Abstract
The authors have been developing an autonomous mobile robot system to be used in the campus of universities. Our robot software adheres stochastic models, where the probability density function is a mixture of several basic distributions. It is well-known that Expectation-Maximization (EM) algorithm is useful for the identification of Gaussian mixture distribution. We apply EM algorithm for our mixture model consisting of Gaussian distribution, two Exponential distributions, delta distribution and a uniform distribution. In the experimental study, we will show some Monte Carlo simulation results for several cases so that this algorithm is shown to be useful for this mixture distribution model. Next, we will apply our algorithm to real data. Finally the parameter values obtained in our algorithm will be applied to the state estimation problem. The usefulness of this algorithm will be discussed.
