Comments on Roxanne Warren's book "The Urban Oasis."

by J. E. Anderson

written January 7, 1998


Note: This review is focussed on an Appendix in this book, pp 165-180.

I recently obtained a copy of The Urban Oasis and thought it worthwhile to comment on the passages that mention Personal Rapid Transit (PRT), particularly on the comment on p. 140 that waiting time will increase with PRT as compared to GRT, and on pages 169-177 where PRT is discussed in more detail. It is interesting that, notwithstanding being a member of the ATRA board, she makes no reference to the 1988 ATRA study of PRT. Also there is no reference to any of the papers in the Proceedings of the 1971, 1973 and 1975 International Conferences on PRT (PRT, PRT II, and PRT III). Based on the references listed, her sources of information on automated transit are almost entirely Tom McGean and Larry Fabian's newsletter, which simply reports on what is going on in AGT. Criticizing this technology while ignoring the bulk of relevant literature is not good scholarship. Roxanne Warren refers to a paper by Ingmar Andreasson, but fails to mention that the problems she discusses are resolved if a minimum operating headway of one half second is used, which many investigators have shown to be practical if done right.

One remarkable thing to me about Warren's words about PRT is that she views it negatively for the very reasons many of us have viewed it positively, i. e., my colleagues and I have seen PRT as advantageous because, designed properly, it minimized the use of land, materials, energy, and cost and at the same time provides an unusually high level of service. She sees it as using too much land, material, energy and money.

PRT has many variable characteristics. Pick too many of them wrong and you have an inefficient system that is not cost effective. Many of these parameters have been picked wrong in the design of some PRT systems, and in this lies a great deal of confusion. The challenge is to select each characteristic based on thorough and open analysis of all reasonable possibilities. This is design optimization, what every engineer should do when faced with a systems problem. When done you have the strong pro-PRT recommendation of reports such as the 1988 ATRA PRT report. It is significant to note that in the SeaTac transit-alternatives study performed about five years ago of bus, Group Rapid Transit (GRT) and PRT, PRT won the unanimous vote of a 17-person steering committee even though the consultant quadrupled the cost estimates for PRT to compensate for his uncertainty. Also, a recent Swedish study of transit alternatives by Göran Tegnér for the city of Umeĺ (Infrastructure, Vol.2, No.3, Spring 1997) showed a benefit/cost ratio of 1.4 for a PRT system while only 0.2 for a large-vehicle monorail AGT system of the type Warren advocates.

Consider the following points raised in Warren's book:

Waiting time

On her page 140, Warren dismisses PRT based on the belief that the waiting time will be longer that on GRT. She references Tom McGean's 1976 book, where the conclusion is not based on comprehensive simulations of whole systems. I refer here to nine simulation studies that address wait time in PRT. Of these, the most detailed comparison of waiting time for different systems is given by an IBM Corporation paper by Martin Ross and Alan Melgaard, PRT III, pp. 369-376. Their results are as follows:

PRT Service Average Wait Time, seconds
  Demand Responsive, Single Party 42
  Demand Responsive, Group 212
  Demand Activated 410
  Scheduled 354
GRT Service  
  Demand Resposive, Group 425
  Demand Activated 565
  Scheduled 402

Other studies that treated true PRT or systems close to it gave the following results. To fully appreciate them it is necessary to read the original works.

Aerospace Corporation, PRT III, pp 345-368 80% of waits, 20-50 seconds

95% of waits, 1-1.5 minutes

99% of waits, <= 2 minutes

K. Thangavelu, DeLew Cather, PRT III, pp 329-344 No real PRT was considered, but for a 3-sec. headway system the waits were less than 2 minutes
Alain Kornhauser, Steven Strong and Paul Mottola, Princeton Univ., PRT III, pp 377-384 Figure 3 shows maximum wait time dropping as vehicle occupancy drops with about 1 minute with an average occupancy of 1.5 persons
Cabintaxi/Cabinlift Final Report, U.S. DOT and SNV, Germany, UMTA-MA-06-0067-77-02 On the bottom of p 4-189 it is stated that with the best selection of parameters, the waits are all less than 3 minutes, with 80-90% less than 1 minute.
Neil Sher and Paul Anderson, Honeywell, PRT II, pp 401-416 With a 2 second headway and 8-passenger vehicles, they show average wait times of 66-75 sec.
Roesler, Williams, FJord and Waddell, APL, Johns Hopkins Univ., PRT II, pp 425-437 With unrestricted routing, they show wait times of 40-85 seconds
Harold York, Bell Labs, PRT II, pp 439-447 40-85 sec. average  wait, maximum about 3 minutes

