Electro-SLAM: Distributed underwater multi-robot SLAM via bio-inspired active and passive electro-sensing

Abstract

This paper presents Electro-SLAM, a novel and fully distributed SLAM system for underwater single- and multi-robot applications, inspired by the electroreception capabilities of weakly electric fish. Departing from conventional vision- or sonar-based methods, Electro-SLAM employs a compact, custom-designed sensing module that supports both active and passive electro-sensing, enabling perception of geometric boundaries and material properties even in visually degraded, dark, or turbid environments, without relying on external infrastructure. At the foundational level, we develop a complete electro-sensing solution, including a compact hardware unit, theoretical electric field models under representative underwater boundary conditions, and estimation methods for localization. Building on this, for single-robot scenarios, we propose an active electro-sensing SLAM pipeline that integrates boundary detection, boundary tracking, and hierarchical localization strategies leveraging both geometric and material-aware features of boundaries. Finally, the framework is extended to multi-robot scenarios, forming the full Electro-SLAM system that integrates passive electro-sensing-based inter-robot relative pose estimation, decentralized semantic map merging, and collaborative localization via merged semantic maps, enabling asynchronous cooperation among multiple low-cost robots. The proposed Electro-SLAM is validated through real-world underwater experiments using custom-built small robots. Results demonstrate that Electro-SLAM significantly outperforms odometry-only baselines in localization accuracy, mapping quality, and semantic map completeness. By combining bio-inspired sensing with distributed multi-robot autonomy, Electro-SLAM offers a scalable SLAM paradigm for dark, turbid, and communication-constrained underwater environments, opening new directions for embodied intelligence in aquatic domains.