The Incorporation of Mobile Technology in Nuclear Security and Renewable Energy - A Study in Reinforcement Learning and Nuclear Engineering Technology




Gu, Siyao

Journal Title

Journal ISSN

Volume Title



The goal of this research is to improve upon how modern sensor networks collect data through reinforcement learning. Oftentimes, static sensors are inadequate in addressing ever-changing ambient conditions, can be major stumbling blocks in nuclear sensor and photovoltaic panel placement. Such configurations are often inadequate when nuclear source lie outside of sensors' ranges or when the weather and sunlight patterns change. On the other hand, moving sensors would able to account for changes in parameters such as sunlight and absence of radiation source. Furthermore, Reinforcement Learning would be account for environmental inconsistencies such as weather pattern and black box scenarios where the radiation source patterns are unknown. This dissertation, divided into two parts. This study also introduces Multi-Dimensional First-Visit Monte Carlo Policy Evaluation, a Reinforcement Learning algorithm when multiple agents are required as well as how to apply Reinforcement Learning when the environment is stochastic. Part 1 begins with a deep dive into the background of both radiation physics and part 2, solar forecasting and the current solutions that exist in addressing both issues. Then the framework for model-based and model-free forms of Reinforcement Learning, First-Visit Monte Carlo Policy Evaluation and Actor-Critic Policy Search respectively, are laid out. And finally, the paper concludes with an application of one or both algorithms being applied to nuclear security and mobile solar farms while evaluating the end result in a systematic way.

这项研究的目标是改进现代传感器网络通过强化学习收集数据的方式。通常,静态传感器不足以应对不断变化的环境条件,可能是核传感器和光伏面板放置的主要绊脚石。当核源位于传感器范围之外或天气和阳光模式发生变化时,这种配置通常是不合适的。另一方面,移动传感器将能够解释参数的变化,例如阳光和没有辐射源。此外,强化学习将考虑环境不一致性,例如天气模式和辐射源模式未知的黑盒场景。本论文,分为两部分。本研究还介绍了多维首次访问蒙特卡洛策略评估,一种需要多个代理时的强化学习算法,以及如何在环境随机时应用强化学习。第 1 部分首先深入探讨辐射物理学的背景,第 2 部分,太阳预测以及解决这两个问题的当前解决方案。然后分别阐述了强化学习、First-Visit Monte Carlo Policy Evaluation 和 Actor-Critic Policy Search 的基于模型和无模型形式的框架。最后,本文总结了将一种或两种算法应用于核安全和移动太阳能农场,同时以系统的方式评估最终结果。


This item is available only to currently enrolled UTSA students, faculty or staff. To download, navigate to Log In in the top right-hand corner of this screen, then select Log in with my UTSA ID.


Machine Learning, Nuclear, Reinforcement Learning, Renewables, Security, Solar



Electrical and Computer Engineering