D. Maio, D. Maltoni, and S. Rizzi. Dynamic clustering of maps in autonomous agents. IEEE Transactions on Pattern Analysis and Machine Intelligence, 18(11):1080–1091, November 1996. [url]
This paper present two incremental heuristic algorithms to maximise the clustering by discovery, a procedure to identify clusters in a map being learned by exploration as an agent moves in the environment.
The author state that clustering an environment map does not only mean partitioning a set of landmarks scattered in space according to some distance measure, but that the topology of the feasible connections should be taken into account.
The article contains a good literature of clustering techniques for autonomous agents and the menstion for the k-means and ISODATA methods. Also, the formalism of definition of the features of the heuristic is of high value. Finally, the proposed method is compared with others using four criteria: optimality, efficiency, robustness and stability.
Tags: spatial clustering