| Title |
A CNN Model for Image based Behavior Learning of Path Tracking Simulated Driving Vehicle |
| Authors |
이종석(Jongseok Lee) ; 조영완(Youngwan Cho) |
| DOI |
https://doi.org/10.5370/KIEE.2020.69.6.930 |
| Keywords |
AutoPilot; Artificial Neural Network; Deep Learning; CNN |
| Abstract |
In this paper, the driving behavior of autonomous driving algorithm is modeled by artificial neural network. Autonomous driving algorithms is determine the best behavior for a destination by entering all the information in the simulator. Thus, the behavior of the autonomous driving algorithm ensures the assumption of the best behavioral decision in each state. In the CARLA autonomous driving simulator, we study the neural network that controls vehicles by applying deep learning. we study the driving behavior of AutoPilot, the autonomous driving algorithm of CARLA simulator, through artificial neural network. We specified a target route for the AutoPilot vehicle and reconstructed the driving data from driving into training data. There are two methods of learning. The first method consists of the position, direction, and driving behavior of the AutoPilot vehicle. In the second method, the neural network learning of the CNN and ANN structure was conducted by constructing the driving image, position, direction, and driving behavior of the vehicle. The artificial neural network selects an action corresponding to each state and drives a path through the selected action. In this study, the experiment analyzed the difference between the driving route of AutoPilot and the driving trajectory of the learned artificial neural network. |