The Congestion Pie: Delay from Collisions, Potential Ramp Metering Gain, and Excess Demand

A method is presented to divide the measured total congestion in any freeway section into three ‘slices’: (1) the delay that collisions cause; (2) the potential reduction in delay at bottlenecks that ideal ramp metering can achieve; and (3) the remaining delay, due mainly to excess demand, but also to all other causes, including non-collision incidents, lane closures, and weather. The method involves two steps. The first step determines the space-time region that each collision affects and then estimates the additional delay it causes. The second step identifies bottlenecks and then estimates the potential reduction in delay that ideal ramp metering can achieve. The fully automated method can be applied to any site. It requires traffic volume and speed data, and the time and location of each collision. Applied to a 22.5-mile section of I-15N in San Diego, the method reveals that collisions, potential reduction by ideal ramp metering, and excess demand respectively account for 31%, 46%, and 23% of the total daily delay. Ideal ramp metering may overestimate the gain, because it assumes that ramps do not have storage limits. Nevertheless, the large gain (46%) signals the great potential to mitigate congestion by ramp metering. In addition to the three congestion pie slices, the method estimates the probability distributions of duration, spatial extent, and additional delay caused by individual collisions, which reveal, for example, that the delay from a collision depends significantly on traffic conditions.