Uncertain-aware Informative Task Planning and Assignment for Multiple-UUVs Cooperative Underwater Exploration

Abstract

This paper presents an uncertainty-aware exploration framework for cooperative underwater operations using multiple unmanned underwater vehicles (UUVs). The proposed framework leverages prior environmental information to iteratively integrate task planning, task assignment, and prior belief updating, enabling efficient exploration in unknown underwater environments. An interest area selection strategy is proposed to balance the exploration of uncharted regions and the exploitation of areas with high target likelihood. To optimize interest area task allocation, a simultaneous auctionbased mechanism is developed that assigns each interest area to the most suitable UUV by maxing potential information gain while minimizing operational costs. Additionally, to address the computational constraints of UUV systems, a Sparse Gaussian Process (SGP) with variationally optimized inducing points is employed, enabling rapid and accurate fusion of real-time observations with prior environmental information. This approach facilitates dynamic updates of the probabilistic environment representation and interest point selection without compromising accuracy. Experimental results in the HoloOcean simulator demonstrate the framework’s effectiveness in refining the probabilistic environment representation, achieving efficient exploration and accurate target detection in complex underwater scenarios. The results highlight the framework’s capability to dynamically adapt to environmental uncertainties, showcasing its potential for underwater exploration applications.