1: \begin{abstract}

2: \Gls{oc} using inverse dynamics provides numerical benefits such as coarse optimization, cheaper computation of derivatives, and a high convergence rate.

3: However, to take advantage of these benefits in \gls{mpc} for legged robots, it is crucial to handle \rev{efficiently} its large number of equality constraints.

4: To accomplish this, we first (i) propose a novel approach to handle equality constraints based on nullspace parametrization.

5: Our approach balances optimality, and both dynamics and equality-constraint feasibility appropriately, which increases the basin of attraction to \rev{high-quality} local minima.

6: To do so, we (ii) modify our feasibility-driven search by incorporating a merit function.

7: Furthermore, we introduce (iii) a condensed formulation of inverse dynamics that considers arbitrary actuator models.

8: We also \rev{propose} (iv) a novel~\gls{mpc} based on inverse dynamics within a perceptive locomotion framework.

9: Finally, we present (v) a theoretical comparison of optimal control with forward and inverse dynamics and evaluate both numerically.

10: Our approach enables the first application of inverse-dynamics~\gls{mpc} on hardware, resulting in state-of-the-art dynamic climbing on the ANYmal robot.

11: We benchmark it over a wide range of robotics problems and generate agile and complex maneuvers.

12: We show the computational reduction of our nullspace resolution and condensed formulation (up to $\mathbf{47.3}\boldsymbol{\%}$).

13: We provide evidence of the benefits of our approach by solving coarse optimization problems with a high convergence rate (up to 10 Hz of discretization).

14: Our algorithm is publicly available inside \textsc{Crocoddyl}.

15: \end{abstract}

16: