Session: 07-15-01: Dynamics and Control of Soft Structures

Paper Number: 147065

147065 - Simulation Model Development and Calibration of Surface Watercraft Dynamics Using Neural Networks and Machine Learning 

The scale model of a surface watercraft with electric propulsion by a dc motor, propeller and rudder for steering was built. A simulation model that can reproduce motion patterns of the boat with sufficient fidelity in various sets of conditions was developed by compounding data sets acquired through various trials and controlled experiments with established governing principles of ship theory. The aim was to reduce the number and cost of instrumentation required to calibrate the model, without unacceptably compromising precision or fidelity. In effect, by using knowledge of boat dynamics, a custom grey-box model was built and calibrated. A pair of nonlinear state equations, capturing the dynamic response of the watercraft of interest were formulated derived. The dynamical model can form the foundation on which AI-inspired control approaches for navigation of a single or swarm of boats can be built. In the AI-inspired control synthesis approach, the control problem is tackled by concept of virtual physics applying on the boat and resulting in motion schemes that are beneficial in particular applications, e.g. homing or visiting waypoints. The design such controllers involves cancellation of actual dynamics governing the boat’s motion complies and then implementation of laws of virtual (i.e. artificial) physics. In the present work, as an initial step in this direction, a series of open-loop tests enable us to identify and formulate the dynamics governing the watercraft’s motion.

Presenting Author: Miguel A Trejos University of New Orleans

Presenting Author Biography: Miguel A. Trejos is a holder of a BSc. in Electrical Engineering. He is currently pursuing a MSc in Engineering at the University of New Orleans.

Authors:

Jonathan Parker Freels University of New Orleans
Nikolaos Xiros University of New Orleans
Miguel A Trejos University of New Orleans