Enhanced vehicle localization with low-cost sensor fusion for urban 3D mapping.

This research paper presents the design and development of an indigenous low cost Mobile Mapping System (MMS) for urban surveying applications. The MMS is comprised of economical Hokuyo-30LX 2D laser scanners, vision sensors, Global Positioning System (GPS) and various odometric sensors that can be installed on car like moving platform. The run time sensorial data is interfaced, processed and recorded using Robot Operating System (ROS). The live laser scan is utilized for the pose estimation using Simultaneous Localization and Mapping (SLAM) technique. In absence of valid SLAM estimation and frequent GPS outages, a multimodal sensor fusion framework for the enhanced pose correction has been developed using Kalman Filter (KF) by incorporating the Inertial Measurement Unit (IMU) and wheel odometric data along with SLAM and GPS data. The corrected pose is utilized for the 3D point cloud mapping by incorporating laser scans perceived periodically from various 2D laser scanners mounted on the MMS. The custom-made installation scheme has been followed for mounting three 2D laser scanners at horizontal, vertical and inclined orientations. The efficacy of the developed map has employed for extraction of road edges and associated road assets by establishing the lucrative classification technique of the point cloud using Split and Merge segmentation and Hough transformation. The surveying to map development time has significantly reduced and the mapping results have found quite accurate when matched with the ground truths. Furthermore, the comparison of the developed maps with ground truths and GIS tools reveals the highly acceptable accuracy of the generated results which have found very nearly aligned with the actual urban environment features. In comparison to the existing global MMS variants, the presented MMS is quite affordable solution for limited financial resourced business entities.