Preprint. This study develops continuous-time primal-dual methods for tracking solutions to constrained optimization problems that change over time. It establishes tracking bounds for strongly convex-concave problems and quantifies the bias and constraint violations introduced by regularization in convex settings. For smooth nonconvex problems, the authors combine local convexification with a two-time-scale flow and derive local tracking guarantees under explicit regularity conditions. Numerical experiments at a traffic intersection compare the approach with standard signal-control schemes, considering both vehicle throughput and pedestrian-aware objectives.