A Hovercraft Visual Simultaneous Localization and Mapping System Performance Evaluation for River Monitoring Application

The dynamic geographical shape of a river provides a unique challenge in adopting an autonomous hovercraft-like robot for river monitoring. Specifically, performing the mapping and localization task in such an environment is non-trivial. Few factors need to be considered including the dynamic of the hovercraft and the severity of the surrounding environment. In this study, an investigation was carried out to evaluate the performance of a visual simultaneous localization and mapping (SLAM) technique in generating the environment model of a simulated river. The main objective of this study is to build a SLAM system of a hovercraft such that the robot can traverse the river surface autonomously. A depth camera was used as the visual sensor, and acts as the main odometry in the process of mapping and localization of the hovercraft. Data from the visual sensor was tuned and optimized to adapt to the river surface environment. The performance of the tuned SLAM was measured by comparing several configuration settings to get the best output. The result of this study shows that with proper tuning and optimization, a SLAM technique can be used to generate an accurate river surface map and monitor the hovercraft trajectories.