Beginning in 1986 I have been developing an accurate PRT network simulation that can be used for any size or configuration and is now sufficient to run an operating system. I have applied it to many PRT systems. Hundreds of runs were made during our studies for the Chicago Area RTA that typically showed average wait times during peak periods under one minute and maximum waits less than three minutes, which are well under those generally experienced in conventional transit systems. During off-peak periods there is no wait for a PRT vehicle, whereas with conventional systems wait time must necessarily lengthen. Wait time depends mainly on minimum headway and the number of vehicles in operation.

Complexity and Sophistication

A lot of both theory and practice needs to be absorbed to design a PRT system optimally. The Aerospace Corporation publications on PRT are still an excellent source, and I cover most of the basics in my updated edition of Transit Systems Theory, which the reader may obtain by contacting me. When one first hears of PRT it may indeed sound complex. But so would a Boeing 767 to an engineer who only knows DC3's. Once you begin to understand PRT you find that it is not as complex as say a modern traffic control system, indeed with today's technical knowledge it is quite straightforward. So dear reader, don't be stopped by PRT complexity. It is well within today's technological capability. Resistance to PRT is more a result of the strong institutionalization of conventional transit than to complexity.


Why is this a negative? If the switches were in the track it would be, but simple, in-vehicle switching systems have made switching as easy as steering an automobile onto different roads.

System Size

In the mythology of AGT it is repeated over and over again that PRT requires large networks. My colleagues and I have laid out and studied dozens of PRT systems large and small, and find again and again that on the basis of cost per passenger-mile they beat the larger vehicle systems. A major reason is that with the lightest practical vehicles, the guideway is much smaller and less expensive. Large vehicles require large, expensive guideways that consume substantially more land than a PRT guideway, thus increasing the land cost associated with the system.


They don't have to be multi-level. We use Y-interchanges almost everywhere because they minimize visual impact. Duncan MacKinnon of the Urban Mass Transportation Administration (UMTA) first stressed that to me in about 1973 while we, at the University of Minnesota, were engaged in a visual impact study of PRT funded by UMTA. [UMTA has now been replaced by the Federal Transit Administration (FTA)].


Where is it written that PRT speeds must be higher than other AGT speeds? Since a PRT trip is nonstop, the line speed needed to achieve a given average speed is lower. I have found many urban applications of PRT for which a line speed of 25 mph is adequate. The ability of a guideway designed for low speed to handle higher speeds depends on minimum headway and random vs. equal spacing of vehicles. Equal spacing gives the effect of a troop of soldiers marching in step across a bridge. By breaking step the dynamic effects are markedly damped. This is one of several reasons we use asynchronous control. In 1975, Snyder, Wormley and Richardson of the M. I. T. Mechanical Engineering Department completed a study of AGT guideway dynamics for UMTA (Report No. UMTA MA-11-0023-75-1). They showed in detail that the ratio of maximum dynamic deflection to maximum static deflection in a guideway varies from very large for large-vehicle, long-headway systems to only about 1.2 for PRT systems running at quarter-span spacings. Based on these results, I found that the deflection of a PRT guideway with fully-loaded vehicles nose-to-tail standing still is larger and determines the design. I recently extended these results in a study (available on request) that shows why PRT systems thus designed can operate without modification at considerably higher speed than mentioned above. I have studied the question: What line speed in a given application minimizes the total cost per passenger-mile? I have found that such an optimum speed is lower than intuition generally suggests, and thus advise that speed not be simply picked but considered a parameter to be determine during a comprehensive study.


I address this topic in a paper being considered by the ATRA Board, in which it is shown why PRT matches capacity to demand. [see the capacity article by Anderson]

Number and Parking of Empty Vehicles

If vehicles were to move all in one direction in the morning and in the other direction in the evening, the number of empty vehicles needed is 50% of the total fleet. If the demand is completely uniform, no empty vehicles are needed. In many applications, simulations have shown that generally about one-third of the vehicles will be empty. We take all of them into account in computing the total cost per passenger-mile. Every transit system has a certain amount of deadheading. A problem with scheduled systems is that to provide reasonable service in off-peak periods vehicles must be run with very few or no passengers. In an AGT system at Las Colinas of the type Warren advocates, ridership was so poor that service was discontinued [It has recently be made operational again]. The daily average occupancy generally runs between 10% and 20% so too much empty weight is carried around per passenger. PRT is demand-responsive so in off-peak periods vehicles need move only if someone wants to be served. This substantially increases the daily average load factor of moving vehicles, and thus improves energy efficiency and reduces operating cost quite substantially. UMTA personnel realized this at least by the late 1970's.

