Return to Andrea Premate's personal page
Return to USAD, main project page

The aim is to make it possible to acquire datasets consisting of the measurements obtained from the on-board sensors of the vehicle, such as cameras, LiDARs, imu and gps. Once the acquisition and registration system is completed, it will be necessary to acquire a dataset to test localization, point-cloud registration and pose calibration approaches between sensors developed within the lab. By achieving the above objectives, it will be possible to build datasets to train and test machine learning models with respect to the navigation environments within which the USAD vehicle operates instead of using only publicly available datasets.
The steps to be addressed are as follows:

1. Using traditional methods to estimate the relative pose between sensors (calibration).
2. Implementation of the acquisition system using ROS (Robot Operating System).
3. Acquisition of the dataset with USAD vehicle in urban context.
4. Any improvements in the results related to point 1 using DNN approaches developed in the laboratory.