Résumé
We propose a control framework for torque control of humanoid robots for efficiently minimizing the tracking error of multi-objective weighted tasks while satisfying dynamic constraints using a Quadratic Programming (QP) solver, such that this optimal dynamically-feasible reference (considering the state of the floating base) can be robustly tracked with exponential convergence in the presence of non modeled torque bias and low-frequency bounded disturbances. This is achieved by introducing integral gains in a Lyapunov-stable torque control which exploit the passivity properties of the dynamical model of the robot, and by the proper consideration of their effect on the dynamic constraints of the QP solver. The robustness of this framework is demonstrated in simulation by commanding our robot, the HRP-5P, to achieve simultaneously several objectives in configuration and Cartesian space, in the presence of non-modeled static and kinetic joint friction, as well as an uncertain torque scale.