Global motion control and support base planning

Advanced humanoid robots capable of operating in complex 3D environments likely utilize an online optimization strategy where joint accelerations are varied to achieve whole-body postural balance. To this end, we propose one such strategy that optimizes global body parameters such as spin angular momentum and body principal angles, or the angles between the inertia tensor principal axes and the lab frame axes. This optimization strategy is easily combined with other optimization objectives (e.g. maximal efficiency) subject to physical constraints such as requiring that the ZMP operates within the support base. To deal with Bellman\´s "curse of dimensionality" we suggest, in parallel, two computational simplifications that may make the optimization problem tractable and easily implemented on today\´s humanoid robots. Finally, we address the problem of support base planning during ground and aerial locomotory phases. We propose novel online strategies for robust coordination of interacting limbs compatible with the proposed optimization strategy.