Implementation of Visual Inertial Odometry Concept for Outdoor Autonomous Mobile Robot and Navigation Based on ROS2

Logistical operations in various industries are progressively getting replaced by the utilization of Autonomous Mobile Robots (AMR), which majorly focus on last-mile delivery in an indoor environment. On the other hand, the demands of operating AMR in an outdoor scenario gradually increase over time. Thus, this thesis focuses on the development of AMR based on ROS2, which is suitable for operation in outdoor settings, with a specialization of implementing the concept of Visual Inertial Odometry (VIO) by using Intel Realsense T265 tracking camera as an alternative approach for mobile robot localization and navigation that usually utilize a fusion of mechanically driven rotary encoders and IMU unit. The odometry data generated from the tracking camera displays an idle pose deviation of 0.05m with an average error percentage of 28.60% when subjected to linear movement and operated outdoors under sunny conditions. Ultimately, the outdoor AMR is capable of localizing itself by projecting its pose by utilizing the odometry data generated by the tracking camera, with a further integration that successfully enables the operation of autonomous navigation. On the whole, the results of this thesis are feasible as findings and sources of data that could be further elaborated on associated mobile robot developments.