This software implements linear, gradient-descent Sarsa(lambda) with tile coding, as described in "Reinforcement Learning: An Introduction". The implementation follows closely the boxed algorithm in Figure 8.8 on page 212. The example application is to the Mountain Car problem, as described on pages 214-215. This code is offered as an example that could guide other applications of reinforcement learning. The code is in C++ and Lisp. The Lisp versions include the possibility of graphical displays when run on a Macintosh.
by Richard S. Sutton, December 18, 2000