Reinforcement learning for self-driving cars

• 主要作者: Ho, Song Yan
• 其他作者: Xavier Bresson
• 格式: Final Year Project
• 語言: English
• 出版: 2018
• 主題: DRNTU::Engineering
• 在線閱讀: http://hdl.handle.net/10356/74098
• 機構: Nanyang Technological University

This project presents the implementation of deep learning model to act as a self-driving car- agent to maximize its speed on a multilane expressway. This project includes the development of traffic environment simulation, the design of neural network model, and the implementation of reinforcement learning algorithm. The proposed model uses the minimal sensory input collected from the environment. The model was trained with reinforcement learning algorithm in the simulation environment to simulate traffic condition of seven-lane expressway. The model successfully learns and applies the optimal policy. The model was tested under three different traffic conditions to determine its performance statistically. The best model is the model with neural network configuration that approximate to the optimal Q-learning function. The source code of this project can be found on https://github.com/songyanho/Reinforcement- Learning-for-Self-Driving-Cars.