Vision-based Monte Carlo - Kalman Localization in a Known Dynamic Environment

Localization is one of the fundamental problems in mobile robot navigation. In this paper, we present a vision-based localization method called Monte Carlo-Kalman localization (MCL-EKF). This method is a combination of Monte Carlo localization (MCL) and extended Kalman filter (EKF) enhancement. We firstly give a detailed implementation of MCL with the emphasis on dealing with multiple types of perceptual information and solving the problem of robot kidnapping. Next, we establish EKFs on landmarks to build a real-time environment around the robot. Information from this real-time environment will be utilized by the perception model of MCL. We also elaborate on our methods of dealing with a single or two landmarks in the perception model. We carry out all experiments on Sony AIBO ERS-7 robots. Results show that the MCL-EKF reduces perceptual errors, increases precision and stability and still keeps a good ability of recovery