A huge influx of interest and funding is supporting advances in autonomous vehicles. This paper employs deep neural networks to extract features from images taken from cameras mounted on the left, right, and center of the front windshield. The synthetic images used in our work are generated from the Udacity platform, which is an open-source simulator developed based on Unity. We find that the features learned can be beneficial to the task of steering prediction. The proposed model has been experimentally verified in several environments. We use the desert environment as the training set and use the desert and mountain environments as the test set.
|Title of host publication||2020 Australian and New Zealand Control Conference (ANZCC)|
|Publication status||Published - Nov 2020|