Reinforcement Learning for Helicopter Flight
Pieter Abbeel has written an original paper about aerobatic helicopter flight using machine learning. He started by studying a pilot flying an helicopter to understand the dynamics model of the helicopter. He then used reinforcement learning to find the optimized controller for the flight.
Abbeel came to EPFL (Lausanne) a few months ago and gave a nice presentation with videos as well. You can find more information here:
http://heli.stanford.edu/
and some videos:
http://www.cs.stanford.edu/~pabbeel/heli-nips2006
This application has the advantage of a nice visual demonstration that may interest people outside the field of data mining.
