# Improvements over FastSLAM2.0

Are there some papers and/or works that describe improvements for the FastSLAM2.0 algorithm? I'm particularly interested in the following areas:

• The map is completely unknown at the beginning of the exploration
• The map is a fixed track (e.g a racing circuit)
• The landmark observations are much noisier than the motion measurements, with a non-negligible number of outlier detections