot is currently unable to defeat advanced players at a table tennis park, its emergence certainly offers a new direction for research in the field of artificial intelligence. The chief scientist at DeepMind noted that the table tennis robot aids in solving problems related to high-speed control and perception, providing a valuable benchmark for the enhancement of robotic technology.
The intelligent agent is composed of a low-level skill library and a high-level controller, achieving automatic skill improvement through data collection from both simulated environments and real-world situations, combined with reinforcement learning techniques. The research also proposes a hierarchical and modular strategy architecture to help the agent adapt in real-time to opponents it has never encountered before during matches.
Looking ahead, as technology continues to advance, we can expect to see further improvements in the robot’s table tennis skills, potentially even challenging professional players in the future. This study not only advances the application of artificial intelligence in sports but also provides new impetus for the development of robotics technology.
【来源】https://www.jiqizhixin.com/articles/2024-08-09-3
Views: 2