AIS: Nonlinear Statistical Control Using Neural Networks

Temple University, Philadelphia PA

Investigators

Abstract

Abstract The objective of this research is to discover new fundamental theories and algorithms for stochastic optimal control with application in solar energy transmission. The approach is based on novel statistical control with cost cumulants and neural networks. In statistical control, one views the performance index as a random variable and shapes the distribution of the performance index. By shaping the performance index, more accurate control performance is achieved. Intellectual Merit. A new optimal control paradigm of performance shaping is proposed. This paradigm will lead to a new direction of research in optimal control for the general nonlinear stochastic systems. By using artificial neural networks, this project proposes to solve the nonlinear optimal control problems. Nonlinear statistical control will allow high performance controllers for various applications. This transformative optimal control theory is applied to a high societal impact application of space solar power harvesting. Statistical control will significantly improve the solar power transmission performance. Broader Impacts. This project will advance the fundamental knowledge in optimal stochastic control. By applying a new optimal control method to the space solar power transmission application, the energy will be transmitted more accurately, which means that the energy is harvested more economically. This research will allow efficient space solar energy transmission by addressing the antenna pointing control problem. Space solar energy harvesting is an issue of national strategic importance. Furthermore, through this project and spacecraft systems engineering course, students of underrepresented groups will be educated about engineering and state-of-the-art research.