ybjoe / pacman-bidirec

pacman(bi-direction)

CSE 571 Spring 2020 Final Project

Team name and members

Quad Js

• Jinyung Hong
• Eunsom Jeon
• Yongbaek Cho
• Jaehyuk Choi

Description of Topic

Bi-directional search

We apply the variants of MM including MMε and Factional MM (fMM(p)) to solve the POSITION SEARCH problems in Pacman domain.

Contributions of each team member

• Jinyung Hong : I developed MM and MMε. I participated in writing the introduction and technical approach in our report.
• Eunsom Jeon : I developed fMM and heuristics. I participated in writing the technical approach heuristics, the results and discussion, and conclusion in our report.
• Yongbaek Cho : I experimented with algorithms and analyzed it. I created tables and graphs participated in writing the Results and discussion in our report.
• Jaehyuk Choi : Analyzed conventional heuristics on MM's behavior and developed customized one for better performance of MM. Contributed to algorithm design, introduction, and technical approach so that how the combined MM impletments the variants of MM.

Dependencies

• Python 2

How to run

python pacman.py -l [Maze] -p SearchAgent -a fn=[search_algorithm],heuristic=[Heuristic],useEpsilon=[useEpsilon],useFractional=[useFractional],p=[pValue]

• Maze (Required) : {Tiny, Small, Medium, Big}

• search_algorithm (Required) : {bfs, dfs, astar, ucs, bis}

  • bfs = breadthFirstSearch
  • dfs = depthFirstSearch
  • astar = aStarSearch
  • ucs = uniformCostSearch
  • bis = bidirectionalSearch

• Heuristic (Optional) :

  • default: nullHeuristic,
  • {manhattanHeuristic, octileDistance, cosineDistance, digonalDistance, customizedHeuristic, mazeHeuristic}

• useEpsilon (Optional) :

  • default : False
  • {True, False}

• useFractional (Optional) :

  • default : False
  • {True, False}

• pValue¹ (Optional) :

  • default : 0.5
  • [0, 1]

¹ The meaning of pValue:

• pValue = 0.5 -> MM
• pValue = 0 -> reverse A*
• pValue = 1 -> Forward A*

Example)

• MM0 (MM Without heuristic)

python pacman.py -l tinyMaze -p SearchAgent -a fn=bis

• MM With manhattan heuristic

python pacman.py -l tinyMaze -p SearchAgent -a fn=bis,heuristic=manhattanHeuristic

• MMε without heuristic

python pacman.py -l smallMaze -p SearchAgent -a fn=bis,useEpsilon=True

• MMε with manhattan heuristic

python pacman.py -l smallMaze -p SearchAgent -a fn=bis,heuristic=manhattanHeuristic,useEpsilon=True

• Fractional MM without heuristic

python pacman.py -l smallMaze -p SearchAgent -a fn=bis,useFractional=True,p=0.8

• Fractional MM with manhattan heuristic

python pacman.py -l tinyMaze -p SearchAgent -a fn=bis,heuristic=manhattanHeuristic,useFractional=True,p=0.3

Comparison table of Heuristics

Graph

References

• AIMA examples
• Video AAAI Presentation - MM: Bidirectional Search That Is Guaranteed to Meet in the Middle
• Paper Bidirectional Search That Is Guaranteed to Meet in the Middle
• Paper Extended Abstract: An Improved Priority Function for Bidirectional Heuristic Search
• Paper A Brief History and Recent Achievements in Bidirectional Search

Pacman-bidirec