| dc.description.abstract | This master's thesis presents the application of a robot localization process for a terminal tractor. To determine an accurate robot localization is essential for autonomous operations. This thesis investigates the use of non-visual localization sensors to estimate the position and orientation. There are several different methods for robot localization, but this thesis focuses on the use of sensors such as Global Position System (GPS), Inertial Measurement Unit (IMU), and wheel odometry.
The contribution of this thesis is to develop accurate robot localization, which can be used for autonomous vehicles. This has been done by investigating all individual sensors, as well as the overall performance of sensor fusion. This thesis gives an indication of the performance of the individual sensors and also the number of sensors necessary for maintaining acceptable performance. The robot localization software was implemented on the Robot Operating System (ROS) framework using Python, C++, Extensible Markup Language (XML), and Cmake as programming languages.
The robot localization process was tested by simulating the vehicle in Gazebo. By simulating the movement of the vehicle and simultaneously collecting measurement data from the sensors, a comparison was made with ground truth values for further analysis.
The results of the robot localization from the wheel odometry was not accurate enough. There was a deviation in the measurement data for linear motion trajectory, which affected the overall performance of the robot localization. However, the IMU sensor showed highly accurate performance for the robot localization.
The conclusion of the master thesis was that the implementation of the robot localization on the terminal tractor was a success, but the accuracy of the robot localization process during linear motion was not adequate. Based on this, more sensors should be incorporated into the system. | |