Localization and feature recognition: Implementation on an indoor navigator robot

Abstract

Oriented FAST and rotated BRIEF, or ORB, is a binary algorithm with rotation invariant and noise resistant abilities. ORB, which was built on the FAST key point detector and the BRIEF descriptor, can be applied for object recognition. This paper studies the feasibility of ORB implementation in indoor navigation with Lego Mindstorms NXT robot. The robot was capacitated to explore and navigate autonomously under a known environment, based on its decision making ability. An Android device was attached to the robot, providing the vision and meanwhile performing complex computation such as an autonomous navigation system. Communication amongst the Android device and the robot was performed with both wireless and Bluetooth transmissions from a computer. In this project, the robot with recognized the landmark using the ORB feature detection. Along with the ultrasonic sensory data, the robot localized itself by utilizing Monte Carlo Localization method. Besides, A* shortest path algorithm was applied to plan the shortest path for the robot to reach the goal. Analysis was made on the collected results and the failure was discussed. Subsequently the conclusion was drawn and some recommendations for improvements were made.