There is a lot more to manipulation than grasping! The term nonprehensile manipulation means "manipulation without grasping", and humans do it often. Pushing an object that is too big to grasp (e.g. an office chair) is a nice example, but we will see more subtle examples in this chapter, too.
For force-closure-grasping manipulation, it is possible to think of a grasped object as effectively an extension of the distal links of the robot; the geometry and inertia of the manipuland effect the kinematics and dynamics of the robot, but as a control problem the manipuland does not contribute additional degrees of freedom to the problem. Nonprehensile manipulation requires considering the additional (unactuated) degrees of freedom of the object. This makes the entire control problem (controlling the degrees of freedom of the robot + the objects) an underactuated control problem. This chapter will focus on this very interesting class of control problems in the context of manipulation.
Note that manipulation does not need to be nonprehensile to be underactuated. As a nice example, Chavan-Dafle et al. considered the case were a two fingered gripper used external environmental contacts in order to perform in-hand re-orientation (see figure below), and referred to the task as "prehensile planar pushing"