End to End learning is a deep learning approach that has been used to solve complex problems that would usually be carried out by humans with great effect. A deep structure was designed within this study to simulate humans’ steering patterns in highway driving situations. The architecture of the network was based on image processing algorithm which is integrated with deep learning convolutional neural network (CNN). There are five aspects in this work, which enables the vehicle to detect the lanes, detect the speed of the vehicle, detect the angle of the road, recognize the objects on the road and predict the steering angle of the vehicle. A self-derived mathematical model is used to calculate the road angles for the prediction of vehicle’s steering angles. The model is trained on 2937 video frame samples and validated on 1259 samples with 30 epochs. The video of the local road was set as the output which will show the difference between actual steering angles and predicted steering angle. The experiments have been carried out in a newly built industrial park with suitable industry 4.0 standard design of urban smart development.
AutoMove: An End-to-End Deep Learning System for Self-driving Vehicles
Advs in Intelligent Syst., Computing
International Conference on Intelligent Computing & Optimization ; 2020 ; Koh Samui, Thailand December 17, 2020 - December 18, 2020
2021-02-08
15 pages
Aufsatz/Kapitel (Buch)
Elektronische Ressource
Englisch
Deep reinforcement-learning-based driving policy for autonomous road vehicles
IET | 2019
|