Predictive digital twins for autonomous surface vessels

Abstract

This paper presents an approach to improve the reliability and efficiency of autonomous surface vessels (ASV) by leveraging predictive digital twin and state estimation techniques. The predictive digital twin is based on a graphical model that shows the relationship between variables, and it is implemented and tested in an ASV based on a nonlinear discrete-time model and an adaptive exogenous Kalman filter to estimate the states and parameters associated with faults in the system. The state estimation method consists of a cascade of a nonlinear adaptive observer and a linearized adaptive Kalman filter, and the predictive algorithm is developed by creating a discrete-time state-space model from measured data and using the validated model to forecast the system response for a future time span. Experimental results in the Otter ASV developed by Maritime Robotics demonstrate the potential of the predictive digital twin as a decision support system for the predictive maintenance of ASVs.