I have studied the volume of space required to store PRT vehicles in off-peak periods in several applications and find that, with the size vehicle we have found practical, the storage volume required per mile is equivalent to the parking-structure volume that will store about three to four automobiles. The volume requirement is modest because it is not necessary to remove a particular vehicle at any time.

Land Use

PRT uses land only for posts and stations, which typically take less than two percent of the land. An automobile system requires one third of the land in residential areas and often well over half the land in central business districts. For this reason alone arguments that PRT duplicates the auto system are specious. PRT is a transit system and can either supplement or replace sections of an obsolete transit system. By using PRT as a circulator in and near the center city, it is possible to achieve a level of environmental quality, safety and transit service far superior to that possible with conventional transit. Advocacy of surface-level light rail ignores the potentiality and experience of many accidents associated with mixing 80,000-lb streetcars with pedestrians and automobiles.

Energy Use

Warren quotes an Office of Technology Assessment (OTA) report, in which I participated but did not endorse, saying "Without a reduction of per passenger weight it will not be possible to reduce energy costs of operation." While weight must be minimized, weight is only one parameter. The nonstop trip of PRT means that the kinetic energy need be put in and removed only once. Transit systems of the same weight per passenger that require intermediate stops use much more energy. Weight is of course very important for energy reduction and many PRT developers are currently realizing that. Road resistance, air drag, construction energy and energy for heating, ventilating and air conditioning are also important. All must be minimized to minimize energy consumption. One of the many reasons I like PRT is that by careful design it is practical to reduce energy per passenger-mile considerably over conventional transit. The frequent assumption of energy efficiency associated with light rail is based on the unrealistic hope of high load factors not borne out in practice.

Vehicle Size

On her p. 176 Warren quotes the above-mentioned OTA report that "there appeared to be no reason why PRT vehicles would be any smaller or lighter . . . than small cars or vans." Actually there are several reasons: 1) a small car or van has to be designed to go over chuck holes and curbs and to resist side collisions and roll-overs. PRT has none of these problems, therefore the cabin can be considerably lighter. 2) A properly designed PRT vehicle will have a lighter, smaller-volume propulsion and braking system than an auto. 3) A PRT vehicle does not have to be designed for an occasional out-of-town trip or to hold as many people as the maximum generally desired in a family automobile. All that is needed is one bench seat about 50 inches wide, like the back seat of a taxi, so the vehicle can be much shorter than an automobile. The daily-average load factor and with it the economy are improved by reducing the capacity of the vehicle. Ride sharing with one or two friends can be encouraged by charging a fare per vehicle. It is surprising to many people to learn that the cost per passenger-mile in a transit system is minimized by going to the smallest and lightest vehicles.

Number of Berths per Station

I cover this in detail in my course "Transit Systems Analysis and Design." Some stations may be large but the average is smaller than one would guess.

Length of Off-Line Guideway

This is an area in which optimization is particularly fruitful. I have found, for example, that by losing only 0.1 second of on-line headway at 25 mph the off-line station lengths reduce by 120 feet. At higher speeds the savings is greater. See my updated Transit Systems Theory. In a typical urban application, I have found that by design optimization the off-line guideway length is only about 12% of the total guideway length without off-lines.

PRT an Anachronism?

For a growing number of people, for reasons well summarized in the Advanced Transit Association's PRT Report, PRT is the wave of the future. That is not to say there is no place for shuttle-loop systems. Some applications are suited to such systems. Nor would we join those who would say that PRT can replace the automobile. It is a transit system. Nor would we say that PRT would eliminate the needs for buses or conventional rail. It can supplement them. PRT deserves to be fully developed so that it can be in the bag of alternatives offered by transportation consultants, and compared fairly with other options. As a society, we understand that free competition is the best way to improve our economy and our quality of life. That philosophy should apply to transit too, yet city planners are told again and again not to even look at anything not fully proven, while conventional transit is losing repeatedly to the automobile. For a full account see The Urban Transit Crisis in Europe and North America by John Pucher and Christian Lefevre, MacMillan Press, Ltd., London, 1996.


Last modified: November 08, 1999