The Development of a Mapping Hovercraft for GNSS Denied Environments Utilizing a SLAM Aided Navigation Scheme

Abstract

Indoor navigation as well as pipeline detection has been under attention in recent years due to the device and technology developments with various requirements from urbanization. Thus, GNSS-denied environments are the targets to complete the mapping constructions or positioning applications. In this research, indoor and underground environments are detected and spatial information is acquired by laser scanner as further Simultaneous Localization And Mapping (SLAM) algorithm input. On the other hand, Inertial Navigation System (INS) is aided for improving position accuracy during scanning processes. The payload is mounted on either mobile robot or hovercraft in relative size of the regions based on the concept of Mobile Mapping System (MMS). Based on the Kalman Filter, the data from INS and SLAM can therefore be combined. The concept of the Kalman Filter is to minimize the variance estimations through conditional expectation. However, the uncertaintiesin the sensors and the effect of the gravity field lead to the errors in the navigation parameters. The drifting problems therefore mounting when the scanning time and distance increase. In order to reduce the influence from mentioned errors, optimal smoothing is included and different error models are under discussion in this research. Smoothing is a method to estimate the values with the measurements from both the past and the future. Once the captured position accuracy is enhanced by these additional methods, the laser scanner is able to provide accurate measurements to reconstruct the environments into the maps. This research provides the rigid/regular payload for various platforms in related scenarios with INS as well as the laser scanner and the combination of the sensors’ data depends on the synchronization of receiving time by Kalman Filter (KF). The main focus emphasizes the capability of the payload on different platforms for further applications and the ability of the algorithm for accurate positioning so far by individual constraints.