Methods for generating training or validation data and for training or validating an advanced driver assistance system (ADAS) or an autonomous driving system (ADS).

Abstract

The invention proposes a method for generating data for the training / validation of an advanced driver assistance or autonomous driving system and characterized by a number, K, of characterization parameters organized in a vector, comprising: a) generation of a number N > 1 of sampling point vectors, each representative of a point of a normalized hypercube of dimension K > 1; for each sampling point: b) determination of allowed intervals for each entry of a vector on the basis of a predetermined condition and a conditional probability of all previously determined entries; c) rescaling of the sampling point vector with respect to the characterization parameter vector to obtain characterization parameters;and d) generation of a set of N scenario maps as training / validation data based on the N characterization parameter vectors with dimension K. Figure for the abstract: Figure 1;

Images