Solving Nonlinear Programs (NLPs) is useful in a wide range of engineering applications. However, NLPs are known to be difficult to solve due to their non-convexity, which prohibits digital computers from solving them efficiently.
Sequential Quadratic Programming (SQP) is the state-of-the-art algorithm for solving NLPs, which works by solving Quadratic Programming (QP) subproblems. We aim to implement the SQP algorithm as an heterogeneous computer, which consists of an analog computer and a digital computer. The analog computer is a physical analog of the QP subproblem, based on a design by UC Berkley [1].
We aim to design and manufacture the analog computer on a PCB, and demonstrate its performance and efficiency by solving a nonlinear optimal control problem using SQP.
[1] https://www2.eecs.berkeley.edu/Pubs/TechRpts/2015/EECS-2015-133.pdf