How can a machine learn to perform a task by observing the way a human being does it? Surprisingly, little has been done to address that question, but the authors of this chapter have taken some significant preliminary steps toward answering it, using experiments involving flying an aircraft simulator.
Data collected from simulated flights by an experienced pilot were filtered and statistically analyzed, and the computer learned to emulate the humans actions and reactions. A control system based on these experiments would be helpful when using an automatic pilot during landings and takeoffs in non-standard conditions, or with noisy controls. The results of the experiment are impressive and interesting, revealing, for example, that humans are not particularly good pilots.
The authors approach to the problem can easily be generalized, and has the potential to be extended to many other areas. The writing is clear, and the authors are refreshingly helpful and candid about the limitations of their experiments. This chapter will be of some interest to the general reader, and for those working in the aircraft-piloting field, it will be of great importance.