A Lightweight, Intelligent Autonomous Exploration System for Indoor Robots Using Improved Random Walk With Back-Off Strategies

Autonomous exploration in unstructured indoor settings requires effective navigation approaches to manage obstacles, dead ends, and resource constraints. Conventional algorithms, like the TurtleBot3 Drive (TB3), frequently struggle with inadequate area coverage, inefficient obstacle management, and substantial computational demands. This study presents the Intelligent Exploration (IE) algorithm, an innovative method that integrates an improved random walk technique, real-time LiDAR-based obstacle avoidance, and a strategic back-off mechanism to optimize exploration efficiency. Detailed simulations using the TurtleBot3 Burger in the Gazebo simulator under the ROS Noetic framework, performed from various starting positions and over different time intervals (ranging from 2 to 15 minutes), demonstrated the efficiency of the proposed method when integrated with SLAM Gmapping. The IE algorithm consistently surpassed the baseline TurtleBot3 Drive (TB3) algorithm across critical metrics, attaining a 35% increment in coverage area (16.25 m2 vs 11.99 m2), traversing 35% longer distances (113.95 m vs 84.07 m), and identifying 17% more obstacles (137.82 vs 118.11). The algorithm maintained a reliable average localization error of 0.005 m while exhibiting memory efficiency by the reduced memory usage by up to 13.7% relative to TB3. By addressing limitations in coverage, obstacle handling, and resource efficiency, the IE algorithm provides a scalable and lightweight solution for autonomous indoor exploration, particularly on resource-constrained robotic systems.