用遗传模糊树提升太空机械臂控制，更高效更可靠。
A Genetic Fuzzy-Enabled Framework on Robotic Manipulation for In-Space Servicing
- 将遗传模糊树与LQR结合，构建可信高效的控制器。
- 平均性能比最优LQR高18.5%，且对不确定性极鲁棒。
- 适合需要高安全性的在轨卫星维护场景。
随着轨道卫星数量增加，在地月空间进行机器人服务的自动化变得愈发重要。安全是执行卫星维护的关键，因此所采用的控制技术必须既可信又高效。本文通过泰勒斯的TrUE AI工具包，将遗传模糊树与广泛应用的LQR控制方案相结合，为一个二维平面机械臂设计了一种可信且高效的控制器，该机械臂理论上可用于执行卫星维护任务。实验发现，遗传模糊-LQR控制器在平均性能上比最优LQR高出18.5%，并且对不确定性具有极强的鲁棒性。
原文摘要 · Abstract (English)
Automation of robotic systems for servicing in cislunar space is becoming extremely important as the number of satellites in orbit increases. Safety is critical in performing satellite maintenance, so the control techniques utilized must be trusted in addition to being highly efficient. In this work, Genetic Fuzzy Trees are combined with the widely used LQR control scheme via Thales' TrUE AI Toolkit to create a trusted and efficient controller for a two-degree-of-freedom planar robotic manipulator that would theoretically be used to perform satellite maintenance. It was found that Genetic Fuzzy-LQR is 18.5% more performant than optimal LQR on average, and that it is incredibly robust to uncertainty.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。