Enhancing Body-Mounted LiDAR SLAM using an IMU-based Pedestrian Dead Reckoning (PDR) Model

With the reduction in motion sensors' cost and power, Simultaneous Localization and Mapping (SLAM) has emerged as a core technology in several applications such as search-and-rescue, first-responders, and defence. Existing SLAM methods were designed mainly for robotic platforms that use wheel odometry. However, wheel odometry is not available for body-mounted platforms. This thesis addresses the challenge of body-mounted SLAM by proposing an integrated sensor fusion scheme. A Pedestrian Dead Reckoning (PDR) model based on inertial sensors is used to enhance LiDAR-based SLAM. This proposed fusion uses the PDR model as a replacement for wheel odometry in vehicular platforms. A system prototype has been developed and used for data collection and experiments. The implemented PDR model was integrated into the Google Cartographer SLAM engine and tested against different tracking systems. Experiments demonstrated that the integration of PDR has significantly enhanced head-mounted SLAM accuracy leading to accurate positioning under different motion scenarios.