차압 및 심층신경망 기반 반력 추정을 이용한 수중 유압 매니퓰레이터 햅틱 피드백 제어

Abstract

In this study, a buoyancy-aware DHP-DNN-based reaction force estimation method is utilized to provide real-time haptic feedback to the remote operator of an underwater hydraulic manipulator, compensating for the visual perception limitations in underwater environment. A method that combines the DHP(Differential Hydraulic Pressure)-based reaction force calculation, which demonstrates rapid sampling rates and operational responsiveness, with the DNN(Deep Neural Network)-based reaction force prediction, which provides accuracy and stability, is used to estimate the final reaction force. This estimated force is then applied to a haptic feedback framework to deliver real-time haptic feedback to the user. The haptic feedback framework was validated through underwater gypsum cushing experiments using an underwater hydraulic manipulator. The results showed more than twice the responsiveness compared to vision-based perception, and the remote operator was able to intuitively understand the work situation through haptic feedback, demonstrating its utility in underwater manipulation tasks.