If we can agree on some assumptions regarding your PRT pulse problem, perhaps a simple and reasonable calculation can be made. Suppose you have a PRT station located near an office building and that 80 lunch-bound persons leave the building between 11:55 and 12:05 and walk (2 minutes) to the PRT station. They arrive at the station between 11:57 and 12:07 according to a suitable arrival distribution.
Suppose the PRT station has five berths, the load/unload time is 15 seconds on the average, vehicles can load/unload simultaneously and some of the 80 people know each other and are going to lunch together. The "system" knows that there is a pulse of about this size at this station every weekday and is able to direct some empty vehicles that are moving on the system to this station, beginning at about 11:50. It can also direct other empty vehicles located at a nearby (5 minutes or so) storage depot to go to this station. And, suppose that the system has sufficient slots available to accommodate these empty vehicles with little wait time.
Let's assume that the average group is 1.5 persons (53 vehicles needed for 80 people). Let's say fifteen vehicles can arrive, load/unload and leave the station in 1 minute, on the average. This would be a station throughput (outbound) of 22.5 people per minute or, say 80, in 4 minutes - assuming that the necessary vehicles arrive in a timely manner, remembering that some arrive with occupants, others arrive empty. Since these people arrive at the station during a 10 minute period, it would appear that they (and even more) could be served with little wait time or delay.
If you don't buy these assumptions, please make your own and use them to make a counter-calculation. There are several simulation models that make this calculation in a more sophisticated way with assumptions like these. They, so far as I know, have produced results that indicate that pulses of reasonable size at stations of the proper size are not a fatal problem for PRT.
Boarding 80 people in a short time is not a "fatal problem", but its analysis shows that PRT cannot perform as efficiently as theories about it claim, and that PRT is distinctly inferior to other modes. The reasons are basically the difference between an ideal theoretical situation and limitations in real-world applications, such as:
-- 1) Provision of 53 vehicles assumes that there are sizable "storage guideways" in the immediate vicinity, or that many vehicles are cruising empty. When you design a transportation system in a city, you see that storage areas are not always easy to find. Cost of many cruising vehicles is not negligible. Scheduling 5-10 vehicles for a certain moment may be feasible, but scheduling 50 of them would be a complex task.
-- 2) The boarding process you describe would require platooning of vehicles, 5 at a time, and there is no way one could get three such platoons per minute. Boarding rates would be considerably lower.
-- 3) Getting 5 vehicles into a platoon would require some efforts; scheduling one such platoon to arrive every 20 seconds - or even every 2 minutes, is quite unrealistic. If platoons "catch up" with each other (a single person delays boarding of one vehicle in a platoon), they will either back up into the main line, stalling the network, or you must have a long off-line guideway which increases the size and cost of the system.
-- 4) Even if we suppose that your assumptions were realistic, try to simulate the same operation with PRT's "first older brother" - the Morgantown system. The simulation should show you much higher efficiency of the latter mode.
Vukan seems to assume that workers in PRT base their findings on theoretical analyses of parts of a network. This may have been true in the mid 1960s; however, since then many investigators have developed simulation models of network PRT systems in which questions such as: "What happens if a bus load of people arrive at a PRT station?" can be fully answered. One of the requirements of the Phase I study of PRT for the Chicago RTA required us to show, based on a network simulation, the performance of a PRT system subject to a pulse of demand at any station. The result was fully satisfactory.
A number of comprehensive simulations of PRT systems were developed in the 1970s. The ones I am most familiar with are those developed by The Aerospace Corporation (networks with up to 1000 stations and 60,000 vehicles), the German joint venture DEMAG+MBB (networks covering medium sized cities), IBM Corporation (networks up to a size needed for Manhattan Island), and the University of Minnesota (networks for the Twin Cities and Duluth). More recently, I developed a simulation in which I can simulate a PRT network of any configuration, and have done so most recently for an application in Korea. Ingmar Andraesson , Gothenburg, Sweden, has developed a simulation in which he has been able to study the performance of PRT networks covering entire metropolitan areas, tested for a PRT system covering the city of Gothenburg and for some other cities. Raytheon Company , in its work with the Chicago RTA (Rosemont), has developed a in which any network can be simulated, and many have been. Such simulations, not the kind of analysis Vuchic describes, are the way to analyze PRT networks, and working with them has led to increased enthusiasm about the potential of PRT.
1) Because of the short trip time with nonstop travel in a PRT system, "immediate vicinity" as implied in Vuchic's first point is within two to three miles.
2) In all responsible PRT cost calculations I am aware of, the requisite number of empty vehicles needed is always included.
3) While the French Aramis PRT program was terminated, it showed long ago that platooning, or batch loading, would provide station capacities that match rapid rail station throughputs. (A given throughput can be obtained as well by a flow of small vehicles at short headways as with a flow of large vehicles at necessarily long headways.) A good available description of station operations can be found in the book Fundamentals of Personal Rapid Transit, published by Jack Irving and others from The Aerospace Corporation in Los Angeles.
4) Based on my analysis and station and network simulations, and comparing my results with others, an average flow of 15 PRT vehicles per minute requires an 8-berth station.
