Simulataneous Localization and Mapping based on Inertial Navigation and Depth Image Information

Abstract

Environment map is needed to accomplish the task independently for mobile robot. However, robot cannot know the priori map of the environment in most cases. So it needs to generate incremental map during the exploration process. This paper mainly discusses mobile robot’s simultaneous localization and mapping (SLAM) problem in an unknown static environment without a priori map on the theory and practice. First, we study the design of strap down inertial unit of navigation system for mobile robot, exploring data of probe gyroscope, accelerometer and electronic compass etc. Due to accelerometer, gyro zero, and random drift error, kalman filter data fusion algorithm is proposed to combine with electronic compass data. Gesture angle and position information are computed after error correction. Secondly, acquisition and sensor of external environment by depth sensor is studied. Depth information collected by sensor is analyzed, and data in specific depth and height range is extracted. For easy description, three-dimensional coordinates in visual range are projected onto the ground plane. Then harris corner detection algorithm is used to extract the feature points of interest to build the feature map. Third, this paper proposes overall design of simultaneous localization and mapping. As it requires higher real-time data in the environment for mobile robot, so this paper chose relatively quick and simple arithmetic based on extended kalman filter SLAM method (EKF-SLAM). The specific process of EKF-SLAM is analyzed in this paper. Finally the method in this paper is proved integrity and feasibility by experiments.