A Review of the Paper "PRT/Déjà Vu"
by Maurice A. Sulkin
Daniel, Mann, Johnson and Mendenhall, Los Angeles, CA
Presented at the Transportation Research Board Annual Meeting, January 10-14, 1999.
Critique by J. Edward Anderson
Ph.D., P. E. President & CEO, Taxi 2000 Corporation
May 5, 1999 (posted January 9, 2000)
The paper by M.A. Sulkin has been published in the Journal of the Transportation Research Board, # 1677, pp 58-63, October, 1999
I finally got a break from my work on PRT to study the subject paper. I found that the analysis given is so simplistic that without looking at its date a person versed in the analysis of PRT systems could assume that it was written in the 1960's at a time when the theory of PRT systems was in its infancy. I cannot believe that it was written with any attempt to seriously understand PRT but rather to wish the concept away so that it won't interfere with hopes of building 19th century rail systems everywheresystems that have been shown again and again to be economic disasters, losing ridership to the automobile. All of the conclusions presented are wrong.
In the introduction to the paper, the writer states "passenger stations (of PRT systems) are depicted as little dots or squares alongside a single line of guideway." This statement shows that he is not familiar with the substantial body of information presented on PRT in many papers and reports over the past 30 years. Much of this information was collected in the proceedings of three international conferences on PRT sponsored by the University of Minnesota in 1971, 1973 and 1975; in an ATRA conference held in Indianapolis in 1978; in the book Fundamentals of PRT by The Aerospace Corporation PRT team; in reports of work on the German, Japanese, French, and British PRT development programs; in my 1978 textbook Transit Systems Theory; in numerous UMTA-sponsored reports; in the papers published from the 1996 Minneapolis Conference on PRT and Other Emerging Transportation Systems; in the work supported by the Northeastern Illinois Regional Transportation Authority; in work sponsored during the past five years by the Swedish Transportation Research Board; in PRT work during the period from 1990 to 1997 in Korea, etc., etc. My own published work on the theory of PRT has been updated through a series of papers developed during the past 20 years, many of which first appeared in the Journal of Advanced Transportation. The contents of my updated Transit Systems Theory (referenced below as TST) may be reviewed at the TAXI 2000 website. Some of the papers may be downloaded there and instructions to purchase are given.
Sulkin makes a point of ridiculing the excellent work of The Aerospace Corporation, which was mainly developed between 1968 and 1974. In preparing to write a paper several years ago, I reread their 1978 book Fundamentals of PRT and found that they were correct on a remarkable number of points even though because of technological advances some things would be done differently now. The Aerospace Corporation, containing perhaps the finest group of system engineers in the world at the time and under the brilliant leadership of Dr. Jack Irving, covered all of the points raised in Sulkin's paper in all necessary detail to prove that their High-Capacity PRT system was worthy of full development. We learned over a quarter of a century ago, however, that even talking about region-wide PRT systems so frightened managers of conventional-rail companies that they saw their dreams of dominating the market with so-called "light" and "heavy" rail systems vanishing out the window.
The substantial opposition to a new solution in the transit field and the reason it has taken so long was explained as well I believe as it can be 500 years ago by Machiavelli in his masterpiece "The Prince," where he said:
"It must be realized that there is nothing more difficult to plan, more uncertain of success, or more dangerous to manage than the establishment of a new order of government [or a new system]; for he who introduces it makes enemies of all those who derived advantage from the old order and finds but lukewarm defenders among those who stand to gain from the new one. Such a lukewarm attitude grows partly out of fear of the adversaries, who have the law on their side, and partly from the incredulity of men in general, who actually have no faith in new things until they have been proved by experience. Hence it happens that whenever those in the enemy camp have a chance to attack, they do so with partisan fervor, while the others defend themselves rather passively, so that both they and the prince are endangered."
