Modeling, Identification, and Estimation of Distributed Parameter Systems Using Mobile Sensor Networks

点击翻译按钮获取中文摘要

This project will formulate a general purpose mathematical framework using mobile sensor networks (MSNs) that will allow an efficient and accurate prediction of information about behavior of complex real world systems such as weather forecasting, wildfire control, disaster recovery, explosive materials detection, etc. Such systems are known to be distributed parameter systems (DPS) and modeling of such systems and accurately predicting their behavior in real-time is highly complex and computationally challenging. The current state-of-the-art techniques use static sensor network to help obtain solutions from complex mathematical models which are often inaccurate and cannot provide timely information. The mobility and adaptability of MSNs make them great candidates for overcoming these challenges. This project will develop a unified framework for modeling, identifying, estimating and predicting behavior of such distributed parameter systems. The project will also develop a strongly integrated research, educational, and outreach program by providing graduate students with interdisciplinary and challenging research experiences, by providing undergraduate students with the opportunity of early involvement in research activities through algorithm development and test-bed experiments, and by motivating K-12 students by giving them hands-on experiences through university's Engineering Ambassadors(EA) program.The outcomes of this project are expected to advance the modeling, identification, and state estimation techniques for distributed parameter systems, offer comprehensive scientific understanding of the connections between DPS and mobile sensor networks, and contribute to generic engineering principles for designing cooperative control and distributed sensing strategies for mobile sensor networks. In particular, the work will: develop novel approaches for online parameter identification and state estimation of DPS using mobile sensor networks with reduced computational and communication cost compared to existing methods developed for static sensor networks; determine information-rich and energy-saving optimal trajectories of mobile sensor networks moving in DPS for simultaneously system identification and state prediction; and design a multi-robot test-bed with controllable advection-diffusion fields for the validation of the strategies, and conduct field experiments to test the strategies under realistic uncertainties and variations. The project also offers great educational opportunities for graduate, undergraduate, as well as K-12 students.