CSR: Small: Human-Centered Synthesis of Provably Correct Controllers for Hybrid Systems

点击翻译按钮获取中文摘要

Hybrid systems are a mathematical framework that has been widely used to model engineering and natural systems whose dynamics have both continuous and discrete aspects. Applications range from modern "fly-by-wire" airplanes to cellular biology. A controller for a hybrid system observes the execution of the system and issues input signals and/or trigger events in the hybrid system so as to meet the system's goals and safety requirements. An outstanding problem is that for complex hybrid systems existing methods for controller synthesis require exceedingly large computational resources.This project seeks to address the high computational demands of constructing controllers for hybrid systems in a new way. The novelty of the approach lies in using human-centered computing (crowdsourcing), while ensuring that the obtained controller is correct. The expected outcomes of this project are mathematical theory and algorithms that will enable the idea above for a wide range of problems, including those involving nonlinear systems and random events. These results are further implemented in simulator computer games, which will be deployed online. The amount of computing resource that can be potentially be harvested through crowdsourcing is very high. Earlier crowdsourcing applications such as reCAPTCHA, FoldIt, and Wikipedia have recorded billions of hours of effort by their human contributors. Beyond solving the controller synthesis problem, the outcomes of this project can lead to new ways to tap into this resource for solving new problems.The broader impacts of this project include improvements in the safety of a wide range of engineered systems. The project also has an educational outreach program involving K-12 students and educators in cutting edge multidisciplinary research. Uniquely, the gaming theme is an ideal vehicle to engage pre-college students in science and engineering.