Haptic Shared Control in Deep Sea Mining

Abstract

Deep sea mining is currently being investigated as a possibility to harvest valuable materials from mineral-rich areas located in water depths up to 2000 meters. One promising mining method is to employ a large crawler on the seabed, remotely controlled by an operator on the supporting vessel. Controlling such a vehicle is expected to be difficult due to unpredictable seabed conditions and limited situation awareness of the operator. In addition, the optimal human-machine interface for controlling the crawler is yet to be determined. A common approach in marine operation is to automate the task as much as possible, leaving the operator in a supervisory role. An alternative approach is haptic shared control, which has shown to be beneficial in vehicle control tasks (automotive, UAVs), yielding improved performance but mitigating traditional human-automation interaction issues such as skill degradation, reduced situation awareness and overreliance. This study aims to compare supervisory control and haptic shared control to manual control of a teleoperated subsea crawler. A simulator was constructed, including a bi-manual control interface capable of rendering haptic feedback, two virtual displays showing primary and secondary task-related information, a mathematical model simulating the dynamics of the slow vehicle, and unpredictable soil properties of the seabed. In a human factor experiment, subjects (n=12) controlled the simulated crawler to complete normal steering, repeated obstacle avoidance, and an unexpected slip event at the end; all with manual control, haptic shared control and supervisory control. During normal steering between obstacles, both haptic shared control and supervisory control improved subjects’ performance and supervisory control allowed a significant decrease in control effort. However, during slip recovery and obstacle avoidance, supervisory control appreciably reduced subjects’ situation awareness. Therefore haptic shared control is a promising approach to assist the operator in underwater teleoperation with improved task performance but not the side-effects from the automation