This is good reason to introduce PRT in small increments and as far as possible in cooperation with the old order, without attacking it if that is possible. Neither Sulkin nor the PRT community can know for sure how far High-Capacity PRT can be expanded until it is tried. A PRT system must in any case start in small increments and if it is to expand it must do so gradually. There is a great deal of evidence that a truly cost-effective PRT system can serve many small applications well, so there is no valid technical reason to hold the development of such systems back. Mr. Sulkin agrees that PRT can start in that way so let us see from experience how far it can expand. Here are my specific comments, taken in the order they appear in the paper:
Page 1: There is no question that The Aerospace Corporation laid out a very elaborate PRT system for Los Angeles. They did so with the enthusiasm of pioneers in the field and to understand the problems of such an elaborate system, and they did so in considerable detail, analyzing all of the necessary feasibility questions. PRT is a transit system. It is not a door-to-door system. The stations, however, can be placed much closer together than the stations of a heavy or "light" rail system. In these conventional rail systems, since the stations are on line, adding stations to improve accessibility lowers the line speed, so the stations must be placed farther apart than would be desirable to provide good service. The result is that people do not ride the system either because the average speed is too low or because the system is accessible at too few points. These difficulties may account for current attempts to urge development of high-rise housing near the stations, i.e., trying to get people to live where they do not wish to live in order to make 19th century rail systems, designed originally to improve transport over horsecarts on mud roads, competitive with the auto. Criticizing PRT for not being able to place stations closer than about 600 feet apart when the stations of conventional rail must be placed usually at least a mile apart is simply laughable. Using the most extreme application ever considered to discredit an evolving new mode discredits only the writer.
Page 2: The issue of the requirements of system and station capacity in PRT systems is addressed in the author's ATRA paper "PRT: Matching Capacity to Demand." The reader may download this paper from the TAXI 2000 website as well as two simulations, one of a PRT station and one of a network containing 25 passenger stations, four storage stations, and 560 vehicles. Beginning with a theoretical analysis of station throughput (available from TST) and using first the station simulation and then the network simulation, we have found that the practical throughput of PRT stations with a single off-line guideway having from one to 14 berths satisfies a wide range of applications and keeps the fraction of wave-offs to less than one per 1000 trips. Indeed, neither The Aerospace Corporation nor we found it necessary to use anything but single-channel stations. The theoretical analysis is useful to clarify the factors that influence station throughput; the station simulation permits quantification of maximum station throughput with vehicle and passenger flows randomized, but is mainly useful now to visualize the operation of a PRT station; and the system simulation, which also uses randomized inputs, is needed to determine realistically the number of station berths required to meet a given demand. Our system simulation is described in a paper that may be downloaded from the TAXI 2000 website.
Two-channel stations such as Sulkin references could be used, but they would be more expensive and would negate the advantage of walking directly from the second or third floor of a building into a station. It makes no sense to use such stations to represent typical PRT stations. Sulkin refers to estimating PRT vehicle load factors. This implies what has been called by UMTA "Group Rapid Transit," such as used in Morgantown where he has had experience. Work mainly by Johnson, Walter and Wilde showed that group-transit service leads to unmanageable station operations from the passenger's point of view so that they are not practical in anything but the smallest systems. The PRT passenger load is the number of people riding together by choice and may be compared to passenger occupancy in automobiles. A 1990 study done by the Twin Cities Metropolitan Council showed average rush-period occupancy of only 1.08 people per vehicle and average daily occupancy of 1.2 people per vehicle. Charging a fare per vehicle rather than per person may raise PRT occupancy. Sulkin refers twice on this page to "main-line merging slots."
This implies synchronous operation as used in Morgantown, in which a vehicle ready to leave a station waits until there is an open position at every merge point all the way to the destination. Even at the first PRT Conference in 1971, speakers showed that such operation is not practical if there are more than a few merge points on the way to the destination. (See paper in TST entitled "Synchronous or Clear-Path Control in PRT Systems" for a calculation of the wait time with such a strategy.) The practical operating strategies used have been quasisynchronous and asynchronous. Our work has led us to prefer an asynchronous point-following strategy, in which case the problem Sulkin references does not exist.
Pages 3-4. Sulkin here mainly discusses the work of another group. In true PRT, the strategy of unloading then moving forward to load at another berth is generally unnecessary and results in unnecessarily large stations. It is generally unnecessary because, with the few people involved at any berth, a bit of human factors study shows that those planning to board can easily stand aside while passengers debark, and it is usually the case that at any station there are more people loading than unloading or vice versa. Sulkin's remarks are understandably biased by his Morgantown experience, in which 20-passenger vehicles were used. Sulkin conjectures on this page that "further reducing headway would probably yield little additional capacity" is not borne out by detailed simulation work. In his section "Capacity Data" at one point he implies that station capacity is "optimized" (I presume he means maximized) by using separate unloading and loading berths, and at the end he states that station capacity could be increased by an "Unload and Load mode." One can't have it both ways.
