Effectiveness evaluation of view-based navigation for obstacle avoidance

In this study, we propose an improved view-based navigation method for obstacle avoidance and evaluate the effectiveness of the proposed method in actual environments existing obstacles. The proposed method is able to estimate the position and the rotation of a mobile robot, even if the mobile robot strays from a recording path for avoiding obstacles. In order to achieve this, ego-motion was employed to the view-based navigation. The ego-motion is calculated from SURF points between a current view and a recorded view by using a Kinect sensor. In conventional view-based navigations, it is difficult to generate other paths to avoid obstacles. By using the proposed method, it is expected that a mobile robot plans flexible paths and avoids persons and objects in actual environments. From experiments performed in an indoor environment, we evaluated the measurement accuracy of the proposed method, and confirmed the feasibility of the proposed method for actual robot navigation.