The bottleneck mechanism of a freeway diverge

This paper describes the behavior of multi-lane freeway traffic, upstream of an oversaturated off-ramp. It is based on empirical evidence from freeway I-880 (northbound) near Oakland, CA. The main findings are: (1) FIFO blockage. Even on wide freeways, an off-ramp queue can grow across all lanes and entrap through vehicles in a first-in-first-out (FIFO) system with similar speeds on all lanes and a well-defined kinematic wave (KW). This can hamper freeway flow much more than an on-ramp bottleneck. (In our case the FIFO regime arose a little over 2 km upstream of the exit and reduced discharge flow approximately to an average of 1500 veh/h per lane across all lanes.) (2) Variable capacity. Under FIFO, the freeway discharge flow can change significantly without a change in the off-ramp flow when the percent of exiting vehicles changes. (In our case it increased to 5750 veh/h after 30 min at 4520 veh/h when the exit percentage declined fom 29% to 24%.) (3) Non-FIFO congested regimes. Multi-pipe traffic states, where queued lanes move at different speeds, can persist for a long time. Presumably, this happens because different drivers prefer different lanes depending on their destination. (In our case, multi-pipe queued regimes were detected immediately upstream and immediately downstream of the FIFO queue.) Semi-congested traffic regimes, where some lanes are queued and others are not, also exist. (They were detected donwstream of the FIFO queue.) (4) Regularity of the flow-density scatter-plots. If one excludes the periods of time corresponding to multi-pipe flow and regime transitions, flow-density points across all lanes are quite close to a ‘fundamental diagram’ with two branches. During multi-pipe periods, flow-density scatter points cluster along an inner line, parallel to the congested branch, because drivers give themselves longer spacings when driving under these conditions.