A Sequential Quadratic Programming Perspective on Optimal Control
By: Abhijeet, Suman Chakravorty
Potential Business Impact:
Finds best robot moves, always improving.
This paper offers a unified perspective on different approaches to the solution of optimal control problems through the lens of constrained sequential quadratic programming. In particular, it allows us to find the relationships between Newton's method, the iterative LQR (iLQR), and Differential Dynamic Programming (DDP) approaches to solve the problem. It is shown that the iLQR is a principled SQP approach, rather than simply an approximation of DDP by neglecting the Hessian terms, to solve optimal control problems that can be guaranteed to always produce a cost-descent direction and converge to an optimum; while Newton's approach or DDP do not have similar guarantees, especially far from an optimum. Our empirical evaluations on the pendulum and cart-pole swing-up tasks serve to corroborate the SQP-based analysis proposed in this paper.
Similar Papers
A Quadratic Control Framework for Dynamic Systems
Systems and Control
Makes robots follow paths perfectly.
Optimal Output Feedback Learning Control for Discrete-Time Linear Quadratic Regulation
Systems and Control
Teaches robots to learn how to control things.
Beyond Quadratic Costs in LQR: Bregman Divergence Control
Systems and Control
Makes robots smarter and safer with new math.