Traffic congestion prediction using deep reinforcement learning in vehicular AD-HOC network | |
| Author | Chantakarn Pholpol |
| Call Number | AIT Thesis no.TC-20-01 |
| Subject(s) | Vehicular ad hoc networks (Computer networks) Traffic congestion Reinforcement learning Neural networks (Computer science) |
| Note | A thesis submitted in partial fulfillment of the requirements for the degree of Master of Engineering in Telecommunications |
| Publisher | Asian Institute of Technology |
| Abstract | As the number of vehicles on the roads are likely to rise dramatically in the future, traffic congestion is inevitable. In the previous works, there are many solutions for this problem; the most popular research topics are Vehicular Ad-hoc Networks (VANETs). The communication among vehicles and road side units is also a part of VANET, which helps to provide each other with information, such as vehicle velocity, location, direction, safety warnings and other traffic information. This results in the vehicles being able to avoid the congested route, as well as increasing their safety. However, the transportation industries will not stop developing. VANET may provide the solutions for traffic congestion, but sometimes it requires more flexibility and reliability to solve complex situations. Therefore, in this thesis, deep reinforcement learning in VANET is proposed in order to enhance the ability to predict traffic congestion on the roads. The target is to reduce travelling time and waiting time so that the vehicles are able to arrive the destination faster. After traffic scenario had been created in traffic simulator named Simulation of Urban Mobility (SUMO), it was integrated with deep reinforcement learning model. In deep reinforcement learning, the performance is getting better and better over the multiple runs until at the end, it will remain constant. In this work, the average travelling time delay and average waiting time delay were successfully decreased. The efficiency of the model is depended on traffic density and deep learning algorithms. If the traffic density is too high or too low, rerouting will not take place. Among three deep learning algorithms used in this thesis, i.e., Convolutional Neural Network (CNN), Multilayer Perceptron (MLP) and Long Short-Term Memory (LSTM), MLP is the most effective one. |
| Year | 2020 |
| Type | Thesis |
| School | School of Engineering and Technology (SET) |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Telecommunications (TC) |
| Chairperson(s) | Teerapat Sanguankotchakorn; |
| Examination Committee(s) | Poompat Saengudomlert;Attaphongse Taparugssanagorn; |
| Scholarship Donor(s) | His Majesty the King’s Scholarships (Thailand); |
| Degree | Thesis (M. Eng.) - Asian Institute of Technology, 2020 |