Maze Solving Robot

Authors:

Jovana Trkulja, Anastasija Rajković

Mentors:

Bojana Malešević, student, School of Electrical Engineering, University of Belgrade

Vladan Bašić, student, School of Electrical Engineering, University of Belgrade

Dimitrije Pešić, School of Electrical Engineering, University of Belgrade

Đorđe Marjanović, graduate engineer in electrical and computer engineering, PFE

Abstract

This work presents a partial implementation of a maze-solving robot simulation using SLAM algorithms. A simulated LIDAR sensor enables obstacle detection and generation of a point-cloud map, while Split and Merge and Seeded Region Growing algorithms provide line segmentation of maze walls. The robot is initially positioned via mouse click and navigates using a frontier-based exploration strategy. Due to time constraints and non-modular code, the project remains partial, with limited testing and no quantitative evaluation. Nevertheless, the results demonstrate the basic functionality of the system and provide a solid foundation for further development, including map optimization, modular code design, and implementation of performance metrics.

Full paper