A Novel Cooperative Localization Algorithm based on Heterogeneous Data Fusion using GNN

The ocean is a natural resource of immense exploratory value, and acquiring precise positional information is critical for effective task execution of Autonomous Underwater Vehicles (AUVs). To address this challenge, this paper proposes a novel cooperative localization algorithm based on heterogeneous data fusion in a collaborative underwater network. The proposed algorithm adopts a lightweight architecture integrated with a graph neural network (GNN) as its core model. The GNN helps to process raw sensor measured data and reduce the dependence on initial location values of the AUV, and hence to enable high-accuracy underwater localization with low computational complexity. Experimental results show that, compared to traditional Extended Kalman Filter (EKF) and Factor Graph (FG) localization methods, the proposed algorithm achieves superior positioning accuracy and robustness.