Principal Component Analysis (PCA) for Data Fusion and Navigation of Mobile Robots

摘要

A mobile robot system usually has multiple sensors of various types. In a dynamic and unstructured environment, information processing and decision making using the data acquired by these sensors pose a significant challenge. Kalman filter- based methods have been developed for fusing data from various sensors for mobile robots. However, the Kalman filter methods are computationally intensive. Markov and Monte Carlo methods axe even less efficient than Kalman filter methods. In this paper, we present an alternative method based on principal component analysis (PCA) for processing the data acquired with multiple sensors. Principal component analysis (PCA) is a procedure that projects input data of larger dimensionality onto a smaller dimensionality space while maximally preserving the intrinsic information in the input data vectors. It transforms correlated input data into a set of statistically independent features or components, which are usually ordered by decreasing information content. The learning rule for PCA is basically non-supervised and was first proposed by Oja in 1982. More advanced learning algorithms were also proposed for PCA including a neural network based approach called PCA network (PCANN). In this paper, we present a PCA network scheme for the sensor information processing of a mobile robot.