Real-Time Clustering for Long-Term Autonomy

摘要: In the future robots will have to operate autonomously for long periods of time. To achieve this, they need be able learn directly from their environment without human supervision. The use clustering methods is one possibility tackle this challenge. Here we present extensions affinity propagation, a algorithm proposed by Frey and Dueck [5], which makes it suitable real-time long-term in robotics applications. extension, called meta-point introduces so meta-points increase performance allows incremental usage. Additionally propose method that enables us obtain probabilistic cluster assignments any propagation based method.We show experimental results on quality speed as well assignments. Furthermore, demonstrate how process data sets much larger then what handle.