Shypd.aiThe largest open index of AI toolsMetaworld is an open-source benchmark for developing and evaluating multi-task and meta reinforcement learning algorithms. It provides collections of robotics environments for continuous control.
Trending
About
Metaworld is an open-source benchmark designed for the development and evaluation of multi-task and meta reinforcement learning algorithms. It focuses on continuous control within robotic manipulation environments, offering various benchmarks to assess algorithm performance in complex robotic tasks. The platform supports Python 3.8 to 3.11 on Linux and macOS, and follows the Gymnasium API for environment creation and interaction. Users can create multi-task benchmarks like MT1, MT10, and MT50, or meta-learning benchmarks such as ML1, ML10, and ML45. It also allows for the creation of custom benchmarks by combining different Meta-World environments, supporting both synchronous and asynchronous modes for varied computational needs.
Capabilities
Pricing & Plans
Open Source
Free
FAQs
Trending
