Uncertainty modeling for urban traffic model predictive control based on urban patterns

In a typical urban traffic network, there are many minor roads which are uncontrollable and unobservable. These small connections can be treated as uncertainties for urban traffic control. Currently, uncertainties are modeled as Gaussian white noise, additive or multiplicative to system states. This paper first proposes an uncertainty modeling algorithm in which the model can be updated by a clustering procedure based on daily measurements, urban patterns and other factors. After we formulate the macroscopic urban traffic model with uncertainties in an urban traffic network, the uncertainty model is added to the system states of a macroscopic urban traffic model (the BLX model). The performance of the model predictive control (MPC) in urban traffic networks using the proposed model is analyzed. The results show that the MPC with the uncertainty model performs better in reducing the total number of waiting vehicles in a network.