Integracija vidnega zaznavanja in koordiancije gibov robota pri igranju igre Križci-krožci

BOC, ANA GAJA (author), Skočaj, Danijel (mentor), Podpečan, Vid (co-mentor)

Diplomska naloga obravnava problem programiranja robota NAO za igranje igre Križci-Krožci. Pri tej igri igralca izmenično označujeta prazna polja na mreži s tremi stolpci in tremi vrsticami. Zmaga tisti, ki mu uspe prvemu postaviti tri znake v vrsto. Če se zapolnijo vsa polja in to ne uspe nikomur, je igra neodločena. Pri razvoju take funkcionalnosti lahko dani problem predstavimo kot sklop treh podproblemov. Prvi je ta, da se mora robot zavedati položaja svoje roke v prostoru glede na ostale objekte, da bi lahko dosegel želeno polje na igralni površini. Problem je rešen z izračunom inverzne kinematike s pomočjo regresijske nevronske mreže. Drugi problem je prepoznavanje situacije na igralnem polju, torej kje se igralno polje nahaja in kakšno je trenutno stanje igre. Pri reševanju tega problema so uporabljeni osnovni principi vidnega zaznavanja. Tretji problem je sama strategija igranja, torej izbira poteze, ki bi robota lahko pripeljala do zmage oziroma to isto preprečila nasprotniku. Strategija je izdelana po principu preiskovalnega drevesa Alfa-beta. V tej diplomski nalogi smo vse tri probleme uspešno rešili in implementirali sistem, ki je v igranju igre Križci-Krožci nepremagljiv.

• Language: Slovenian
• Keywords: robotika, inverzna kinematika, vidno zaznavanje, teorija iger
• Work type: Undergraduate thesis
• Organization: FE - Faculty of Electrical Engineering
• Year: 2019
• PID: 20.500.12556/RUL-110525
• Publication date in RUL: 16.09.2019

Secondary language

• Language: English
• Title: Integration of robot visual perception and hand-eye coordination for playing the game of Tic-Tac-Toe

Abstract

This thesis aims to program the NAO robot so that it will be able to play the game of tic-tac-toe, where two players alternately mark empty fields on a game board with three columns and three rows. The goal is to place three crosses or circles in a row before the opponent does so. If there are no more empty fields and no one succeded in doing so, the game ends in a draw. When teaching the robot, this problem can be split into three sub-problems. Firstly, the robot must be aware of the position of its arm in space, so that it is able to reach his goal on the gaming surface. This problem is solved by calculating inverse kinematics using a regression neural network. The second problem is recognizing the current situation, which includes locating the gaming surface and recognizing the state of the game. The program for vision recognition implements basic principles of robot vision. Lastly, the robot plays the game using such a strategy for choosing its next move, that it is most likely to get him victory or to prevent its oponent from doing so. The algorithm used for choosing the optimal move to win the game is based on alpha-beta search trees. In this thesis we succesfully solved all three problems and implemented a system, that is unbeatable at Tic-Tac-Toe.

• Keywords: robotics, inverse kinematics, visual recognition, game theory