engleski

Model based Kalman Filter Mobile Robot Self-Localization

Mobile robot localization is an important module for mobile robot navigation. Known mobile robot pose is needed for path-planning and task planning purposes. Localization can be relative or absolute. This chapter will describe several methods that combine relative and absolute mobile robot localization assuming that initial pose is given also known as self-localization method. Self-localization means that after the mobile robot receives its initial pose it can determine its pose using only onboard sensor measurements and respective environment model during its movement. As a case study to demostrate proposed concept following onboard sensors are used: encoders for odometry, sonar sensor and monocular-camera. Sonar sensor is used combined with an occupancy grid model and Extended or Unscented Kalman filter for sensor fusion. Mono-camera is used combined with a 3D world model and Extended Kalman filter for sensor fusion. Implemented self-localizations methods are experimentally evaluated using a Pioneer 2DX/3DX mobile robot in a corridor like environment. Obtained results are compared regarding to obtained pose estimation accuracy, memory consumption and computational complexity.

mobile robot, odometry calibration, localization, Kalman filter, monocular vision, sonar sensor

nije evidentirano