Sequential Quadratic Programming

Sequential Quadratic Programming (SQP) is an iterative method for solving constrained optimization problems by approximating the objective function and constraints with quadratic and linear models, respectively, at each iteration. Current research focuses on enhancing SQP's efficiency and robustness, particularly for stochastic optimization, large-scale problems (e.g., in robotics and control), and applications involving Gaussian processes or neural networks. These advancements are significantly impacting fields like model predictive control, scientific machine learning, and game theory by enabling the solution of complex, real-time optimization problems previously intractable with traditional methods.