Y. Chen and D.J. Braun, Iterative Online Optimal Feedback Control, vol. 66, no. 2, pp. 566-580, IEEE Transactions on Automatic Control, 2021.
This paper introduces a data-driven feedback control method for solving finite-horizon nonlinear optimal control problems with input constraints. The approach combines an approximate system model with real measured state information to compute near-optimal control inputs along the true system trajectory rather than relying only on potentially inaccurate model predictions. The authors present an algorithm that implements this method, prove its convergence and optimality, and demonstrate its generality on systems with unknown and changing dynamics.
Why it matters: Many control problems involve models that are incomplete or inaccurate. By directly incorporating real system data into the optimization process, this method improves reliability and performance compared to traditional approaches, making it valuable for robotics and other applications with complex or uncertain dynamics.