1: \begin{abstract}

2: Optimal control is a widely used tool for synthesizing motions and controls for user-defined tasks under physical constraints.

3: A common approach is to formulate it using direct multiple-shooting and then to use off-the-shelf nonlinear programming solvers that can easily handle arbitrary constraints on the controls and states.

4: However, these methods are not fast enough for many robotics applications such as real-time humanoid motor control.

5: Exploiting the sparse structure of optimal control problem, such as in \gls{ddp}, has proven to significantly boost the computational efficiency, and recent works have been focused on handling arbitrary constraints.

6: % [wxm] Recent work did more than just general inequality, e.g., there is work on projection based for equality

7: Despite that, \gls{ddp} has been associated with poor numerical convergence, particularly when considering long time horizons.

8: One of the main reasons is due to system instabilities and poor warm-starting (only controls).

9: This paper presents \emph{control-limited Feasibility-driven DDP} (Box-FDDP),

10: %This paper presents Box-FDDP, 

11: a solver that incorporates a direct-indirect hybridization of the control-limited \gls{ddp} algorithm.

12: Concretely, the forward and backward passes handle feasibility and control limits.

13: We showcase the impact and importance of our method on a set of challenging optimal control problems against the Box-\gls{ddp} and squashing-function approach.

14: \end{abstract}

15: