Finding the optimum path to an indoor object by locating it with an autonomous vehicle

This thesis aims to develop an integrated system enabling an autonomous mobile robot to detect a specific target object in an indoor environment, navigate it via an optimal path, and return safely to its starting point. In such environments, where GPS signals are unavailable and challenges like narrow spaces, static and dynamic obstacles, and variable conditions exist, reliable operation requires the combination of object detection, path planning, motion control, and obstacle avoidance algorithms. In the existing literature, these algorithms are often addressed independently, and comprehensive systems that integrate them in real-time under indoor conditions are limited. This study fills that gap by systematically integrating SLAM, AMCL, A*, DWB, HSV color detection, and depth estimation algorithms under the ROS2 Humble framework to create a unified and functional architecture. In the methods section, a simulation scenario is implemented using the TurtleBot3 Waffle robot in the Gazebo environment. The robot is tasked with detecting a red-colored object, estimating its approximate position, navigating toward it, and returning to its original position. Cartographer is used for mapping, AMCL for localization, A* and DWB for global and local path planning, and object detection is performed using HSV filtering and depth sensing. The process is managed via a state machine that guides each step, from detection to return. Simulation results demonstrate that the system successfully performs in environments with both static and dynamic obstacles. After parameter optimization, the success rate of reaching the target improved to 90%, while the average path length and completion time decreased. The originality of this study lies in the real-time integration of diverse algorithms in a modular ROS2-based system.