Page 5. The work in the section "Station Configuration" is rarely useful in serious PRT work.
Pages 6-7. System Availability. The definition given for system availability is useful only in a system in which one failure would shut down the entire system, i.e., in very small automated shuttle-loop systems. By analogy such a definition would be like assuming that in an automobile system one failure such as a flat tire would shut down all the cars in the town. It is true that a car with a flat tire can be quickly driven out of the way of other cars. It is not as easy to remove a failed car in a PRT system, but, with a pushing strategy such as was devised over a quarter of a century ago, the number of PRT cars involved in a failure is generally much smaller than the total number of cars in the system. (See the TST paper "Redundancy, Failure Modes and Effects, and Reliability Allocation.") Thus a new way of calculating system availability has been needed to determine the on-time performance of PRT systems. Indeed, workers in conventional transit have looked for a way to take into account the delay time of passengers in a measure of on-time performance. I have developed such a measure. It is refined in my paper "Dependability as a Measure of On-Time Performance of PRT Systems" (Journal of Advanced Transportation, 26:3, pp. 201-212.)
This method was accepted and used in our Phase I study of PRT sponsored by the Northeastern Illinois Regional Transportation Authority. In my paper undependability is defined as the ratio of person-hours of delay due to failures to person-hours of operation. Dependability is one minus undependability. This quantity can't be measured in large-vehicle transit systems without asking each person where he or she is goingnot a good idea. In PRT, the system knows where each vehicle is going and it knows the expected trip time, so undependability can be measured continuously and automatically. Now a few words about Mr. Sulkin's development of service availability, As. In the first sentence under his equation for As he has the definition of MTBFs up-side-down. MTBFs is a time period such as one year divided by the number of failures in that period. In the second sentence he defines Mean Time to Restore (MTTRs) as a ratio of times. MTTRs is the mean down time per failure, i.e. the total downtime caused by failures divided by the number of failures in the same period, not by the number of operating hours. The formula given at the bottom of his page 6 is wrong. The mean time between failures is simply the reciprocal of the number of failures per unit of time. His formula has units of time squared per failure rather than time, as it should be.
Using his simplistic equation Mr. Sulkin comes to the erroneous conclusion that system availability decreases as the number of vehicles increases. In the above-mentioned paper I show in detail why this is not true for either a system-applicable definition of system availability or dependability. It is hardly truer than to say that the system availability in an automobile system in a large city is lower than in a small city. The number of person-hours of failure due to vehicle failures is proportional to the number of vehicles and the number of person-hours of operation is proportional to the number of vehicles, so the number of vehicles drops out of the equation. Mr. Sulkin's simplistic equation is simply not applicable to a PRT system. Safe High-Capacity PRT does require high reliability components. I have analyzed this problem in detail in two papers in TST: "Redundancy, Failure Modes and Effects, and Reliability Allocation," and "Effect of Redundancy on Failure Frequency in PRT Systems." Technology available today can do the job, and the computer systems required get smaller and cheaper every year.
Pages 7-12. Station Interval Requirements. Mr. Sulkin insists that PRT requires very close station spacing and then argues that PRT isn't practical because his analysis shows that such close station spacing isn't possible. Station spacing in PRT is much more flexible than in conventional rail because adding a station does not reduce the average speed of any of the trips. I have shown in my paper ""Optimization of Transit-System Characteristics" (Journal of Advanced Transportation, 18:1(1984):77-111, Appendix B) that the economics of adding stations in PRT is very good; however, there is no need to place them closer than dictated by requirements of the off-line guideway length and the required distance between branch points. The correct way to arrive at such lengths is to use the exact equations for acceleration and deceleration and for the transition lengths.
