California Partners for Advanced Transportation Technology

Traffic data measured near the junction of a single-lane on-ramp (with metered inflows) and a three-lane freeway were carefully studied for four days during the rush. The data showed the area around this merge junction became a bottleneck each day when the on-ramp's meter allowed its inflows to rise in the presence of high flows arriving from the freeway. Detailed study during these times further showed that queueing actually arose some distance downstream of the merge and that these queues were caused by drivers who, having just entered the freeway's shoulder lane from the on-ramp, slowed-down to negotiate discretionary lane-change maneuvers. The slowing was observed to spread quickly to all three freeway lanes and this had marked effects on the discharge flows from the bottleneck; i.e., average queue discharge rates were as much as 10 percent lower than the flows that had departed the merge prior to the bottleneck's activation. The latter of these flows, moreover, could be observed for many minutes (and therefore cannot be attributed to statistical variation). Even after the bottleneck had activated, its discharge flows continued to be influenced by vehicle slowing at, or just downstream of, the merge; i.e., queue discharge rates temporarily diminished with each recurring slow-down. Reductions of 800 vph or more could accompany such a slow-down and these lower rates could persist for some minutes. These undesirable effects were again triggered by vehicles from the on-ramp changing lanes soon after having merged. And notably, the discharge reductions became more (less) pronounced as the ramp's inflows surged (diminished). The findings show that service rates at this freeway site can be made higher 1) by eliminating, or at least postponing, the bottleneck's activation; and 2) by mitigating downstream slow-downs if or when the activation occurs. They further show, at times in dramatic fashion, that these objectives can be realized by carefully metering the on-ramp to limit the rates its vehicles join the freeway traffic stream. Certain issues concerning ramp metering in light of the present findings are discussed at the conclusion of this manuscript.