Research area
System Identification
Extracting hydrodynamic coefficients and manoeuvring models from model tests and full-scale field data.
A manoeuvring simulation is only as good as the coefficients inside it. We develop methods to identify hydrodynamic derivatives and manoeuvring models directly from experimental data — captive model tests, free-running model trials, and full-scale field measurements.
Learning models from free-running data
Support vector regression and related techniques are used to identify hydrodynamic coefficients from free-running datasets produced by high-fidelity simulators, with the goal of extending the same pipeline to experimental data. That would reduce reliance on large captive facilities such as towing tanks and rotating-arm basins for every new hull. Follow-on work focuses on parsimonious models via LASSO and SVR so that identified models stay interpretable and usable for control design.
Reduced-order models and waves
For path-following control we have derived simplified reduced-order yaw models from high-fidelity simulation data and used them to design PID, sliding-mode and backstepping controllers, comparing performance in calm water and under strong wind disturbance.
Another thread studies how manoeuvring dynamics change in waves. With regulations pushing ships toward lower installed power, it matters whether a vessel can still keep course in a seaway. Preliminary experiments in the wave basin at IIT Madras support this line of work.