New LiDAR simulation framework for GNSS/INS/LiDAR sensor fusion

20 July 2021

In order to aid our navigation module, which fuses data of the GNSS and INS, and implements a  PPP solution that uses HAS to get the satellite clock and orbit corrections, we have added a LiDAR sensor.

LiDAR can be quite advantageous in GNSS denied environments. We have decided to use the velocity estimates coming from the registration of a pair of point clouds as updates in our loosely coupled Kalman Filter (see picture below).

PictureLiDAR2.png

 

As LiDAR highly depends on the features of the environment, it is of our interest to convey an assessment of the reliability of these updates under different sensor configurations, scenarios and dynamics of the ego-vehicle. We use CARLA in order to get not only the generated LiDAR point cloud but also the vehicle state, which allows us to make the comparison between the actual velocity of the car and the estimated one. 

PictureLiDAR3.png

Moreover, it allows us to measure if the computed estimated position error fulfills our integrity requierments.

PictureLiDAR4.png

Together with the other simulation tools that have been developed at our institute, we intend to focus in the constant enhancement of our algorithms.