Optimal traffic control

Project members: B. De Schutter, T. Bellemans (K.U.Leuven, Belgium)

Sponsored by: FWO project ICCoS (Identification and Control of Complex Systems)

Congestion and traffic jams are one of the major socio-economic problems of today. Since building new roads is not always a feasible option, one of the most effective measures in the battle against traffic congestion seems to be a better control of traffic. Possible approaches to control traffic flows are traffic signals, variable message signs, dynamic route information panels, ramp metering, ``green waves'', route directives, radio broadcast messages, etc.

In this project we concentrate optimal ramp metering. In ramp metering a traffic signal is put at the on-ramp of the highway. When the signal is green, one car at the time is allowed to enter the highway. The switching scheme of the traffic signal (i.e., the occurrence of green periods) should be controlled in a such a way that the additional flow coming from the on-ramp does not cause the traffic flow on the highway after the on-ramp to exceed the critical density. In this way a smooth flow of traffic is guaranteed and traffic jams are prevented. Note however that the waiting time for the vehicles on the on-ramp should also be minimized.

The resulting problem can be solved using nonlinear optimization. However, because of its computational complexity this approach is not feasible for on-line adaption of the ramp metering policy to changes in traffic patterns. Therefore, we are now using a model predictive control (MPC) approach, which provides a balanced trade-off between computational complexity and global performance. In addition, we have improved an existing simple first-order model to describe the evolution of traffic flows in a motorway network [2]. The results of the computed ramp metering strategy are verified using micro-simulation, in which the movements of each individual vehicle are simulated (see Figure 26).

Figure 26: Screenshot of micro-simulation of ramp metering.