I have used the ROS system extensively, developing my rover for the NASA Lunabotics competition. One thing everyone should know about robots, they are freaking complicated.
They have programs on every level of abstraction from low-level embedded code to high-level executive task manager to create pre-defined behaviours.
ROS has gone through quite a few changes, and still has some huge architecture changes coming to fix some of the short comings of the ROS protocol especially for integration of embedded devices which robotics heavily rely on.
Unfortunately I'm not going to be doing much robotics anytime soon. NASA has cancelled the competition for any International teams(I'm Canadian and we came 8th last year). And I'm not to keen on developing systems for DARPA.
source: A friend works at the Open Source Robotics Foundation (osrfoundation.org)
I have used Willow as an example of the good that newly minted millionaires can do when explaining to my non-tech friends how this area keeps going. It's amazing to watch someone with a big chunk of change from their previous endeavour to turn around and plunge it into new cool things. Non-techies think that folks just buy a yacht and spend the rest of their days in the Bahamas (or Beleize :-)) but the reality is people are driven to create cool stuff.
What will take its place with respect to research into human-level, mobile, autonomous manipulation platforms?
[1] http://osrfoundation.org/ [2] http://osrfoundation.org/blog/ros-at-osrf.html