Skip to content

In the Pick-up and Drop-off (PD) World, our goal is to design a route from the agent so that it could use the least steps to send all the blocks to drop-off cells. To solve reinforcement learning problems, we use a statistical approach and dynamic programming, especially Q-learning, to estimate the utility of taking actions in the states of the …

Notifications You must be signed in to change notification settings

Xiaoyang-Rebecca/Artificial-intelligent

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

6 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Artificial-intelligent

The projects for AI

Thanks for using reinforcement learning to solve PD world !

Contact XiaoyangLi (xiaoyang.rebecca.li@gmail.com)

%%%%%%%%%%%%%%%%%%%% Q_learning %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

Language : C++

Program compiler : GNU GCC (IDE-codeblocks) http://www.codeblocks.org/downloads

[ output.txt ]

Header line: The experiment name, the numbered execution (each experiment is executed twice), the seed used for this execution.

	#Subsequent lines are the q-tables:

	- One line indicating step number, Reward and Blocks Delivered after every 40 steps

	- The q-table using state representation 2: (i, j, x), ordered as: N S E W after first 100 steps, after 1st dropofff is filled and after each termination.

[Q_learning results]

    Exactly the same content as Output.txt except we put all the result in separate files and 6 folders

%%%%%%%%%%%%%%%%%%%% Visualization %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

Language : Matlab

[QtableFullstate . m] is the main function to generate screen shot of Q table and Fullstate

 Input template: (We need to copy the Q table from txt file to the corresponding templates)

	Qread0.dat for Q table x=0

	Qread1.dat for Q table x=1

 Output:

     original    fig images in [Visualization \ QtableFullstate Fig] folder

     compacted   jpg images in [QtableFullstate JPEG] folder 

[PerformanceMeasure.m] is the main function to generate performance measurement

Input template: (We need to copy the Q table from txt file to the corresponding templates)

	Perf.xlsx (variablenames = { 'steps','Reward','BlocksDelivered','BankAccount','Note'}

Output :

     original   fig images in [Visualization \ PerformanceMeasure Fig] folder

     compacted   images in appendix chaps of report.

About

In the Pick-up and Drop-off (PD) World, our goal is to design a route from the agent so that it could use the least steps to send all the blocks to drop-off cells. To solve reinforcement learning problems, we use a statistical approach and dynamic programming, especially Q-learning, to estimate the utility of taking actions in the states of the …

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published