5) In an N-berth PRT station, the off-line guideway must accommodate at least N + 3 vehicles stopped and waiting to enter the station if the frequency of wave-off is to be less than about one in 1000. By sacrificing a very small amount (about 0.1 sec) of on-line headway, the length of the off-line guideway can be shortened significantly. The needed lengths of off-line guideway are taken into account in our cost calculations.
Since Vuchic has spent his career studying and promoting conventional rail transit systems, particularly streetcar systems, albeit under the new name "light rail," and has tried since the early 1970s through various publications to "derail" PRT, he has a strong interest in trying to insure that PRT will never happen. To understand this, the reader of this piece should read the section on PRT in his textbook entitled Urban Public Transportation (1981). There he implies that a PRT system is one that schedules vehicles with a certain frequency from every station to every other station, and thereby concludes that PRT would not work. I agree that if PRT were arranged that way that it wouldn't work, but no serious PRT investigator I know of would dream of proposing such a scheme. In Vuchic's response, he repeats that error in talking about scheduling PRT vehicles.
PRT operates in response to demand, not to a schedule of any kind. All of the questions Vukan mentions in his response, and many more, have been considered in the design of PRT operating procedures, and the more we work with them the simpler we find they can be. Certainly, there are ways to design a PRT system so that it won't work, but the objective of the serious PRT investigator has been to consider every "what if" in order to design a system that will work, and work far better than conventional modes of transit. The result, we have found, is that in typical situations in which we have compared PRT with LRT, PRT can provide a considerably higher level of service than LRT for the range of about one third to one tenth the total cost per passenger-mile of an LRT system.
In his piece in Urban Transport International Vuchic states that "spacious vehicles" are needed to make transit efficient. What kind of efficiency? If he were to mean high daily average load factor, i.e., high utilization, LRT is not the system to select. Federal data from the Section 15 data report shows daily average load factors in the range of ten to twenty percent. If he were to mean energy efficiency, LRT would be okay if the vehicles were always full, but with the load factors that can be practically attained, the energy efficiency of LRT is equivalent to an automobile system in which the vehicles achieve around 10 miles per gallon. If he were to mean safety, in the surface-level streetcars promoted as a way to reduce costs, the number of accidents per vehicle-mile is about 60 times the figure for exclusive guideway systems, for which the costs are now upwards of $100M per mile. If he were to mean land efficiency, streetcars are better for a given flow of people per unit of time than automobiles, but they still take about a 38-foot strip of valuable surface street, now clogged with automobiles, whereas an optimally designed PRT system requires a tiny fraction of that land. If he were to mean total cost per passenger-mile, that number, which conventional rail people consistently avoid mentioning, for systems installed in the United States over the past two decades is outrageously high, about 5 to 10 times that of bus systems.
The fundamental reason is that transit systems using "spacious vehicles" are inherently highly inefficient for a variety of reasons, mainly high cost and low daily average utilization. I described this in detail in my paper "Optimization of Transit-System Characteristics," which appeared in the Journal of Advanced Transportation in 1984 (18:1). As to passenger carrying capability, there is no streetcar system in the United States reported in the Federal Transit Administration Section 15 report that actually carries - in the same corridor and with the same length - more people per hour than could be carried by an optimally designed PRT system.
In his fourth point, Vuchic argues that the larger vehicle (20 passenger) Morgantown system would be more efficient than PRT. This question was studied in great detail by computer simulation in the large study of transit alternatives that took place in the Colorado RTD in 1974-1975. The major results are reported in a paper by Johnson, Walter and Wild in Personal Rapid Transit III (University of Minnesota, 1976), pp 269-282, which showed that the shared-ride versions of AGT in the kinds of networks that were of interest in Denver, simply don't work.
The reason is that as a network grows the number of possible routes increases roughly as the square of the number of loops. Thus a group-riding vehicle coming into an off-line station will in general have people on board headed to stations along one of many routes. So, a person waiting at a station can't board any vehicle but must wait for one assigned to the route the destination is on. To maximize throughput, it is not practical to match stopping berths and routes so a person, perhaps your grandmother, must stand near the center of the platform ready to move quickly, with little time to spare, to a specific berth. In the Denver study, this operation was simulated in great detail, and the conclusion was that it is impractical. The kind of system that does work in a network, they clearly showed, is a true private-party PRT system, where one can wait to board at any berth and board the next vehicle coming along.
The Morgantown system works in its current setting because it has only six stations, but a PRT system operating even at two seconds headway would have provided better service at lower cost mainly because the guideway would be much lighter and therefore much less expensive, as we have found repeatedly by consultation with steel fabricators, one of which fabricated a section to obtain a better feeling for configuration and cost.
Today, we live in an era of great change, and, because of increasing pressure on resources, there will be even greater change in the decades ahead. Sooner or later the rail transit industry will find that its interest lies in becoming involved in the changes, however great they may be, rather than in fighting them - in this particular case by fighting new transit systems while trying to convince policy makers to install systems that were effective a century ago when the competition was a horse cart on a mud road rather than an automobile on a freeway.
Return to the PRT Debate page.
Last modified: December 24, 1996