izpis_h1_title_alt

Drevesno preiskovanje Monte Carlo pri namizni igri Scotland Yard
ID Ilenič, Nejc (Author), ID Šter, Branko (Mentor) More about this mentor... This link opens in a new window

.pdfPDF - Presentation file, Download (327,25 KB)
MD5: 83AFB1572CAC4C69E70B5A045A55C6E2
PID: 20.500.12556/rul/8647a469-9592-44ae-a4d0-64e24c2dca74

Abstract
Drevesno preiskovanje Monte Carlo zaradi uspeha pri računalniški igri Go postaja vse bolj uveljavljena metoda odločanja v različnih domenah. Za zelo uspešno se je izkazala pri igrah s popolno informacijo za enega, dva ali več igralcev, pri igrah, kjer igralcem v danem trenutku ni na voljo vsa informacija, pa je za večjo učinkovitost potrebno uvesti domensko specifične izboljšave. V diplomskem delu so opisani in empirično preizkušeni obstoječi pristopi k problematiki uporabe drevesnega preiskovanja v namizni igri Scotland Yard. Izkazalo se je, da hevristična izbira možne lokacije igralca s popolno informacijo v največji meri vpliva na uspeh igralcev z nepopolno informacijo. Po poskusih se je MCTS igralec z vsemi izboljšavami izkazal kot konkurenčen nasprotnik človeškemu igralcu.

Language:Slovenian
Keywords:drevesno preiskovanje, Monte Carlo, nepopolna informacija, Scotland Yard, odločanje, zgornja meja zaupanja pri drevesih, umetna inteligenca
Work type:Bachelor thesis/paper
Organization:FRI - Faculty of Computer and Information Science
Year:2015
PID:20.500.12556/RUL-72440 This link opens in a new window
Publication date in RUL:17.09.2015
Views:1393
Downloads:1719
Metadata:XML RDF-CHPDL DC-XML DC-RDF
:
Copy citation
Share:Bookmark and Share

Secondary language

Language:English
Title:Monte Carlo tree search in the board game of Scotland Yard
Abstract:
Because of its success in the computer game of Go, Monte Carlo Tree Search is becoming a progressively popular decision making algorithm in various domains. It has proven its strengths in singleplayer and multiplayer games with perfect information, however domain specific improvements must be introduced in games with imperfect information. In thesis existing approaches to the problem of applying the tree search to the Scotland Yard board game are described and empiricaly tested. It has turned out that heuristic selection of the possible location of the hider has the most impact on seekers performance. After testing, the MCTS player with all improvements has proven itself as a competitive opponent against the human player.

Keywords:tree search, Monte Carlo, imperfect information, Scotland Yard, decision making, Upper Confidence Bound for trees, artificial intelligence

Similar documents

Similar works from RUL:
Similar works from other Slovenian collections:

Back