Modelado e identificación de vehículos móviles usando modelos de baja complejidad basados en datos

MARIANO DE PAULA; GERARDO G. ACOSTA; ALEJANDRO ROZENFELD; IGNACIO CARLUCHO

Buenos Aires

Congreso; ARGENCON 2016; 2016

Autonomous vehicles are attractive platforms forseveral applications such as aerial, terrestrial, aquatic andunderwater applications. The system modeling andidentification is paramount to the success of the model-basedcontrollers. Reliable control strategies require faithful modelsto achieve a good performance. Classical modeling representsthe system dynamics by ordinary differential equations andoften requires extensive human knowledge. Many times, thedynamics are complex and nonlinear and also manysimplification assumptions are made during system modeling.In this paper we compare different data-driven techniques tomodel the system dynamics. Particularly, we use the wellknownartificial neural networks, multilayer perceptron andradial basis functions, as well as Gaussian process regression tomodel the vehicles dynamics. These techniques learn theunderlying structure of the vehicles dynamics from theexperimentally measured data offering a natural framework toincorporate the unknown nonlinearities. In this paper aterrestrial vehicle is identified, the Pioneer 3at and theobtained model is validated with the real vehicle.