Fully-Autonomous, Vision-based Traffic Signal Control: from Simulation to Reality

Keyphrases

• Traffic Condition 100%
• Vision-based 100%
• Deep Reinforcement Learning (deep RL) 100%
• Traffic Signal Control 100%
• Simulation to Real 100%
• Domain Randomization 100%
• Training Data 50%
• Generalization Ability 50%
• Signal Control 50%
• Real Environment 50%
• Fundamental Challenges 50%
• Congestion 50%
• Unseen 50%
• Control Agents 50%
• Effective Techniques 50%
• Traffic Control 50%
• Visual Data 50%
• Economic Constraints 50%
• Safety Requirements 50%
• Validation Set 50%
• Traffic Situation 50%
• Traffic Environment 50%
• Reinforcement Learning Agent 50%
• Complex Traffic Environment 50%
• Trained Agent 50%
• Positive Transfer 50%
• Live Visuals 50%
• Reality Gap 50%
• Live Traffic 50%
• Urban Road Network 50%
• Traffic State Estimation 50%
• Dynamic Traffic Environment 50%
• Adaptive Signal Control 50%

Computer Science

• Traffic Condition 100%
• Control Signal 100%
• Deep Reinforcement Learning 100%
• Traffic Environment 100%
• Traffic Signal 100%
• Training Data 50%
• State Estimation 50%
• Dynamic Traffic 50%
• Learning Agent 50%
• Traffic Control Agent 50%
• Validation Set 50%