A Topological Map based Autonomous Exploration Method for Mobile Robots

Abstract

This paper proposes a novel topological map-based autonomous exploration method for mobile robots to address the limitations of traditional grid-based and feature-based approaches in computational efficiency, noise sensitivity, and scalability. The core of our method leverages a Generalized Voronoi Diagram (GVD) with a sparse sampling strategy and incremental map denoising to rapidly extract environmental features while minimizing redundant nodes. A constrained Mean Shift clustering algorithm is introduced to generate topological nodes, ensuring connectivity and avoiding overlap with obstacles. The topological map is efficiently connected using a bidirectional A* algorithm, enabling robust navigation in complex environments. To guide exploration, a fused cost function integrates topological path length, node count, and angular deviation to prioritize frontier points. The experimental results in three simulated environments demonstrate significant improvements over RRT Exploration, achieving 14.8%–15.8% higher efficiency and 6%–12.9% shorter path lengths. The proposed method exhibits strong adaptability across varying environments, offering a scalable and computationally efficient solution for autonomous robot exploration.