Development of Optimal Parameter Estimation Methodologies Applied to a 3DOF Autonomous Surface Vessel

Abstract

Autonomous Surface Vessels (ASVs) are reliable and robust vehicles. They perform autonomous missions in lakes, rivers and even open waters. Those are dangerous environments that requires precise and secure navigation. Under these conditions, the knowledge of a robust and accurate mathematical model is a fundamental aspect for adjusting the control system for reaching safety and performance. Moreover, traditional mathematical models disregard assimetries and coupling between the degrees of freedom. While those models work fine for bigger vessels, to ignore these characteristics in small ASVs compromises the model’s quality. In this context, this work presents a new methodology for modeling and identifying the dynamics of ASVs along with the uncertainties arising from disturbances and non-modeled dynamics. As for the uncertainties and disturbances, this work considers the coupling parameters into the mathematical modeling, which synthesizes the divergences between the model and the real application, allowing to incorporate asymmetries and model deficiencies. Regarding the parameter identification, the proposal is based on (i) the design of optimal input excitation signals from a double layer optimization methodology and (ii) a parametric estimation concept in two steps, dividing the original set of parameters into two partially coupled sub-problems. Finally, this work also presents a full discussion and analysis about the importance to manage the trade-off between precision and complexity of mathematical models, its respective solution spaces and the impact over the optimization algorithms. To validate the approach, a real 3 Degree of Freedom ASV with aerial holonomic propulsion system is used. The results show that it is possible to successfully capture the complex set of parameters and identify physical characteristics not considered by the model.

Keywords

• Autonomous surface vehicles
• real case parameter identification
• 3 DOF ASV
• optimal input signal design
• solution space complexity management
• parametric estimation