Simultaneous Localization Auto-Calibration and Mapping of Ground Vehicles
Type of DegreeDissertation
MetadataShow full item record
This dissertation advances the state of the art of sensor auto-calibration by presenting a generic solution and analysis to the 3D sensor auto-calibration problem, as well as to the 2D simultaneous localization and auto-calibration (SLACAM) problem . Proper sensor calibration can be key to mission success when attempting to navigate robots in challenging environments. It allows for maximum information correlation between two sensors leading to higher fidelity representations of the environment thereby facilitating more precise navigation and obstacle avoidance, without which sub-optimal solutions such as taking longer paths to avoid obstacles to account for sensor error and other similar work arounds are employed. Additionally the ability to calibrate on the fly can greatly increase the robustness of a robot operating in challenging environments where canceling a mission prematurely is simply infeasible or not an option. The ability to auto-calibrate two rigidly mounted sensors while on a moving platform will be validated using a rigidly mounted lidar and an inertial navigation system (INS) on a sport utility vehicle (SUV) while driving through an urban center. The presented technique will be compared to an independent party’s static laboratory calibration, which will show agreement to within tenths of degrees in angular accuracy and centimeter level translational accuracies. Additionally, the observability of this technique will be assessed and unobservable maneuvers will be highlighted. Additionally the three (DOF) alignment of an inertial measurement unit (IMU), lidar, and odometer system on a ground robot that is performing simultaneous localization and mapping (SLAM) will be assessed. Additionally the performance of the SLAM algorithm will be assessed as well as the observability of this system. This technique will demonstrate the ability to produce a navigation solution that accurately navigates the ground robot through a structured environment to within a robots length of the initial position while calibrating the lidar and odometer systems relative to the IMU to within the state of the art of single sensor calibration techniques.