The paper deals with the problem of planning movements of a two-hand system, considering the possibility of using one hand to remove potential obstacles in order to grasp a desired object with the other hand. The approach is based on a Probabilistic Road Map that does not rule out samples implying collisions with removable objects but instead classify them according to the collided obstacle(s), and allows the search of free paths with the indication of which objects must be removed from the workspace to make the path be actually valid. The approach has been implemented and different tests were performed with considering a real two-hand robotic system with one hand in charge of grasping a desired object and the other in charge of removing the potential obstacles. Some running examples both in simulation and a real workcell are presented in the paper using simulations and real experimentations.