College - Author 1

College of Engineering

Department - Author 1

Electrical Engineering Department

Degree Name - Author 1

BS in Electrical Engineering

College - Author 2

College of Engineering

Department - Author 2

Electrical Engineering Department

Degree - Author 2

BS in Electrical Engineering

Date

6-2026

Primary Advisor

Jason Poon, College of Engineering, Electrical Engineering Department

Abstract/Summary

This paper presents a field-programmable analog array (FPAA) implementation for solving linearly constrained quadratic programs (LCQPs) directly in the analog domain. The solver is based on a continuous-time primal-dual control architecture with integral action, anti-windup compensation, and a piecewise-linear nonlinearity for enforcing affine inequality constraints. A switched-capacitor implementation using three AN231E04 FPAAs is developed, and coefficient scaling methods are introduced to keep internal and output signals within the voltage limits of the hardware. A global scaling factor is used to reduce internal signal excursions, while solution-space scaling is shown to modify the implemented optimization coefficients and alter the local closed-loop pole locations. The resulting hardware solver is evaluated on a benchmark two-variable LCQP and compared with ideal circuit simulation and a prior discrete-component implementation. For the benchmark problem, the FPAA implementation converges with steady-state errors of 0.30% and 2.53%. Dynamic reconfiguration through a microcontroller-controlled SPI interface is also demonstrated, enabling new LCQP coefficients to be programmed without rebuilding the analog circuit. Experimental results on additional LCQPs show that the FPAA platform can solve multiple constrained quadratic programs with repeatable convergence while improving compactness, reconfigurability, and implementation consistency relative to fixed discrete analog hardware.

Download

Share

COinS