RUF Network Design Experiment for the Seattle Metropolitan Region

by J. B. Schneider

July, 1997

This paper briefly describes some of the results obtained from a design exercise conducted as part of a freshman design class at the University of Washington during the Spring Quarter of 1995. This was a "design to specification" type problem that was given to three four-person groups to be accomplished during a two-week period at the end of the quarter. The students were asked to work as "design teams", called Alpha, Beta and Gamma.

Scope and Objectives

The purpose of this exercise was to examine the potential utility of the RUF dual-mode urban transportation concept (presently being developed in Denmark) in the context of the existing conditions present in the Central Puget Sound Region (CPSR) located in the western part of the State of Washington. This region contains about 3 million people and includes the cities of Seattle, Everett, Tacoma, Bellevue and Bremerton as well as many smaller cities. At present, the conventional roadway network in this region includes approximately 6,750 miles of high level roadways (freeways and arterials) as well as a much larger mileage of local streets. This roadway system is often very congested during peak periods and weekends with average speeds being around 30 mph or less on most days of the year. Significant levels of air pollution are also present in this region several days of the year.

The problem assigned involved the design of a RUF network (i.e. locating stations and connecting them with RUF-rail guideway segments) that could serve the entire Central Puget Sound Region (i.e. the urbanized parts of Kitsap, Pierce, King and Snohomish counties). RUF is a Danish dual-mode transportation concept, currently under development.

The design was to be evaluated with respect to several performance measures, including cost, coverage and travel time indicators. A computer program, called EMSIMS was provided to assist the students code their designs and compute its expected performance for subsequent comparative evaluation. The essence of this design problem involves resolving the trade-off between the number of stations provided, the estimated cost and the estimated performance of the system relative to travel via conventional auto over the existing roadway network. More RUF stations and more guideway will usually provide more performance. Spatially extensive systems also cost more but they are needed to provide good service to the highly dispersed and low density Seattle-Tacoma metropolitan area.

The data needed for this problem were provided. They consisted mainly of a 1994 regional highway network that was obtained from the Puget Sound Regional Council . Figure 1 is a map of this network. This network data includes 3,398 nodes, 10,484 links and speeds for the PM peak, AM peak and Off-peak periods. Population data (by residence) were used to represent the demand for service. A more inclusive dataset would have also included the number of jobs in each subarea of the region but such data were unavailable when needed for this exercise. Eighty feasible and desirable candidate locations for RUF stations were also specified by the instructor. Figure 2 shows the locations of these 80 station candidates.

The students were asked to design a RUF network that would place at least 15% of the population of the region within a 5-minute drive of a RUF station and at least 50% within 10 minutes of a RUF station. In addition, the design must not place more than 20% of the population more than 15 minutes from a RUF station (i.e. 80% within 15 minutes). These performance measures were called "coverage objectives". An important objective of the design exercise was to satisfy these coverage objectives for the least possible cost. The cost of the RUF design was measured by the number of stations used and the length of the mainline and connecting guideway needed. It was assumed that the average cost of a RUF station would be $4 million and that the cost of a two-way mile of RUF guideway would be $7 million. RUF routes that cross a lake were assigned a cost of $21 million per mile. Thus, a design that contains 15 stations and 120 miles of RUF guideway (none crossing a lake) would cost $900 million. It was assumed that RUF vehicles could use the cross-sound ferries like conventional vehicles. No definite upper limit to the budget was specified but it was suggested that designs that were estimated to cost more than $1 billion are not likely to be very popular with the taxpayers of the region.

To solve this problem, the students first had to determine how many stations to include in their design. Since there were 80 candidate locations, the problem was to determine how many should be selected and where should they be located. The students used a automated search routine included in the EMSIMS software to find a near-optimal solution to this problem. Typically, an intuitive solution was developed on paper and then given to the computer to see if its search routine could find a better one. Initially, a large (e.g. 30 ) and small (e.g. 15) number of stations was used to establish upper and lower bounds for the problem. Then, additional searches were conducted to find solutions that had the most performance for the least number of stations. For this problem, an optimal solution is one that places the specified number of stations as close as possible to the people who would want to use them.

Once a suitable set of stations was identified, they had to be connected to each other and to the conventional street network. This was done intuitively and then the results were entered into the computer. Special attention was given to keeping the number of routes connecting at each station as minimal as possible.Then, some "delay penalties" were added to simulate the reduced speed of RUF vehicles as they pass through a station, switching from one RUF line to another or exiting. Once a database representing the design was completed, it was possible to compute a variety of performance measures and generate a set of maps that could be used to evaluate the expected utility of the design. If the design was found to be unsatisfactory, then some design changes were made and the modified design was reevaluated. This process was repeated until satisfactory results were obtained. When this point was reached, the design was said to be the "preferred design" and it was used as the basis to develop the team's design report.

