BUSINESS WIRE (SEP 2018) - Kindred, an AI and robotics company that builds intelligence for robots, has announced today the launch of SenseAct, the first open-source toolkit to set-up reinforcement learning tasks on physical robots. Kindred’s SenseAct was created to provide robotics developers and researchers with a consistent, learnable interface that efficiently controls for time delays, a factor that simulation environments aren’t hindered by.
“SenseAct is an important new step in machine learning research on robots, enabling consistent and reproducible results on physical robots for the first time. It will establish a standard that may greatly accelerate machine learning research on physical robots, pushing reinforcement learning to a new level of performance just as large standard data sets have for supervised learning,” said Richard S. Sutton, professor of Computing Science and AITF Chair in Reinforcement Learning and Artificial Intelligence at University of Alberta.
SenseAct allows reinforcement learning agents written for OpenAI’s popular “gym” simulator to learn behaviors for robots, by insulating them from the complexity of real-time control of robotic components. SenseAct’s guiding principles of minimizing delays and maximizing timing consistency via proactive computation lead to responsive learned behavior and reliable learning via state-of-the-art algorithms.
“This initial release focuses on tasks that explore timing, control frequency, action representation, partial observation, and sparse reward. We intend to update SenseAct as we define new tasks to reflect challenges in developing efficient, general, intuitive behaviors for our current and future products,” says James Bergstra, Head of AI Research at Kindred.
SenseAct allows general reinforcement learning algorithms to learn diverse tasks on diverse robots, and is made freely available for anyone to explore and extend. For more information, visit www.kindred.ai/senseact