System of IMU, GPS, LIDAR, stereo camera and magnetometer data fusion for self-driving car localization

Izvin Andrei Valerevich

This work describes the localization system of an unmanned vehicle, which consists of IMU, GPS, LIDAR, stereo camera and magnetometer sensors. Unscented Kalman filter is responsible for merging the data produced by these sensors. System performance is tested on synthetic data generated by the Carla simulator. The car’s state values predicted by the system are compared with the ground truth values. As a result of comparison it is established that the system determines the position, speed and direction of the vehicle with accuracy ± 4.13 cm, ± 0.08 m/s and ± 0.01 rad, respectively. This makes it possible to use that system in practice.

Keywords:localization, Kalman filter, GPS, IMU, stereo camera, LIDAR, magnetometer, unmanned vehicle, simulator, data fusion, Carla