Development of Autonomous NavigationAlgorithms for Service Robots

Abstract

The increasing deployment of service robots in healthcare, hospitality, logistics, and domestic environments requires robust autonomous navigation capabilities. Traditional navigation approaches often fail to deliver sufficient adaptability to dynamic and unstructured spaces, creating the need for new algorithmic frameworks capable of real-time decision-making. This study presents a comprehensive development and evaluation of a hybrid navigation model combining probabilistic mapping, deep-learning-  based perception, and behavior-based motion planning. The proposed system integrates an enhanced Simultaneous Localization and Mapping (SLAM) module with a convolutional neural network for obstacle detection and a hierarchical reinforcement learning structure for path optimization. Experiments were conducted using a TurtleBot3 platform, simulated environments in Gazebo, and real-world corridor and room layouts. Results demonstrate improved localization accuracy, reduced collision rates, and higher
trajectory smoothness compared to conventional and basic SLAM frameworks. Findings indicate that hybrid navigation systems offer significant reliability in environments characterized by moving humans, changing object positions, and variable lighting conditions. This research contributes to scalable algorithmic designs for next-generation service robots intended for human-centric tasks.