Implementation of Lane Change Maneuver System using Neural Network

In order to give the user a safe driving experience, research on autonomous vehicles and advanced driver assistance systems (ADAS) has increased recently. To prevent vehicle collisions and to drive safely on highways, one of the trickiest moves, lane changes, must be anticipated in advance. For these kinds of applications, where different types of neural networks are utilised to anticipate lane changes in autonomous vehicles.A new structured deep neural fusion network is proposed by combining two Convolutional Neural Network (CNN) pre-trained models which are trained on BDD100K dataset. The fusion network predicts the possible lane change by calculating the distance between the ego lanes and surrounding vehicles. In comparison to the fine-tuned RseNet-50 and GoogLeNet models, the proposed fusion network performs better with accuracy of 96%.