A robust backfitting algorithm for additive models based on (robust) local polynomial kernel smoothers. It includes both bounded and re-descending (kernel) M-estimators, and it computes predictions for points outside the training set if desired. See Boente, Martinez and Salibian-Barrera (2017) <doi:10.1080/10485252.2017.1369077> and Martinez and Salibian-Barrera (2021) <doi:10.21105/joss.02992> for details.
