Research On Global Localization Of Home Robot Based On Single Camera

The home robot will be used more and more widely with the development of the science and the improvement of the acquirement of people's life. In order to save the cost of the home robot, the robot with single camera as its outer sensor has become a popular research field during the past few years.We need a method to describe the environment so that the single camera can be used effectively. In this paper a novel algorithm named scale invariant feature transformation(SIFT) is used to extract the key points of the environment. The SIFT features are invariant to image scale, rotation and partially invariant to affine distortion. Thus, the features are highly distinctive. The environment can be correctly described using SIFT features and a map based on these features can also be obtained.Navigating safely in the home environment is the essential function of home robot. During the navigation, robot need to know its position and orientation in the environment. This process is global localization. The precondition of global localization is to have a map of current environment. The robot with stereo vision system can get the 3D information of the environment without moving, and the precision of the 3D information is satisfied. However, the robot who only uses single camera as its sensor has to move for a distance to obtain the 3D information of environment. This is structure from motion. While building map, the robot has to update its odometry data, then continue to create map. This process is called simultaneous localization and mapping(SLAM),and the localization during SLAM is local localization.Both local and global localization can be identified as an optimization problem. Here, the particle swarm optimization (PSO) was adopted to solve the local and global localization. A reasonable fitness function is proposed to optimize the poses of robot. The optimization updates the poses of robot in the SLAM process and makes the map more precise. The experiment results show that the PSO algorithm can also improve the precision of global localization.In addition, to speed up the search of map, a local sub-map search approach based on the principle which human use to search map is proposed. This approach improves the search speed greatly.