Capital Costs and Ridership Estimates of Personal Rapid Transit
Supin L. Yoder, Sidney E. Weseman, and John DeLaurentiis
January, 2000
Abstract
Preparing capital cost estimate for new transit technologies requires refining and refitting traditional estimation tools. The personal rapid transit (PRT) system project in Rosemont, Illinois – a suburb of Chicago near O’Hare International Airport – required (1) the development of estimates of PRT capital costs, (2) establishment of a reasonable cost range for the PRT system, and (3) an order-of-magnitude evaluation of existing cost projections.
A method was developed to compare PRT system components (versus the entire system) with the components of existing automated-guideway transit (AGT) and automated people mover (APM) systems were examined – (1) guideways; (2) stations; (3) maintenance and control facilities; (4) power and utility systems; (5) vehicles; (6) command, control, and communications systems; and (7) engineering and project management.
Three analysis techniques were used – (1) statistically significant regression analysis; (2) measurement of a central tendency for "comparable" systems; and (3) statistics from all AGT systems. The results show that the combination of the three techniques worked well for component-level studies and show promise for use in other cost analyses involving new technologies or application of existing technologies on a scale outside the bounds of previous experience. In addition to the cost study, a PRT ridership forecasting approach and projections were evaluated, providing another key element of decision support for potential PRT deployment in Rosemont.
For additional details and illustrations, see the supplemental presentation consisting of 12 slides
The Northeastern Illinois Regional Transportation Authority (RTA) is the metropolitan Chicago area’s transit financing, planning, and oversight agency. The RTA transit system operates in a six-county area of 9580 km² (3,700 mi²) and nearly 8 million residents. Guided by the RTA Strategic Plan’s principles promoting experimental solutions to evolving transit markets and a statutory mandate to study and develop new public transportation technology, the RTA began its personal rapid transit (PRT) development program in May 1990. The project has three phases. Phase 1 included system design studies and Phase 2 involved system development and testing of the preferred PRT concept. Phase 3 will consist of a site demonstration and deployment in Rosemont, a suburb adjacent to O’Hare International Airport. Raytheon, in partnership with the RTA, is developing the PRT 2000 technology.
In late 1997, the PRT 2000 technology was in the final developmental stage of Phase 2 with the initiation of prototype testing. With the development phase ending, decision support for Phase 3 deployment of the PRT system at Rosemont became a concern. RTA planners were asked to provide an independent order-of-magnitude check on the capital costs projected by Raytheon (Rosemont DTUPG Update, Raytheon, 1997, unpublished data) and to evaluate ridership potential estimated by Wilbur Smith Associates (Estimated PRT Ridership and Revenue Base Scenario, Task Report No. 6, Wilbur Smith Associates, 1997, unpublished data). The assessment had to address uncertainties inherent in a new technology, provide a level of comfort to RTA decision makers, and be completed in a short time.
PROPOSED ROSEMONT PRT SYSTEM
Among the four competing municipalities in the Chicago metropolitan area, the village of Rosemont was selected for the PRT demonstration site because of its higher ridership potential, local commitment, and desirable construction site. With 4,000 residents and the employment of 23,000 workers, Rosemont is adjacent to O’Hare International Airport and is about 24 km (15 mi) from downtown Chicago. Employment-based rather than residential-based land use characterizes the PRT services area:
3,000 hotel rooms,
167,000 m² (1.8 million ft²) of office space,
73,000 m² (780,000 ft²) of convention facilities,
4,200-seat performing arts center,
6,000 Chicago Transit Authority (CTA) rail station boardings, and
200 condominium units.
The initial PRT system configuration was proposed in 1997. The system consisted of a 5.5-km (3.4 mi) guideway, 7 stations, 45 vehicles, 3 power substations, and 1395 m² (approximately 15,000 ft²) of maintenance and control facilities. After extensive studies of capital costs and ridership potential, the configuration was changed slightly. The guideway was shortened to 4.8 km (3 mi), and the number of stations was reduced to six. The study results of the 1997 PRT system configuration are presented.
