BLOM: The Berkeley Library for Optimization Modeling

Abstract

We present the Berkeley Library for Optimization Modeling (BLOM), an open-source tool for optimization-based modeling and control formulation implemented in Simulink. The underlying structure for BLOM is a novel way of representing linear and nonlinear mathematical functions that allows for easy computation of closed form gradients, Jacobians and Hessians. This formulation provides an efficient problem representation for optimization-based modeling and is scalable to large optimization problems. With BLOM, an optimization-based controller for a dynamic system can be developed and exported from the same model that is used in forward simulation. BLOM is capable of solving several types of optimization problems, including static optimization problems and optimization problem with dynamics. Its intended use is for nonlinear model predictive control. We present results where BLOM is able to handle problems with tens of thousands of variables.