CRII: CPS: High-Performance Adaptive Hybrid Feedback Algorithms for Real-Time Optimization and Learning in Networked Transportation Systems

University Of Colorado At Boulder, Boulder CO

Investigators

Abstract

Recent technological advances in the areas of sensing, computation, and communication, as well as the modern data revolution, are bringing transformative changes to the operation and control of several engineering systems that are tightly connected with society. In particular, intelligent transportation systems stand as prominent examples of socially integrated engineering systems where the combination of large-scale data sets, automation, and feedback control could have a tremendous societal impact in terms of improvement in travel efficiency, safety, and reliability. Nevertheless, to fully exploit this potential, different fundamental open questions at the level of algorithmic design and analysis need to be rigorously addressed. These questions include how to design data-driven feedback-based optimization and control algorithms that are robust, efficient, and suitable for deployment in highly dynamic and complex systems; and how to certify and characterize the fundamental limitations of these algorithms. Motivated by these questions, this project will develop novel analytical and data-driven algorithmic tools for the real-time optimization and control of intelligent transportation systems. The proposed research is motivated by two particular applications where feedback-based algorithms have shown tremendous potential: dynamic pricing control, which has recently been implemented in several cities across the United States, Europe and Asia; and urban traffic light control, which is now a fundamental infrastructure technology in dense urban cities. This project includes outreach and mentoring of high school and graduate students, as well as an active dissemination of the results via educational initiatives. The main objective of this project is the design and analysis of novel adaptive and robust data-driven control and optimization algorithms with provable performance guarantees in transportation systems. Unlike traditional approaches, the algorithms will incorporate acceleration to maximize exploitation, while simultaneously using information-rich data sets to relax traditional aggressive exploration requirements. All of this, without sacrificing structural robustness properties, which are fundamental for a safe interconnection with the dynamics of the transportation network. To achieve this, the feedback-based algorithms will be developed and analyzed using tools from hybrid dynamical system's theory, which is suitable for the study of complex cyber-physical systems. The adaptive nature of the proposed algorithms will allow them to cope in real-time with the complex discrete-time and continuous-time dynamics that emerge in transportation networks. To certify closed-loop stability and robustness properties, the project will exploit time scale separation techniques and passivity tools. The performance of the algorithms will be validated by extensive experimental and numerical simulations in realistic environments. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.