24.01.2024
skrl: A Modular and Flexible Library for Reinforcement Learning
Abstract
skrl is an open-source modular library for reinforcement learning written in Python. It is designed with a focus on readability, simplicity, and transparency of algorithm implementations. In addition to supporting environments that use the traditional OpenAI Gym interface, skrl allows loading, configuring, and operating NVIDIA Isaac Gym environments. This enables the parallel training of several agents with adjustable scopes, which may or may not share resources, in the same execution. The library is built on top of PyTorch and JAX, and it is highly modular and flexible.
For more information, please visit skrl.readthedocs.io – its source code is available on GitHub.
Bio
Nestor Arana-Arexolaleiba (PhD CNRS-LAAS – Toulouse-France - 2002) is a researcher at the Robotics and Automation Research Group at Mondragon Unibertsitatea in Basque Country. He was a guest researcher at Aalborg University in Denmark from 2018 to 2019. His research focuses on collaborative robotics, machine learning, and image processing for sustainable manufacturing and remanufacturing. He has a keen interest in human-robot collaboration and has researched deep learning strategies like reinforcement learning for collaborative robot. He has two patents and 50+ publications on Researchgate or Google Scholar.
Nestor also teaches in the Robotics and Automation Master program, which covers ROS, Robotics AI-based Control, and Mobile Robotics. He also participates in seminars on Collaborative Robots and Image Processing for companies in the context of the Basque Digital Innovation Hub.
Photo: Private
skrl is an open-source modular library for reinforcement learning written in Python. It is designed with a focus on readability, simplicity, and transparency of algorithm implementations. In addition to supporting environments that use the traditional OpenAI Gym interface, skrl allows loading, configuring, and operating NVIDIA Isaac Gym environments. This enables the parallel training of several agents with adjustable scopes, which may or may not share resources, in the same execution. The library is built on top of PyTorch and JAX, and it is highly modular and flexible.
For more information, please visit skrl.readthedocs.io – its source code is available on GitHub.
Bio
Nestor Arana-Arexolaleiba (PhD CNRS-LAAS – Toulouse-France - 2002) is a researcher at the Robotics and Automation Research Group at Mondragon Unibertsitatea in Basque Country. He was a guest researcher at Aalborg University in Denmark from 2018 to 2019. His research focuses on collaborative robotics, machine learning, and image processing for sustainable manufacturing and remanufacturing. He has a keen interest in human-robot collaboration and has researched deep learning strategies like reinforcement learning for collaborative robot. He has two patents and 50+ publications on Researchgate or Google Scholar.
Nestor also teaches in the Robotics and Automation Master program, which covers ROS, Robotics AI-based Control, and Mobile Robotics. He also participates in seminars on Collaborative Robots and Image Processing for companies in the context of the Basque Digital Innovation Hub.
Photo: Private