Fuzzy SARSA learning (FSL) is a robust reinforcement learning (RL) technique that represents successful solutions in various industrial problems. Water management in irrigation canals is one of these problems where an FSL agent interacts with the canal environment to control the gates. FSL often requires a large number of interactive experiences and takes a long time in real-life problems. To reduce the iteration and speed up the learning process, an enhanced FSL (EFSL) was developed to accelerate the process of policy learning. A MATLAB program was written, and combined with the irrigation canal conveyance system simulation (ICSS) model. To evaluate the proposed idea, the E1R1 Dez canal, located in the south-west of Iran, was considered as a case study. Standard performance indicators were used for assessing the results based on considered water delivery scenarios, showing a shorter learning time with reasonable performance in controlling water depth changes within the canal.

中文翻译：

---

使用增强型模糊 SARSA 学习的灌溉渠控制

模糊 SARSA 学习 (FSL) 是一种强大的强化学习 (RL) 技术，代表了各种工业问题的成功解决方案。灌溉渠中的水管理是这些问题之一，其中 FSL 代理与渠环境交互以控制闸门。FSL 往往需要大量的交互体验，在现实生活中的问题需要很长时间。为了减少迭代并加快学习过程，开发了增强型 FSL (EFSL) 来加速策略学习过程。编写了一个MATLAB程序，并与灌渠输送系统仿真（ICSS）模型相结合。为了评估提议的想法，将位于伊朗西南部的 E1R1 Dez 运河作为案例研究。