Intelligent Automation, Inc. — Department of Defense SBIR Phase I: N172-123

Description

Navys ship-to-shore cargo transfer operations, supported by the INLS, is largely dependent on local wave characteristics for decision making and throughput planning. In this proposal, Intelligent Automation Inc. (IAI), along with its subcontractor Lockheed Martin (LM), propose to develop a Wave Characteristics Measurement System (WCMS) based on state-of-the-art deep learning algorithm to estimate the wave characteristics in real time only using warping tug motions as input. The proposed deep learning architecture combines a Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks together. This combined structure makes it possible to learn the composition of physical system and noise model directly from noisy sensor data. Deep neural networks need large amount of data to train. In this project, we propose to develop high-fidelity ship motion simulation software to generate large amounts of data for training and evaluation. The proposed approach is built on top of IAIs previous success and experience with deep learning predictive analytics and ship motion simulation, prediction and control demonstrated in various ONR, DARPA and AFRL programs.