Self-driving through a Time-distributed Convolutional Recurrent Neural Network

This paper proposes an approach based on the use of time dimension within a Convolutional and Recurrent Hybrid Neural Network to carry out the driving of a simulated vehicle in real time. Convolutional layers are transformed into time-distributed layers that process a series of images in parallel and are related in time and space by recurrent layers. The experimentation showed an approximate 10% improvement over other autonomous driving approaches.