Efficient Multi-robot Active SLAM
Muhammad Farhan Ahmed, Matteo Maragliano, Vincent Frémont, Carmine Tommaso Recchiuto
TL;DR
This work addresses efficient exploration in unknown environments with multiple robots by introducing a frontier-sharing, ROS-based active SLAM framework. The method combines a D-Optimality–driven pose-graph uncertainty term with frontier-path entropy in a unified utility to select exploration targets, while filtering frontiers to reduce computation. A centralized server coordinates frontier merging, filtering, and reward updates to promote distributed coverage, achieving higher mapping quality and exploration efficiency than state-of-the-art baselines. Experimental validation in both simulated and real environments demonstrates substantial improvements in map coverage and SLAM accuracy, along with a marked reduction in frontier processing overhead, highlighting its practical impact for scalable multi-robot exploration.
Abstract
Autonomous exploration in unknown environments remains a fundamental challenge in robotics, particularly for applications such as search and rescue, industrial inspection, and planetary exploration. Multi-robot active SLAM presents a promising solution by enabling collaborative mapping and exploration while actively reducing uncertainty. However, existing approaches often suffer from high computational costs and inefficient frontier management, making them computationally expensive for real-time applications. In this paper, we introduce an efficient multi-robot active SLAM framework that incorporates a frontier-sharing strategy to enhance robot distribution in unexplored environments. Our approach integrates a utility function that considers both pose graph uncertainty and path entropy, achieving an optimal balance between exploration coverage and computational efficiency. By filtering and prioritizing goal frontiers, our method significantly reduces computational overhead while preserving high mapping accuracy. The proposed framework has been implemented in ROS and validated through simulations and real-world experiments. Results demonstrate superior exploration performance and mapping quality compared to state-of-the-art approaches.