SmoothRL: Asynchronous Online Reinforcement Learning

SmoothRL: efficient online reinforcement learning with asynchronous execution. Continuous learning. The robot keeps moving and the model keeps inferring; SmoothRL learns precisely from the actions executed on the real robot. Uncut footage and technical report: https://www.astribot.com/research/SmoothRL #Astribot #SmoothRL #RL #physicalAI #robot