Data efficient reinforcement learning and adaptive optimal perimeter control of network traffic dynamics

Multiobjective Model Predictive Control Based on Urban and Emission Macroscopic Fundamental Diagrams

“…The Perimeter Control problem has become the main object concerning network-level control and different aspects of the problem have been deeply investigated [28], [29], [30] [31], [32]. Indeed, the Perimeter control problem is still nowadays capturing the attention of researchers [33], [34], [35], [36]. An alternative to Perimeter Control is the Routing strategy, also known as Route Guidance, used to suggest alternative paths offered by the road infrastructure.…”

Distributed Signal Control for Multi-Modal Traffic Network: A Deep Reinforcement Learning Approach

“…As the Reinforcement Learning (RL) technique enables traffic control, various studies applied two-region and multiregion PC based on the RL framework, showing comparable performance to model predictive control and high transferability [17,[24][25][26]. In [17,25,26], the perimeter of each PN is regarded as a centralized RL agent learning the metering policy from explorations during the training process, and all gates in the same perimeter act with a uniform metering rate.…”

“…In [17,25,26], the perimeter of each PN is regarded as a centralized RL agent learning the metering policy from explorations during the training process, and all gates in the same perimeter act with a uniform metering rate. In [24], the agent is designed to learn the macroscopic urban traffic dynamics for multi-region networks, and then a uniform metering rate is calculated for every perimeter based on the learned macroscopic traffic models.…”