A stochastic optimal control based approach to real-time incident-responsive local ramp control

This paper presents a stochastic optimal control based methodology for real-time local ramp control in response to lane-blocking incidents on freeways. A discrete-time nonlinear stochastic system, which characterizes the time-varying relationships among lane traffic states, control variables, and point detector data, is proposed. In addition, a stochastic optimal control-based algorithm is developed to update the time-varying ramp metering control variables and incident-induced lane traffic states in real time. Preliminary test results indicate that the proposed ramp control method permits alleviating incident impacts efficiently particularly under low-volume and medium-volume incident conditions. Moreover, utilizing the proposed method, time-varying lane traffic states together with incident impacts on queue lengths and capacities can be estimated in parallel with incident-responsive ramp control to monitor the status of incident impact and provide incident-related traffic information for further applications, e.g., incident management.