An Adaptive Clustering Algorithm for Autonomous Agents

P. Baruah and R.B. Chinnam (USA)


Adaptive clustering, kernel density estimation, intelligent systems.


: This paper proposes an adaptive clustering method for development of completely autonomous agents. The method uses a non-parametric clustering algorithm that can automatically detect the number of clusters present in the dataset. For detecting the number of clusters, the algorithm employs Gaussian kernel density estimation in combination with a mode hunting procedure. The proposed algorithm is also adaptive in that it can perform cluster tracking by employing an exponential weighted scheme for updating the cluster parameters. There is a provision to optimize the adaptation rate as well. The proposed set of methods for clustering is particularly suitable for developing intelligent systems where the primary goal is to make the system autonomous and adaptive. The paper also presents results from several simulation studies and a real world equipment diagnostics problem.

Important Links:

Go Back