8
Artificial WINtelligence By: Jamaal Alleyne, C. Barrett Ames, Daniel Sullivan

Artificial WINtelligence By: Jamaal Alleyne, C. Barrett Ames, Daniel Sullivan

Embed Size (px)

DESCRIPTION

Input/Output Specification 0 = Game Initialization (followed by 3+Nx*Ny Integers) – Total Time available – Board Size – List of cell Status Numbers 1 = Request Move (followed by 1 integer) – Remaining time 2 = Opponent Move – Start Cell – End Cell 3 = Error 4 = End Game Input (Managed Via stdin)Output (Managed Via stdout) 0 = Move (followed by 3+2NIntegers) – Number of Hops – Start Cell Coordinates – If N>1, list coordinates of N-1 intermediate cells – End Cell Coordinate

Citation preview

Page 1: Artificial WINtelligence By: Jamaal Alleyne, C. Barrett Ames, Daniel Sullivan

Artificial WINtelligence

By: Jamaal Alleyne, C. Barrett Ames, Daniel Sullivan

Page 2: Artificial WINtelligence By: Jamaal Alleyne, C. Barrett Ames, Daniel Sullivan

Problem Statement & Motivation

•Problem Statement•Project Scope•Project Objective•Deliverables•Behaviors•Technical Requirements•Limitations & Constraints

Page 3: Artificial WINtelligence By: Jamaal Alleyne, C. Barrett Ames, Daniel Sullivan

Input/Output Specification

• 0 = Game Initialization (followed by 3+Nx*Ny Integers)– Total Time available– Board Size– List of cell Status Numbers

• 1 = Request Move (followed by 1 integer)– Remaining time

• 2 = Opponent Move– Start Cell– End Cell

• 3 = Error• 4 = End Game

Input (Managed Via stdin) Output (Managed Via stdout)

• 0 = Move (followed by 3+2NIntegers)– Number of Hops– Start Cell Coordinates– If N>1, list coordinates of N-1 intermediate cells– End Cell Coordinate

Page 4: Artificial WINtelligence By: Jamaal Alleyne, C. Barrett Ames, Daniel Sullivan

Background Reading

Artificial Intelligence: A Modern ApproachStuart Russel Peter Norvig

Page 5: Artificial WINtelligence By: Jamaal Alleyne, C. Barrett Ames, Daniel Sullivan

General Approach

•Where’s The AI?•Adversarial Search•Minimax algorithm•Alpha-Beta pruning•Real-Time Imperfect Decisions

•Reinforcement Learning •Passive •Learn utility functions for minimax

Page 6: Artificial WINtelligence By: Jamaal Alleyne, C. Barrett Ames, Daniel Sullivan

System Architecture

• Input System – Data Structure, API

• Learning System– Dependent

• Goal System– Build correct plans, dependent

• Output System – Independent, API

Page 7: Artificial WINtelligence By: Jamaal Alleyne, C. Barrett Ames, Daniel Sullivan

Evaluation Plans• “Toy” Task: successfully move pieces to the other side of the board• “Hard” Task: defeat a human player

Possible evaluation metrics include the degree of grouping, efficiency measured as the number of jumps, and the number of moves needed to reach the goal state

Page 8: Artificial WINtelligence By: Jamaal Alleyne, C. Barrett Ames, Daniel Sullivan

Schedule• Agile Programming

– Individual sections due by Sunday September 11th 1:00pm – Presentation and revisions done by Monday September 12th

10:00pm – Preliminary proposal due September 13th 2011revisions done by

Thursday September 15th at 10:00pm – Final proposal due September 16th 2011– Agree upon API for I/O September 20th – Create data structure using built-ins for java September 23rd 10:00pm– Create output action functions with unit tests - Sept 23rd 10:00pm – Create Opening moves system with unit tests - Sept 23rd 10:00pm– Internal Code review / Performance Tests