Feasible pattern generation method for humanoid robots

This paper proposes a feasible pattern generation method for humanoid robots. One of the difficulties in pattern generation for humanoid robots is that generated patterns must satisfy many constraints such as physical limits, self-collision and so on to be feasible in addition to constraints to achieve a specified task. In reality, some of these constraints are not often taken into account during the pattern generation and they are just checked afterwards and unsatisfied constraints are fixed by hand. It is not easy to find a parameter set to get a feasible motion for humanoid robot and these pattern generators need to be used carefully when they are used online. The proposed method integrates the feasibility constraints into the pattern generation algorithm and enables to use it online more safely and releases human from parameter tuning. Moreover, a stiffness varying constraint is introduced to improve the feasibility.