Session: 06-17-03 AI Technology for Ocean Engineering III

Paper Number: 128231

128231 - Data Driven Identification and Model Predictive Control for a Holonomic Catamaran Surface Vessel 

Autonomous Surface Vehicles (ASVs) have received increasing attention in various maritime applications, and their ability to operate autonomously in complex and confined environments, such as narrow inland waterways, presents a unique set of opportunities and challenges. This paper presents a comprehensive study on the dynamics identification and subsequent development of a Model Predictive Controller (MPC) for a holonomic catamaran surface vessel. The catamaran surface vessel under investigation was designed and developed in-house with a holonomic thruster configuration and a comprehensive sensor suite.

Traditionally, dynamics models are identified using various experiments in specialised facilities which are costly to build and maintain. With recent advances and new techniques developed in data driven identification, it is possible to identify the maneuvering models of surface vessels purely from free running maneuvers. This study focuses on experimental identification of a 3 degree of freedom maneuvering model of a fully actuated surface vessel with holonomic thrust configuration and a comprehensive sensor suite. Experimental data from these sensors is utilised to estimate surge, sway and yaw velocities. These velocity estimates along with the thrust commands are used to identifying hydrodynamic coefficients of the dynamics model which is formulated as a regression problem. This study focuses on identification of a parsimonious model of the system dynamics using the SINDy methodology.

With a well-defined dynamic model in place, we proceed to design a Model Predictive Controller for efficient trajectory tracking. The MPC framework is chosen due to its ability to optimize control inputs over a finite prediction horizon, ensuring the catamaran follows desired trajectories while adhering to operational constraints and actuator constraints. This paper outlines a holistic study on the dynamics identification and Model Predictive Control development to impact a wide range of applications, from oceanographic research to coastal security, by providing a robust control framework that can navigate the challenges of dynamic marine environments.

Presenting Author: Vallabh Deogaonkar Indian Institute of Technology Madras

Presenting Author Biography: I am a doctoral student in Department of Ocean Engineering at Indian Institute of Technology Madras.
I have completed my bachelors in Naval Architecture and Ocean Engineering with an Integrated Masters in Robotics from IIT Madras.
Prior experience and projects : [https://www.youtube.com/watch?v=Bcgicbp_92E]
My research interests include
> System Identification of nonlinear systems
> Autonomous marine vehicles control and path planning
> Physics Informed machine learning for engineering applications
> Reinforcement Learning
I am always eager to collaborate on these topics and learn new ideas.

Authors:

Vallabh Deogaonkar Indian Institute of Technology Madras
Mohammed Ibrahim M Indian Institute of Technology Madras
Akash Vijayakumar Indian Institute of Technology Madras
Abhilash Somayajula Indian Institute of Technology Madras