Astribot's foundation model team has released an online reinforcement learning framework called SmoothRL, which is specifically designed to address the critical issue of speed mismatch between asynchronous large model inference and the real-time control demands of robots. The framework innovatively allows the reinforcement learning algorithm to continue executing actions even before the model returns its inference results, thereby eliminating the motion stuttering that typically occurs when the robot has to wait for model outputs. According to the team, SmoothRL has been thoroughly validated on both simulation and real robot platforms, demonstrating its effectiveness in reducing motion stuttering and improving response speed as well as task success rate in dynamic environments. This news was reported by Techub News.
Astribot, the robotics company behind the Astribot brand, has rolled out SmoothRL, an online reinforcement learning framework built by its foundation model team. It goes after a hard robotics problem: the speed gap between asynchronous large-model inference and the split-second timing robot control demands. SmoothRL lets the reinforcement learning agent keep carrying out actions before the large model’s inference results come back. No waiting around. That removes the stuttering that happens when the system pauses for model outputs.
The team says SmoothRL has been tested in both simulations and on real robot platforms. The results, they say, show the framework cuts motion stuttering and lifts the robot’s response speed and task success rate in dynamic environments. Techub News reported the story, citing Qubit.
This article was originally published by Bit.Fan. For more cryptocurrency news and market insights, visit www.bit.fan. Disclaimer:
The market information, project data, and third-party content displayed on this platform are for industry information sharing only and do not constitute any form of investment advice or return commitment.
Cryptocurrency trading carries high risks. Users should fully assess their risk tolerance and make independent decisions. All profits, losses, and legal responsibilities are borne by the users themselves.
