Chinese Robotics Vendors Bypass Simulations with Real-World Reinforcement Learning
Chinese robotics company Astribot released a video demonstrating its new 'SmoothRL' technology, a reinforcement learning method enabling robots to learn precisely from their actions while continuously operating in real-world environments.
AsiaAI Publisher
·
September 7, 2026 ·
2 min read · Source: ロボスタ Robot Start · Issue #89
Robotics & Automation
East Asian Technology Intelligence
Japan & China tech news — translated, contextualized, and delivered for Western readers.
Free. Unsubscribe anytime.
This story ran in Issue #89, alongside three other stories.
📊 Featured Chart
Success rates with SmoothRL, versus initial performance (39%, 8%, 30%)
Chinese robotics company Astribot released a video demonstrating its new ‘SmoothRL’ technology, a reinforcement learning method enabling robots to learn precisely from their actions while continuously operating in real-world environments. Unlike traditional methods relying on simulations or static trials, SmoothRL directly applies reinforcement learning in live deployment, significantly improving task success rates for dynamic actions like throwing objects and fine manipulation.
This development from a Chinese firm challenges the perception that real-world, dynamic robot learning is solely the domain of Western or Japanese research. It signals a leap in practical application, potentially accelerating robot integration into logistics, manufacturing, and even household tasks, bypassing the limitations of simulation-to-real gaps.