All of this is developed in detail in papers in TST. A few of the points made on these pages are worthy of comment: The correct way to calculate acceleration and deceleration lengths is to ramp up to or down from the maximum comfort acceleration at the maximum comfort jerk, rather than to use an approximate exponential velocity profile, which must be chopped at an arbitrary point to keep the formula from predicting an infinite distance to reach line speed. The exponential formula implies infinite jerk at the start of acceleration and much too small an acceleration near line speed. Rather, at a predetermined point, such as at half the line speed, to limit acceleration power, the acceleration can be decreased at a constant negative jerk and then close to line speed at maximum jerk. This profile is derived in TST in the paper "Kinematics of PRT Vehicles." It is shown there that acceleration power can be reduced by about 45% over the case where comfort acceleration is used up to the point that maximum negative jerk must be applied. In this way, the extra length for acceleration is increased by only about 15%, not by the huge factor Sulkin derives. We take the distance between one merge or diverge of two guideways and the next merge or diverge to be the speed multiplied by the switch time plus the emergency stopping distance if the switch position cannot be verified, including suitable tolerances. Sulkin takes the line speed in a downtown application to be 13.5 m/s or a bit more than 30 mph. He gives no justification for this speed. Our observations shows that 25 mph and possibly even lower will be adequate in a downtown where the auto and bus traffic generally averages not much more than 10 or 12 mph. Kinetic energy, air drag, curve radius, and stopping distance vary as the square of the line speed, and (30/25)2 = 1.44 which is a significant ratio and will affect his calculations of off-line guideway length. Line speed is a parameter that should be determined in a planning process. Ideally the line speed should be selected to minimize the total cost per passenger-mile. While this is a complex calculation, estimates I have made indicate that the optimum line speed is lower than one might guess. The advantage of small station spacing is that it cuts walk distance. This saves much more time than going a little faster on line. For example, walking a quarter of a mile at the average speed of 2 mph takes 7.5 minutes, whereas the time increase in going 25 mph rather than 30 mph is only 40 seconds per mile. Sulkin states that with short headways it is not possible to have any on-line deceleration or acceleration. I found in my paper "Plane Curved Guideways" in TST that by sacrificing only one tenth of a second on-line headway one can save about 110 ft of off-line guideway length at 25 mph line speed and 130 ft at 30 mph. The conclusion is that it is wasteful not to do some on-line deceleration into and acceleration out of off-line stations, and further that the off-line guideway becomes short enough so that one could fit one station per block if it were found to reduce the cost per passenger-mile by so doing.
Page 12-13. Conclusions. Sulkin concludes that only small PRT systems are practical. He gives six bases for his conclusions, all of which are wrong. Here are my summary comments on each of his six points: Based on the ATRA paper "PRT: Balancing Capacity to Demand" there are many circumstances in many cities in which PRT station capacity is more than adequate to match demand. For example, in a study of PRT for Downtown Indianapolis, we found that none of the stations required more than three berths whereas we can use up to about 14 berths. In our simulation for Downtown Cincinnati, we found that four 14-berth stations could match the demand into the Cincinnati Reds Ballpark. PRT stations are generally not large and complex. I have found no practical case in the cities we have studied where we would need multiple-channel stations. They could be used but are an exception, not the rule. We design for no more than one wave-off in 1000 trips where the waved-off vehicle has a reasonably short return path. If the wave-off path happens to be long, we design for no more than one wave-off in 10,000 trips. The acceptable level in any application must be determined as a policy input based on understanding of the extra trip time involved. Perhaps the system could give a prize to a person waved off. It would be like hitting the jackpot! Even when we permit fractional-second on-line headways, we find that we can save over 100 ft of off-line guideway by sacrificing only one-tenth of a second of on-line headway. The conclusion given that system availability decreases with fleet size is based on a simplistic equation that assumes that every failure stops the entire system, which applies only to systems consisting of only one loop. A correct analysis shows that system availability or dependability in network PRT systems do not reduce as the fleet size increases. The conclusion that one could not fit one station per block in a downtown application, if that were desired, is based on inadequate analysis of the factors that go into calculating the off-line guideway length and on too high a line speed.
The only thing light about "light" rail is its name. Changing its name from "streetcar" was a masterstroke for the uninformed. The cars are heavier than the cars of typical heavy rail systems and the rail can be lighter than the rails of heavy rail only when the speed is much lower, however, in many current applications the speed between stations is designed to reach the same range as used by heavy rail systems and hence the rail must be just as heavy. Streets must be reinforced to support these "light" 80,000-lb vehicles. "Analysis and Simulation of Automated Vehicle Stations," Personal Rapid Transit III, University of Minnesota, 1976, pp. 269-281.
Last modified: January 09, 2000