Florida International University
Cesar Rojas was born in Miami and graduated from Miami-Dade College, and FIU, earning his A.A. in C.I.S., and B.Sc. in I.T. respectively. He graduated from FIU Summa Cum Laude through the Honors College in the Spring of 2012 and was a member of GEP, UPE, ACM, and other student organizations. After graduation, Cesar went to work for Costa Farms, Ultimate Software, and Microsoft. He left Microsoft in 2016 to pursue an M.Sc. C.S. degree at CUNY City College. He returned to FIU to finish his M.Sc. program in the Fall of 2019 and begin his Ph.D. program under Dr. Leonardo Bobadilla, researching applications of remote sensing and Secure Multiparty Computation(MPC) in the field of Robotics. Cesar is a DHS CAESCIR fellow and a GEM.
Environmental parameters (e.g., water quality, vegetation, and weather) measuring, monitoring, and estimation are critical for threat detection and mitigation. As a result, many robotic platforms with multiple sensors are deployed globally and in outer space to record measurements of observations. Modern-day monitoring systems and robots can collect in-situ data with fine spatial and temporal resolutions, but only in small areas limited by battery life, budget, time, and risks. Observational satellites collect remote sensing data over larger areas at the expense of spatial and temporal resolutions.
To extend our monitoring capabilities, we must also integrate external sources of data. There are two external sources of data: 1) data from third-party stakeholders (e.g., residents, tourists, law enforcement, and cruise ships) in the region of interest and willing to instrument a sensor package, collect, and share information provided that we can guarantee privacy. 2) Observational Satellites (e.g., Sentinel, LANDSAT, and AQUA MODIS) collecting remote sensing data. This thesis aims to create the scientific foundations and systems to integrate these data sources for wide-area, fine temporal environmental monitoring.
First, we present a method for oblivious sensor fusion to solve filtering problems. Second, we present a technique for privacy-preserving planning tasks. Second, we will show a solution for secure sharing of a policy to solve a Markov Decision Process (MDP). These methods serve as the fundamentals for Secure Multi-Party Computation (MPC) applications in robotics. Thirdly, we present a method for using remote sensing data for state estimation by incorporating a sensor-based 3-D map into an Extended Kalman Filter (EKF). We will show how remote sensing data is collected, selected, processed, and fused with in-situ data to create 3-D sensor-based maps for robot localization and navigation. Fourthly, we present a method for using remote sensing data for planning by creating a sensor-based map that identifies areas of interest for a robot. Finally, we will present a case study of monitoring Biscayne Bay by integrating these methods into a solution that fuses data from multiple robotic systems with privacy guarantees.
View on Zoom:https://fiu.zoom.us/j/93441531056?pwd=dld6Ti8rLzA3TndBZm1qWXJQVzRiQT09
Meeting ID: 934 4153 1056