Highway performance

In an article in ACCESS (UCTC Spring 2000 p22) Pravin Varaiya, a computer sciences engineer at Berkeley writes about a Performance Management System (PeMS) he has designed to analyze loop data. It goes beyond measuring speeds and delay – already being done at scores of traffic management centers so-called – and measures aggregate highway service: vehicle-miles traveled/ vehicle hours-traveled weighted by number of vehicles. That’s average speed times vehicles or vehicle-mphs. With vehicle occupancy data you get person-mphs, and the item carpooling enthusiasts like to maximize.

Now, the familiar parabolic shape of the graph of speed versus volume in a motorway lane describes how beyond the point of capacity, more entering vehicles reduce the volume of vehicles handled per hour as traffic flow breaks down and average speeds drop precipitously. This is the heart of the case for traffic management – that unmanaged entry to a busy traffic lane beyond the point of ‘capacity’ causes service for all to quite suddenly drop away. Each entering vehicle at that point not only suffers stop-&-go itself, but imposes stop-&-go on hundreds if not thousands of other motorists, as drivers progressively hit their brakes to maintain a comfortable headway from the vehicle in front. There’s a classic economic externality, and the opportunity for a win-win deal for motorists and a toll authority alike in raising the toll rate enough to prevent saturation of capacity.

The point at which saturation is reached seems to vary by as much as 20% from day to day and from road to road, Varaiya notes. Various hypotheses are being tested to track down the causes of premature saturation. A cure for this condition could yield the equivalent of an extra lane on a 4-lane roadway.

He writes: “We know in principle the freeway system can be operated (at optimum loading to prevent saturation) by the proper control of ramp metering and advisory messages.” In principle, maybe, but in practice only in limited circumstances. Ramp metering can help to reduce turbulence in the traffic flow by breaking up the flow of entering traffic into separate vehicles that can merge into the mainline. It also works to prevent overload on the mainline, but at the cost of transferring the congestion to the ramps. When traffic backs-up down the ramps to the cross-streets at the bottom of the interchange, then chaos threatens on the approaches and exits. The signal managers soon have to give up on ramp metering. That’s apparently why ramp-metering is of limited use in the congested Los Angeles area. Advisory messages only help where there is useful advice to be given - like a less congested alternate route. If all the alternates are clogged up, there’s nothing useful for the signs to say.

The only known method of reducing the demand for trips - as opposed to storing it on ramps, or diverting it with messages - is through a variable toll rate. Varaiya should have a look at 91X and I-15’s HOT lanes and see if his PeMS has more applicability in that setting.