Authors :
Abhinav Singh; Nishant Prakash; Kanishk Bhaskar; Krishnan Rangarajan; Aditya Raj
Volume/Issue :
Volume 8 - 2023, Issue 2 - February
Google Scholar :
https://bit.ly/3IIfn9N
Scribd :
https://bit.ly/3IuyzKa
DOI :
https://doi.org/10.5281/zenodo.7664562
Abstract :
The papers that we are surveying have many
methods that have been presented with different solutions
for autonomous driving. One of the few novel
representations helps in proving the reasoning for
imitation learning in a certain scene where the cameras
are used to highlight a certain location which coordinates
to waypoints and semantics. In this method the camera
follows the car and will show the waypoints at a certain
distance ahead of the car at all times while the car is
moving. The papers have used attention fields to
compress two-dimensional images with features which
are best suited for cognitive processing on a discrete
aspect of information or in other words obstacles that
may appear in front of the car. Therefore, the other
model being a Multi-Modal Fusion Transformer is used
to combine two separate datasets such as image data and
topography data from cameras and distance sensors
respectively using attention mechanism. This helps in
integrating image data and the topography data that is
being received through the camera and distance sensors.
The distance sensor maps the surface of all the
surroundings where the car is being driven.
Keywords :
End-to-End Autonomous Driving, Transformer, 2D Imaging, Self-Attention Model, Imitation Learning.
The papers that we are surveying have many
methods that have been presented with different solutions
for autonomous driving. One of the few novel
representations helps in proving the reasoning for
imitation learning in a certain scene where the cameras
are used to highlight a certain location which coordinates
to waypoints and semantics. In this method the camera
follows the car and will show the waypoints at a certain
distance ahead of the car at all times while the car is
moving. The papers have used attention fields to
compress two-dimensional images with features which
are best suited for cognitive processing on a discrete
aspect of information or in other words obstacles that
may appear in front of the car. Therefore, the other
model being a Multi-Modal Fusion Transformer is used
to combine two separate datasets such as image data and
topography data from cameras and distance sensors
respectively using attention mechanism. This helps in
integrating image data and the topography data that is
being received through the camera and distance sensors.
The distance sensor maps the surface of all the
surroundings where the car is being driven.
Keywords :
End-to-End Autonomous Driving, Transformer, 2D Imaging, Self-Attention Model, Imitation Learning.
