Title Page
Contents
Abstract 10
Chapter 1. Introduction 12
1.1. Backgrounds 12
1.1.1. Introduction of V2X communication 12
1.1.2. Introduction of RSU in V2X communication 14
1.1.3. Reinforcement Learning (RL) in V2X 15
1.2. Contributions of this paper 16
Chapter 2. Related Works 19
2.1. Resource Allocation Using RL in V2X Communications 19
2.2. RSUs in V2X communications 22
2.3. Heterogeneous Traffic 23
Chapter 3. System Model 27
Chapter 4. Multi Agent Reinforcement Learning Model 31
4.1. Agent 33
4.2. State 33
4.3. Action 34
4.4. Reward 35
Chapter 5. Performance Evaluation 40
Chapter 6. Conclusions 49
Bibliography 51
TABLE Ⅰ. V2X MESSAGES AND THEIR CHARACTERISTICS 26
TABLE Ⅱ. SIMULATION PARAMETERS 42
TABLE Ⅲ. MARL PARAMETERS 43
Figure 1. System model with heterogeneous V2X traffic in an underload situation (N=10, K=3, M=3 and D=2) 28
Figure 2. Used MARL structure 32
Figure 3. Global reward per episode 44
Figure 4. PRR performance of optimal, random, and proposed MA-DQN resource allocation schemes. (a) Underload situation with 6 transmitting vehicles and 1 RSU. (b) Overload situation... 45
Figure 5. PRR performance of CAM and DENM messages with the proposed MA-DQN scheme. (a) λR=0.5. (b) λR=0.7.[이미지참조] 46
Figure 6. Communication range for CAM and DENM messages with different DENM priority constant λR.[이미지참조] 48