In this paper we present a method for an appearance-based modeling of the e
nvironment of a mobile robot. We describe the task (localization of the rob
ot) in a probabilistic framework. Linear image features are extracted using
a Principal Component Analysis. The appearance model is represented as a p
robability density function of the image feature vector given the location
of the robot. We estimate this density model from the data with a kernel es
timation method. We show how the parameters of the model influence the loca
lization performance. We also study how many features and which features ar
e needed for good localization. (C) 2001 Elsevier Science B.V. All rights r
eserved.