Mobile robot map building in eigenspace — A pea-based approach

Detection and compression of the environmental information incrementally is useful when the mobile robot needs to continuously update its perception database online. In this paper, we propose an incremental subspace fusion method for abstract map building of mobile robot. We firstly adopt PCA to preprocess the panoramic images taken from omnidirectional camera at corresponding location, which can optimally extracts the local map features formed in the subspace and reduces the uncorrelated features. This enables robot to memorize the environmental features incrementally when it travels in an unknown environment. Experimental results are shown finally.