Mobile scouts sensing for military operations in mountain environments

다국어 초록 (Multilingual Abstract)

In military operations, an accurate localization system is required to navigate and direct troops to destinations in mountain environments. Traditional terrain information detection uses manual annotation of terrain features on maps, but this method is time-consuming and requires advanced knowledge of map construction. Previous work has used laser sensors to detect terrain information such as obstacles. However, the laser-based method is not applicable for human-centric tracking in mountain environments and requires expensive hardware. The global positioning system (GPS) is now widely used for localization, but GPS signals are unreliable and even susceptible to deception in the operational field. In addition, it is difficult to preserve the robustness of GPS-based localization against jamming of signals. With the widespread use of embedded sensors, a sensor-based pedestrian tracking scheme has become an attractive option. Pedestrian tracking technology provides practical location-tracking services in non-GPS environments. A typical pedestrian tracking system does not require any training ahead of use and is independent of environmental changes. However, pedestrian tracking systems using motion sensors such as accelerometer and digital compass suffer from a major drawback in that errors in estimated displacement accumulate over time because of noisy sensor readings. This thesis presents a group-based standalone system that automatically constructs terrain maps and simultaneously localizes soldiers in mountain environments. First, a pedestrian tracking system is described that uses group mobility information from several pedestrians forming stable moving clusters. This information is shared and coordinated to overcome the accumulated tracking error inherent in the inertial measurement unit (IMU). Experimental results from a prototype implemented in a smartphone show that location error is reduced by up to 43% as compared to traditional dead-reckoning tracking methods. Second, a pedestrian tracking system is described that automatically detects terrain features and constructs a terrain map for a mountain. The goal is to track user location and to provide an accurate terrain map for military operations without relying on the existing infrastructure. The scheme estimates terrain features using an inertial smartphone sensor and uses group information to improve the robustness of map construction. In a prototype implemented in a smartphone to evaluate its performance on a mountain, the proposed system successfully constructed a terrain map despite inherent tracking errors caused by noisy sensors. Finally, the system estimates landmarks using inertial sensors and utilizes group information to improve the robustness of map construction. Extensive evaluation confirms that our system successfully corrects system localization drift error without use of existing infrastructure.