Computer Vision for Underwater Remote Teleoperation

Abstract

Assistive Artificial Intelligence (AI) systems can be useful for autonomous real-world underwater navigation, exploration, and manipulation. However, the human-in-the-loop approach, which relies on human expertise to make decisions even in challenging situations, remains vital. As a result, remote teleoperation is one of the most important robotics tasks required in harsh environments, such as retrieval and manipulation. Communication is a crucial aspect of teleoperation, however, because of the restrictions of underwater channels, Remotely Operated Vehicles (ROVs) must be connected to units on the surface through tethers or cables. An acoustic channel would enable ROVs to be used without the need for a physical connection to other units, enhancing their performance in many applications, such as examining complicated sea bottoms or wrecks. However, the introduction of acoustics induces severe delays in the system, which degrades the operator’s performance. In this project, we propose a delay-mitigation scheme for end-to-end teleoperation, in which the user controls the robot in a virtually simulated underwater environment, and in conjunction, the robot replicates the commands in the real world after receiving them through the acoustic channel.

Date of Award	Apr 2023
Original language	American English
Supervisor	JORGE Dias (Supervisor)