PERCEPTION-BASED LOCALIZATION SYSTEM FOR AUTONOMOUS VEHICLE USING NATURAL LANDMARK TO SUPPORT NAVIGATION SYSTEM BASED ON HIGH DEFINITION MAP

• Main Author: Lukmanul Hakim, Muhammad
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/75423
• Institution: Institut Teknologi Bandung

Localization systems have developed rapidly along with the development of autonomous vehicles. The need for navigation makes the localization ability of the vehicle a must-have. For the development of the current localization system, it is satellite-based, namely using GPS and using SLAM. During the development of the satellite-based system, it experienced a decrease in accuracy when it was in an environment that was not conducive, such as when it was in an area with many tall buildings or when it crossed a bridge. In autonomous vehicles, features for mapping have been developed, namely HD Maps which have several levels. The feature level is one that is developed by utilizing existing objects around the environment with the help of a perceptual system. So that a localization system can be developed that can utilize natural landmarks to obtain the location of the vehicle on HD Maps. This localization system will utilize existing landmarks in the environment so there is no need to add new landmarks as reference points. Natural landmark used in this system is road sign. This system will use a stereo camera as an object detector and image processing will be carried out. This processing is done to read the text contained in the object. Processing is done using the EAST algorithm that has been developed to detect text on natural landmarks. Position estimation is carried out by autonomous vehicles against natural landmarks that have been detected so that the relative localization of the vehicle will be obtained. Global localization can be obtained when knowing the geographic coordinates of landmarks globally. The results obtained from the system show a fairly good value for the accuracy and precision of the 11 points obtained. This is indicated by the standard deviation values obtained which are 2.3 cm in the lateral direction and 4.7 cm in the longitudinal direction. And the average MAE obtained is 16.9 cm for the lateral direction and 2.7 cm for the longitudinal direction