Added methods for parsing command line arguments so that testing can be done without recompilation
Valid command options are
vi=true (run value iteration)
pi=true (run policy iteration)
ql=true (run QLearning)
converge=true (run until consecutive policies are equal to one another)
cc=5 (integer representing the number of policies need to be equal to consider convergence)
Created a method for checking if two policies are equal to one another to test convergence. Policies are equal to one another if for all valid states the same action is selected.
Notes:
For value iteration convergence seems to be working properly. For policy iteration the policies converge quickly but are not optimal. Haven't wrapped my head around this yet.
Added methods for parsing command line arguments so that testing can be done without recompilation Valid command options are vi=true (run value iteration) pi=true (run policy iteration) ql=true (run QLearning) converge=true (run until consecutive policies are equal to one another) cc=5 (integer representing the number of policies need to be equal to consider convergence)
Created a method for checking if two policies are equal to one another to test convergence. Policies are equal to one another if for all valid states the same action is selected.
Notes: For value iteration convergence seems to be working properly. For policy iteration the policies converge quickly but are not optimal. Haven't wrapped my head around this yet.