ai practicum proposal anand bheemarajaiah, m.eng. chet mancini, m.eng. felipe osterling, junior ta:...

8
AI PRACTICUM PROPOSAL Anand Bheemarajaiah, M.Eng. Chet Mancini, M.Eng. Felipe Osterling, Junior TA: Jason Yosinski

Upload: lambert-walker

Post on 31-Dec-2015

216 views

Category:

Documents


3 download

TRANSCRIPT

Page 1: AI PRACTICUM PROPOSAL Anand Bheemarajaiah, M.Eng. Chet Mancini, M.Eng. Felipe Osterling, Junior TA: Jason Yosinski

AI PRACTICUM PROPOSALAnand Bheemarajaiah, M.Eng.

Chet Mancini, M.Eng.

Felipe Osterling, Junior

TA: Jason Yosinski

Page 2: AI PRACTICUM PROPOSAL Anand Bheemarajaiah, M.Eng. Chet Mancini, M.Eng. Felipe Osterling, Junior TA: Jason Yosinski

Problem Statement

The motivation for this project is to practice implementing various AI ideas into how to play a game against another agent. Various and unique sub problems and factors in this task will be to:

• Find a feasible solution to the game, where the total waiting time between moves is less than 10 minutes.

• Interpret the opponent’s moves and respond by evaluating new strategies for each move.

• Try to block the opponent as much as possible.• Optimize both on strategy and on time. i.e. The best possible move

in the shortest possible time.

Page 3: AI PRACTICUM PROPOSAL Anand Bheemarajaiah, M.Eng. Chet Mancini, M.Eng. Felipe Osterling, Junior TA: Jason Yosinski

IO Specification• Input (stdin)

• Game initialization• Request move• Opponent move information• Error information• End Game notification

• Output (stdout)• Our intended move.

• Error (stderr)• Our custom debugging information

• (Summary of specification in Wiki).

Page 5: AI PRACTICUM PROPOSAL Anand Bheemarajaiah, M.Eng. Chet Mancini, M.Eng. Felipe Osterling, Junior TA: Jason Yosinski

System Architecture| State Representation

Credit: http://hem.passagen.se/baolan/release/china.pdf

Page 6: AI PRACTICUM PROPOSAL Anand Bheemarajaiah, M.Eng. Chet Mancini, M.Eng. Felipe Osterling, Junior TA: Jason Yosinski

System Architecture | Other• Other Components we plan to build and use

• IO adapter – This is the component which interacts with the server through stdin and stdout.

• Data access abstraction module – This module abstracts out the data structures which are used internally.

• Successor Function evaluation module – This module again will be generating the next best possible moves. Going forward we will be able to test various algorithms and their performance by replacing the contents of this module.

• Error and debug data enabled GUI – This module will be responsible for handling Logging, GUI for human testing.

• Testing Framework – We will have a framework using which we will be able to add new testcases very easily. We might build this using JUnit Framework or TestNG framework in Java.

Page 7: AI PRACTICUM PROPOSAL Anand Bheemarajaiah, M.Eng. Chet Mancini, M.Eng. Felipe Osterling, Junior TA: Jason Yosinski

Team Task Breakdown• With three team members we plan to break up the

responsibilities in the following manner• Members A and B will work on bootstrapping the system into

workable modules.• Member C will spend time researching and digesting the multitude

of information on available algorithms, and then begin to implement those algorithms in our agent.

• A or B will each implement logging and testing modules.• All three members will work on documentation.

Page 8: AI PRACTICUM PROPOSAL Anand Bheemarajaiah, M.Eng. Chet Mancini, M.Eng. Felipe Osterling, Junior TA: Jason Yosinski

Basic Schedule• Milestone 1

• Construct essential data structures• Implement basic IO• Extend GUI to have custom debugging information.• Extend random AI to a basic goal seeking agent.

• Milestone 2• Construct an essential AI• Develop additional heuristics and methodologies based on testing

and preliminary competition.

• Final• AI revised.• Project finished