الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 119 |  |  |
| | | from | 119 | | |
Abstract
Recently, there has been a lot of interest in autonomous vehicles (AVs) because of their overall safety and comfort. However, the unstructured environments, with no prior high-detail maps available, are still challenging to AVs and can affect the performance of the localization system.
This research aims to develop and validate different AVs localization systems in unstructured environments. The specific research purpose of this work is to take steps towards a more capable and efficient autonomous driving localization system suitable for autonomous driving in unstructured environments and explore potential areas for future work.
This is done by conducting real-world experimental testing of different algorithms in a challenging unstructured environment.
Also, vehicle understanding of the surrounding environment is done using segmentation based on a very deep convolutional neural network, which is accompanied by a very high computational load.
This research proposes a bio-inspired novel method to reduce the computational time of the segmentation task by mimicking the human eye’s vision in terms of dividing the wide field of view into two zones: the peripheral vision and the central vision based on the target vehicle path. This method resulted in a reduction of the computational time by 19.7%.
Keywords:
Autonomous vehicles, global localization, on-road localization, sensor fusion, and unstructured environments.