Real-time estimation of delay at signalized intersections

To evaluate improvements at signalized intersections it is important to know the resulting change in vehicular delay. However, it is difficult to collect delay data during over-saturated conditions even though this is when knowledge of delay levels is critical. Extensive peak hour queuing thwarts our ability to collect key data, such as arrivals at the back of queue. This incomplete information makes it impossible to calculate the resulting delay.;The research presents a real-time procedure for estimating delay during over-saturated conditions with limited information. The procedure utilizes a series of adjustments to the measured arrival rate entering the field of view to estimate the true arrival rate at the back of the queue. An advantage of the procedure is that estimated queues and associated delay are calculated on a second-by-second basis in real time. A disadvantage is that no theoretical relationship exists between the measured arrival rate and the real arrival rate.;Fortunately, it is possible to calculate a set of theoretical upper and lower bounds on the solution space by using historical minimum peak hour factors. The theoretical bounds take the form of cumulative arrival curves. Delay is obtained through consideration of the area between these arrival curves and the associated departure curve. Trajectory analysis during over15 saturated conditions is used to reconcile the difference between stopped delay and the area between the curves.;This research also demonstrates that the Highway Capacity Manual (HCM) definition of an initial (residual) queue is incorrect. To identify the true residual queue, the situation must be evaluated at the end of the red interval and thruput during the subsequent green interval must be deducted. Failure to do so leads to overestimation of both the initial queue and the corresponding delay.;Another finding is that the random component of the HCM's incremental delay term incorrectly contributes to delay during over-saturated periods preceded by an initial queue. A remedial modification to the d2 term is proposed.;Finally, it is demonstrated that the HCM's period-based queue accumulation procedure has drawbacks that can produce substantial errors in delay during over-saturated conditions. A remedial cycle-based counting technique is proposed.