2P1-D05 Path generation for environment survey using UGV : Feasibility study of prediction model for discrimination of survey spots

Abstract

In the environment survey using autonomous vehicle, vehicle path should be generated from measurement results after the rough search of large space. To reflect the results we propose a method to compare them with some prediction to compare something with results. The prediction model of the field needs to deal with many variables with complex relationship. Applying 2 prediction methods, a historical database model and a decision tree model, we discuss which prediction model is suited for our research. In addition we consider the form of the data used for calculation of prediction value. Finally, database model is suited more than the decision tree model and the form of the data has less squared error than normal method.