In addition to cost and coverage performance measures, the overall quality of the designs was assessed by comparing RUF travel times with those provided without RUF (i.e. by conventional auto using the existing roadway system). This was done by calculating the minimum path travel times between 20 O-D (origin-destination) pairs twice, once for a network with RUF and again for a network without RUF. Finally, the total travel time required for a "grand tour" was calculated for the two (with RUF and without RUF) networks. This grand tour consisted of a trip from a starting node to 20 other nodes in the region - in a specified sequence - finally returning to the starting node.

Critical Assumptions

Since the RUF technology is not yet fully developed, it was necessary to make a number of assumptions about it for the purposes of this network design exercise. The speed of a RUF vehicle on the mainline RUF rail was assumed to be 70 mph (113 kph). As a RUF vehicle approaches a station, it must decelerate to 19 mph (30 kph) to pass through the station or to exit at the station. To represent this slowdown, a delay penalty of 30 seconds was assigned to the RUF trip time for every station that its O/D path passed through. A RUF vehicle would also experience some delay in entering or exiting the station to/from the conventional street system. This delay was represented in the design by assuming that each access/egress link to the street system would be 0.1 mile long with a speed of 10 mph. This means that the average time required to enter or leave a RUF station was assumed to be 36 seconds. The speeds used to represent auto travel are those that have been derived for 1994 by the Puget Sound Regional Council for the PM-peak time period (i.e. the slowest speeds). Understanding and remembering these assumptions is essential to making a correct interpretation the results obtained.

Results Obtained

The results obtained were extensive but are only briefly summarized here. Figures 3 and 4 and 5 show the RUF network designs from each of the three design teams. The small triangles on these maps represent RUF station locations. Two teams used 18 stations, the Beta team used only 13. The Beta team decided to meet the $1 billion budget goal. As a result, the performance of its design is somewhat less that the performance of the Alpha and Gamma teams, as their costs were significantly more than $1 billion. All three designs were able to exceed the coverage objectives by a significant amount. Note that the coverage provided by the Alpha design exceeds that provided by the Gamma design, even though its costs were about $200 million less.

The following table gives a summary of some of the results from all three design teams:

The average driving time to a RUF station was highly similar in all three designs, ranging from about 8 to 9 minutes with a similar standard deviation. The average service area populations varied directly with the number of stations provided, as expected. It is interesting to note that the Beta design had coverage statistics that are much higher than one would expect, given its small number of stations. The total miles of guideway is directly related to the cost and extensiveness of the network design, being highest for the Gamma design and lowest for the Beta design.

The results from the RUF-auto travel time comparisons were quite similar in that the RUF network provided a travel time reduction of between 22 and 31 % with an average reduction of 27% for both the 20 O/D pair and Grand Tour calculations. One must not extend the interpretation of these travel time reduction calculations very far. Their utility is quite limited since the sample of all possible trips in the region that was used is very, very tiny and was intuitively, not randomly, derived. The O/D matrix for this problem is 3,395 by 3,395 in size and defines 11,522,630 possible O/D pairs. So, even a 1 percent random sample would have to contain over 100,000 observations. Moreover, the trip length distribution of these observations should closely match the observed trip length distribution in the region. This is an analysis task that was far beyond the scope of this simple design exercise.

Clearly, short RUF trips would seldom use the RUF rail and so their travel times would be approximately the same as conventional auto. The longer the trip, the more likely RUF travel times are to be shorter than conventional auto. The O/D pairs included in this design exercise calculation tended to be mid-range in length (i.e. 25-50 minutes). A more comprehensive and detailed examination of this critical travel time comparison question is needed.

If one simply divides the total miles of RUF mainline guideway by the number of stations included in the design, it seems that the average station spacing is approximately 10 miles. If the average station spacing were to be reduced to five miles, the number of stations would probably be about twice that (i.e. about 30 or so) included in these designs. One would expect the coverage values to rise as the number of stations is increased, although at a declining rate of increase. The added cost of another 15 stations would not be great relative to guideway costs but the added delays in travel time that would arise as the number of stations is increased might be large enough to be of concern. Certainly, travel time is an important attribute of urban travel but it should not be the only one of concern. RUF travel would be highly reliable as compared to auto travel. It would also rate highly on being a low stress and comfortable ride, with opportunity to see interesting views of the city and to relax enroute in a variety of ways. It might be more costly or less costly that travel via auto; that is yet to be determined. It is the sum of all of these factors that will ultimately determine the utility of a RUF system to the residents of a metropolitan area.

For more details on the current status of RUF, visit the Danish RUF website

Written by J. Schneider , Professor Emeritus of Civil Engineering, Department of Civil Engineering, University of Washington, Seattle, WA 98195 on June 13, 1995 - updated in July of 1997. Comments and criticisms are welcome.

Return to the Innovative Transportation Technologies home page .

Last modified: December 09, 2000