Adaptive networked model predictive control of freeway traffic systems

Abstract

A networked traffic control scheme is presented in this paper. We refer to a realistic case in which the control actions are computed in a control center located far from the traffic system and then transmitted, through a wireless communication channel, to the actuators placed along the road, i.e. on-ramp traffic lights in the considered case of ramp metering control. More specifically, we assume that the communication channel is affected by delays and packet loss. We propose to adopt a control strategy of model predictive control (MPC) nature based on the use of a buffer to compensate for the time delays, so that the designed traffic networked control scheme provides performance significantly close to those of the ideal case (i.e. that with no delay). Moreover, in order to limit the computational load, a modified version of the control strategy is proposed, in which the length of the optimization horizon of the predictive control algorithm, and, consequently, the buffer length, is updated relying on a delay estimation. A simulation analysis of the different networked control proposals applied to a Dutch freeway is provided, along with the comparison with a conventional MPC equipped with an hold mechanism. The superiority in terms of performance of the proposed networked MPC with adaptive buffer length is apparent.