readme

CMAC神经网络机械臂控制的设计matlab源码

FUNCTION:
This program is a simulation of a car learning the proper acceleration to
get up a mountain.  It uses Q-learning with CMAC as a function
approximator.  It is based on papers by Rich Sutton and Satinder Singh and
was developed by Sridhar Mahadevan.

REQUIREMENTS:  This program requires a C++ compiler and the X11 windowing
libraries and include files.


INSTALLATION:
To install:

1. Edit the file Makefile, and change the line that begins:
	   INCLUDE =
       to the X11 include file path on your system.

2. Set the loader library path environment variable to the X11 library
   path on your system:
	setenv LD_LIBRARY_PATH <X11-lib-path> 

3. Type make.  The executable will be called "xcar"

4. To run the program with the pre-defined defaults, type "xcar".  
   To change the defaults, see the next section, INPUT.


INPUT:

	By default, one run of 1,000 trials is done.  The simulation
parameters (which are particular to this car implementation) are in the
mc.h file.  The reinforcement learning parameters are in qlearn-cmac.cpp. 
Of course, you must execute the "make" command again after you change any
parameter.


OUTPUT:
   A new window with a visual display of the current action of the
simulation will be displayed, while the window from which it was launched
will show (text) the position and velocity interval values.

   The best ongoing qvalues are written to the data files "qvalue-x-y",
where x is the run number, and y is the trial number.  A new data file is
started for every 100 trials. 	


BUGS AND LIMITATIONS: