Portfolio · Projects
Selected projects
Exact algorithms, control applications, and interactive tools developed around continuous-time optimization and control.
Continuous-time explicit optimal control
An algorithm that solves Pontryagin’s optimality conditions directly: it identifies the sequence of active constraints, computes the switching times, and constructs the critical regions without converting the model to discrete time.
Exact curved critical-region boundaries
The critical regions are characterized by the switching times, the instants at which the active constraints change. For a given sequence of active constraints and fixed switching times the optimality conditions are linear in the initial state, so every boundary, hyperplane or curved, is computed exactly from linear equations rather than optimal control problems. Whether a boundary is a hyperplane or curved is known from the two sequences of active constraints alone, before any computation.
Building the whole partition
Builds the complete map of critical regions from one region outward, without solving optimal control problems and without gridding the initial-state space, and verifies with the method’s own optimality check that no region is missing. Demonstrated on process models with up to ten states, where the map has a handful of regions and is built in seconds.
Continuous time versus discrete time
The same explicit controller built both ways, on forced-circulation evaporators in series with one, three, and five effects. The continuous-time critical regions are few, fixed, and faster to compute, and the controller holds the constraints at every instant. The discrete-time regions multiply with the number of discretization intervals and with the number of states until the computation becomes intractable, and the resulting controller is less accurate and leaves the constraints between discretization instants.
Continuous-time multiparametric MPC within PAROC
Integration of the continuous-time controller-synthesis step with process modeling, model approximation, critical-region construction, and closed-loop validation. Current work includes a binary distillation benchmark, forced-circulation evaporators in series with up to five effects, and preparation for intensified column applications.
Neon Gate Challenge
An interactive website experience that lets visitors feel the effect of timing, constraints, continuous control, and changing initial conditions before introducing the mathematical language behind the problem.