AN INTEGRATED TRAFFIC RESPONSIVE RAMP CONTROL STRATEGY VIA NONLINEAR STATE FEEDBACK

Ramp metering has been an effective tool for combating freeway congestion, and past research has produced a variety of ramp metering algorithms. One particular approach based on the Linear Quadratic (LQ) theory is appealing because it is traffic responsive and robust. However, this approach is based on a linearized equation of the nonlinear traffic flow system, thus its performance is limited when traffic conditions deviate far from steady state. We propose here a nonlinear controller to overcome this problem. The nonlinear controller is realized by a series of neural networks, whose parameters can be obtained through nonlinear optimization. Preliminary traffic simulation results are presented and demonstrate the controller's potential effectiveness.