Collaborative Radio SLAM for Multiple Robots based on WiFi Fingerprint Similarity
This addresses the problem of efficient and accurate navigation for multiple robots in large environments, offering a domain-specific incremental improvement over existing SLAM methods.
The paper tackles the inefficiency of single-robot SLAM and high computational costs of visual/LiDAR methods by proposing a collaborative SLAM approach using WiFi radio signals, introducing a novel similarity model combining RSS and AP detection likelihood to improve localization accuracy, with extensive experiments demonstrating effectiveness.
Simultaneous Localization and Mapping (SLAM) enables autonomous robots to navigate and execute their tasks through unknown environments. However, performing SLAM in large environments with a single robot is not efficient, and visual or LiDAR-based SLAM requires feature extraction and matching algorithms, which are computationally expensive. In this paper, we present a collaborative SLAM approach with multiple robots using the pervasive WiFi radio signals. A centralized solution is proposed to optimize the trajectory based on the odometry and radio fingerprints collected from multiple robots. To improve the localization accuracy, a novel similarity model is introduced that combines received signal strength (RSS) and detection likelihood of an access point (AP). We perform extensive experiments to demonstrate the effectiveness of the proposed similarity model and collaborative SLAM framework.