SE PhD Final Defense: Yingqing Chen

SE PhD Final Defense: Yingqing Chen

TITLE: Safety-Critical Control and Optimization of Traffic Systems: Sequencing, Coordination, and Recovery

ADVISOR: Christos Cassandras (ECE, SE)

CHAIR: David Castanon (ECE, SE)

COMMITTEE: Ioannis Paschalidis (ECE, SE, BME), Roberto Tron (ME, SE), Emiliano Dall’Anese (ECE, SE)

ABSTRACT: Urban traffic systems are becoming increasingly complex, requiring control strategies that simultaneously improve efficiency while guaranteeing safety. This dissertation develops control and optimization methods for traffic systems with an emphasis on safety-critical operation, vehicle coordination, and recovery from unsafe conditions. The first part addresses adaptive traffic light control for urban intersections and traffic networks. Using stochastic hybrid system models with parametric signal controllers, Infinitesimal Perturbation Analysis (IPA) is used to derive data-driven gradient estimators of performance metrics with respect to controllable signal parameters. These estimators enable online gradient-based optimization of traffic signal timings while accounting for turning movements, transit delays, blocking effects, and pedestrian flows. The resulting framework provides scalable adaptive traffic light control capable of improving performance in multi-intersection networks. Building on the adaptive principles established in the first part, the second part focuses on coordination of Connected and Automated Vehicles (CAVs) at critical traffic locations such as roundabouts. A framework is developed that jointly determines vehicle sequencing and motion control using Model Predictive Control (MPC) integrated with Control Lyapunov–Barrier Functions (CLBFs). The controller optimizes efficiency while enforcing speed-dependent safety constraints and avoiding infeasibility issues associated with conventional barrier-function-based approaches. The framework is further extended to mixed traffic environments through a safe sequencing policy that guarantees merging safety without requiring knowledge of human-driven vehicle behavior. Finally, an exact-time safety recovery framework is introduced using time-varying Control Barrier Functions with optimal barrier tracking, guaranteeing recovery to the safe set at a prescribed time while preserving feasibility under input constraints. This recovery mechanism is applied to traffic coordination scenarios in which safety constraints are initially violated and must be restored before vehicles reach conflict points. Together, these contributions provide control and optimization methodologies for safety-critical traffic systems involving signal control, vehicle coordination, and safety recovery.