Abstract
Safety assurance plays an important role in the successful and widespread deployment of safety-critical systems. However, safety is a bare minimum requirement for safety-critical systems. Many safety-critical systems are also performance-critical systems for which a control design must not only satisfy safety constraints but also deliver a desirable closed-loop behavior with acceptable performance. When safety and performance are of concern, a popular approach that has gained a surge of attention is to blend a control barrier function with a control Lyapunov function (CLF) by solving a quadratic program (QP). However, the QP-based approaches are functions of current states, and we need to solve the optimization problem at every time. Once the conflict occurs between safety and performance, the QP algorithm relaxes the stability by adding a relaxation factor. However, inappropriate selection of the CLF might result in significant conflicts between safety and stability, which in turn requires increasing the magnitude of the relaxation factor to resolve conflicts. This, however, can compromise the convergence of trajectories to the equilibrium point. To address these issues, we presented a data-driven approach to maximize the region over which the conflict between safety and optimality is resolved in the presence of disturbance. The proposed approach can greatly improve the performance of the safe reinforcement learning algorithm.
Biography
Bahare Kiumarsi is an assistant professor in the Electrical and Computer Engineering Department at the Michigan State University. Dr. Kiumarsi earned the B.S. degree in Electrical Engineering from Shahrood University of Technology, Iran, in 2009, the M.S. degree in electrical engineering from the Ferdowsi University of Mashhad, Iran, in 2013, and the Ph.D. degree in electrical engineering from the University of Texas at Arlington, Arlington, TX, USA, in 2017. In 2018, she was a Post-Doctoral Research Associate with the Coordinated Science Laboratory, University of Illinois at Urbana–Champaign, Urbana, IL, USA. Her research is in the general area of reinforcement learning and control with specific attention to safe and resilient learning-enabled control design in challenging and unknown environments. Dr. Kiumarsi was the recipient of the 2023 Withrow Teaching Excellence Award in the Department of Electrical and Computer Engineering, Michigan State University. She also was a recipient of the UT-Arlington N. M. Stelmakh Outstanding Student Research Award and the UT Arlington Graduate Dissertation Fellowship in 2017.
