Multifingered robotic hands have net yet made significant inroads into prac
tical applications, partly due to the complexity of dextrous manipulation t
asks, and also due to control software shortcomings. High level task contro
llers exist, as do low level grasp controllers, but neither of these fully
address the problems of changing kinematic and dynamic constraints that ari
se during a grasping and manipulation task. This paper develops a framework
, utilizing phases, events, and transitions, that bridges the gap between h
igh- and low-level central. Results from constraint handling and transition
experiments conducted with a two-fingered hand are included.