Virtual Tethered Autonomous Underwater Vehicle Using Multimodal Communications

The collection of seafloor data, especially high-resolution images and video, can help researchers better understand the sea environment. Considering mobility, autonomous underwater vehicles (AUVs) are widely used to carry various data collection tools for seafloor data collection. Data collected by the AUV is challenging to retrieve in real time due to limitations in underwater communications. We propose the use of multimodal communications incorporating both acoustic and optical to allow the AUV to be virtually tethered to the surface vehicle during underwater surveys. Based on the path of the surface vehicle, the AUV carries out underwater data collection and then real-time data upload. We trained the reinforcement learning policy to control the AUV tracking the surface vehicle during underwater surveys, maintaining virtual tethers through multimodal communications. The trained policy is tested in the simulated environment and will be deployed on actual machines in the future.