| As an advanced control methodology, Model Predictive Control (MPC)
offers a lot of advantages for controlling urban traffic networks. MPC
predicts the future traffic states based on the prediction model, so as
to make long-term control decisions. MPC is robust to the uncertainty of
the process, which can be caused by the unpredictable disturbances, the
(slow) variation over time of the parameters, and model mismatches in
the prediction model. MPC can easily coordinate multiple intersections
and also multiple control measures. Another advantage of MPC is that one
can easily select and replace the prediction model based on the control
requirements. However, one problem that needs to be overcome when
implementing the MPC algorithm in a real-life traffic environment is the
on-line computational complexity.
Therefore, our research focuses on designing MPC controllers that are
effective and also efficient for application in large-scale urban
traffic networks. In order to build more efficient MPC controllers for
urban traffic networks, the following approaches are considered: First,
reducing the urban traffic model, which is taken as the prediction
model, to achieve more efficiency for the MPC controller. Second,
approximating the optimization problem by one that can be solved more
efficiently. Third, dividing the network into small sub-networks, and
building distributed controllers. Moreover, as performance objectives
for the MPC controllers we consider the reduction of the travel times,
the reduction of the traffic emissions, or a combination of both.
|
