Analog Solver for Embedded Model Predictive Control with Application to Quadruple Tank System

Continuous-time dynamic solvers have emerged as a viable alternative for online optimization of model predictive control (MPC). This class of solvers emulates the behavior of MPC by tracing its solution trajectories to the Karush-Kuhn-Tucker (KKT) optimality point. In this paper, we develop an analog circuit architecture for the hardware realization of such dynamic controllers. Using the gradient dynamics as a case study, we develop a prototype solver using a Field Programmable Analog Array (FPAA) and we validate its efficacy on a quadruple water tank system.