Autonomous driving through non-uniform steering angles nodes determination by deep learning

In this work, an autonomous driving model using only one camera was implemented by combining a CNN (Convolutional Neural networks) and a YOLO (You Only Look Once) framework. Hyper-parameters in the structure were adjusted to improve driving performance. Autonomous driving in a corridor was performed by applying the improved model. An appropriate dropout and deep learning structure associated with non-uniform steering angle intervals as output is proposed. The proposed algorithm was implemented, and through experiments resulted in successful obstacle avoidance and stable driving.