Probabilistic mapping in agricultural environments using kernel estimators with recursive subsampling

SANTIAGO TOSETTI; LUCIO SALINAS; RICARDO CARELLI; JAVIER GIMENEZ

Mar del Plata

Workshop; XVII Reunión de trabajo en Procesamiento de la Información y Control; 2017

This work presents a non parametric probabilisticmapping based on kernel estimators which does not use grids.Theproposed methodology characterizes the map with a cloud ofpoints obtained from several observations of the environment.In order to maintain a bounded number of observations inmemory, a recursive subsampling algorithm is proposed. Theprocedure is included in an SLAM, in order to localize therobot as well. An application example is the presented, where theproposed methodology is applied in an agricultural environment.Simulation and experimentation results are presented to validatethe proposal.