An efficient loop closure detection method for communication-constrained bathymetric cooperative SLAM

Cooperative bathymetry simultaneous localization and mapping (SLAM) technologies hold high potential for autonomous underwater vehicles (AUVs) to accurately map underwater terrains in large-scale environments, by sharing bathymetric measurements among multi-vehicles. However, only compressed data can be transmitted with low-bandwidth and unreliable underwater acoustic channels, and compressed or even polluted bathymetric data is at high risk of invalidating loop closures and ultimately leading to catastrophic SLAM failure. To deal with this problem, an efficient 6-degree-of-freedom loop closure detection (6-DoF ELCD) algorithm is proposed. The algorithm is based on a genetic algorithm (GA), wherein a robust fitness function is devised to flatten potential additional matching errors arising from compressed or contaminated maps and adaptive genetic parameters are employed to prevent premature convergence. Notably, the significant effect of unknown tidal on the fitness value was considered, hence the population was redesigned to decrease the search dimension. Compared to state-of-the-art methods, the proposed algorithm shows smaller localization and registration errors as well as higher computational efficiency in field data playback simulations.