Risk-averse perimeter control for alleviating the congestion of an urban traffic network system with uncertainties

The perimeter control method is an effective way to alleviate traffic congestion that is based on Macroscopic Fundamental Diagrams (MFDs). However, the strategy may lead to congestion when it ignores the uncertainty of MFDs. To address this problem, this paper presents a risk-averse perimeter control method. First, the urban traffic network system with uncertainty is modeled using a neural network and scenario tree. Then, this research quantifies the congestion risk caused by uncertainty using an average value-at-risk. The next step sees the design of a risk-averse model predictive control (MPC) controller that takes the multi-stage risk as the optimization objective and improves robustness by interpolating between the conventional stochastic and worst-case MPC formulations. Finally, this paper analyzes the risk-sensitive stability of an urban traffic network system and gives a solvable form of risk-averse optimal control for this system. Finally, two simulations are conducted to verify the presented method's validity and superiority for an urban traffic network system with uncertainties. The simulation results show that the risk-averse perimeter control method presented by this paper is superior because it reduces the total travel time by 12.98% compared to Stochastic MPC, by 15.96% compared to bang-bang control, and by 14.54% compared to proportional-integral. This paper designs a risk-averse model predictive controller to track the urban traffic network system with uncertainty. This method significantly alleviates congestion and reduces the travel time of the traffic system.