Autonomous Exploration of Complex Underwater Environments Using a Probabilistic Next-Best-View Planner

Autonomous underwater vehicles (AUVs) have been extensively used for open-sea exploration. However, the mapping or inspection of complex underwater structures, which have an interest from the scientific and the industrial point of view, is still carried out by professional divers or remotely operated vehicles. We propose a probabilistic next-best-view planner, targeted to hover-capable AUVs, that will allow them to explore these complex environments without an apriori model. The proposed method is based on scanning the area from different viewpoints in an iterative way. At each step, a viewpoint is chosen from a set of random samples according to a utility function. An obstacle-free path to the selected viewpoint is planned, and the vehicle navigates to it to gather a new scan that will be registered with the previous ones. To evaluate the proposed method, we present four different tests using the Girona 500 AUV, both in simulation and in real scenarios. The results demonstrate the capability to explore complex environments autonomously, producing models of the environment with a high degree of coverage that can enable mapping and inspection applications.