PhD candidate Wei Xiao (SE) is the first author on a paper titled \u201cFeasibility-Guided Learning for Constrained Optimal Control Problems.\u201d\u00a0The piece was published in Proceedings of 59th IEEE Conference on Decision and Control<\/em> and earned the Outstanding Student Paper Award.<\/p>\n Over 1300 papers were presented at the conference, and 64 were nominated for the Outstanding Student Paper award. Xiao’s submission is one of four to be recognized for displaying\u00a0\u201coriginality, clarity and potential impact on practical applications or theoretical foundations of control.\u201d<\/p>\n Xiao co-authored the paper with his advisors: CISE faculty affiliates, Professors Christos Cassandras<\/a> (SE, ECE, Division Head of Systems Engineering) and Calin Belta<\/a> (ME, ECE, SE). The paper addresses a fundamental and challenging problem in control theory: stabilizing a dynamical system while optimizing a cost and satisfying constraints. \u201cTypically, such problems include autonomous driving in traffic and robot safe exploration in unknown environments,\u201d says Xiao.<\/p>\n \u201cWhen safety becomes critical, it is desired to prioritize the strict satisfaction of constraints over optimality.\u201d<\/p><\/blockquote>\n In the paper, the researchers adopt the control barrier function (CBF) method to improve the feasibility of optimal control problems with stringent safety constraints and tight control limitations in an unknown environment. Since control functions are often too conservative in their response to unknown aspects of the environment, the researchers apply machine learning techniques to improve the feasible regions that a system can access. With this approach, an appropriate and safe response to unpredictable obstructions is created. Xiao tested these methods on a robot for the purpose of the study, although his work could be applied to improving the ability of autonomous vehicles to navigate around obstacles, ensuring safety for passengers and surrounding individuals.<\/p>\n “This paper addresses a crucial dilemma in modern engineering: how to bridge the gap between control planning and powerful motion learning techniques?,” explains Professor Cassandras. “This paper is original in its approach to marrying classical control engineering with emerging machine learning techniques.\u201d<\/p>\n![\"SE](\"\/eng\/files\/2020\/12\/Wei-Xiao.png\")