ASSESSMENT OF PRT CAPITAL COSTS
Objective
The objective of this cost analysis was to provide an accurate check of the capital costs associated with the PRT project. Specifically, the three analysis goals were to (1) develop estimates of the PRT system capital costs; (2) establish a reasonable cost range for the PRT system; and (3) conduct an order-of-magnitude evaluation of Raytheon’s cost projections. The PRT capital cost analysis is uniquely useful in decision-making support. An independent evaluation of the costs projected by Raytheon and the development of reasonable ranges for RTA decision makers based on similar technology enhance the assessment of PRT system deployment at Rosemont. Given the limited time available and that PRT is a new technology, an investigation using the costs for automated-guideway transit (AGT) and automated people mover (APM) systems was undertaken. Research indicated that PRT compares better with AGT and APM systems than over existing technologies in its system characteristics and operating environment as shown in Table 1.
Table 1: PRT Comparable Transportation Mode
System Characteristics RTA Rosemont PRT System
North American AGT/APM Systems
Light Rail Transit (LRT) Systems
Heavy Rail Transit (HRT) Systems
Type of Service Collector/ Circulator
Shuttle/Collector/ Distributor/Circulator
Line-Haul or Distributor
Line-Haul
System Size Small
Small
Medium
Large
Capacity Medium
Medium
Medium
Large
Trip Length Short
Short
Medium
Long
Operating Environment Driverless, Fully Automated Vehicles
Driverless, Fully Automated Vehicles
Drivers
Drivers & Conductors
Security Systems CCTV,PA, & Intercom Systems
CCTV,PA, &Intercom Systems
PA System
PA System
High capital costs have been major obstacles to the wide application of AGT and APM systems in North America. With their applications generally in medium- and low-density development areas, their ridership levels have been too low to justify the development of high-capital-cost AGT and APM systems. If the proposed PRT system can be shown to have similar or lower capital costs than AGT and APM systems, its deployment chances would be better because of its superior service characteristics. PRT offers travelers rapid, affordable, personal, nonstop travel on demand. Such service-oriented characteristics would result in higher ridership and system revenue.
Analysis Approach
The initial research effort found variations in PRT-like AGT or APM designs as far as guideway and vehicle dimensions, propulsion and suspension mechanisms, control and communications technology, and switching philosophy. Because of these significant technological differences, comparing total system costs would have small value. Although these systems were not comparable at the system or project level, research indicated some similarities at the component level. Comparing PRT component parts to the component parts of similar AGT or APM systems would provide more accurate PRT cost estimates.
Data availability is the most important factor in determining system components for analysis. The more detailed the component categories, the better the capital cost analysis will be. A literature review found two studies that compared AGT and APM capital costs at the component level. The first study provided insight into important capital cost variations of 16 AGT systems in the United States (1). The second study compared five urban and eight airport APM systems in the United States by analyzing site conditions and construction techniques that affected capital costs (2). The two studies provided the basis for determining the components for the PRT cost analysis. Seven components were selected because of data availability – (1) guideways; (2) stations; (3) maintenance and control facilities (MCF); (4) power and utility systems; (5) vehicles; (6) command control, and communications (CCC) systems; and (7) engineering and project management (1-4). Unlike the two studies identified in the literature, Raytheon’s PRT designs and functionality had more influence on the AGT systems selected for the component cost comparison analysis. Three analysis techniques were used – (1) statistically significant regression analysis, (2) measurements of a central tendency for comparable systems, and (3) statistics for all AGT systems.
Component Analysis Methodology
For each component analyzed, the following tasks were performed:
Identification of the major factors affecting the cost of the component category,
Selection of comparable components and their associated characteristics and capital costs,
Identification of features unique to the PRT system compared with components of other systems,
Assessment of the capital cost impact of specific PRT features, and
Development of reasonable ranges of capital costs using the three analysis techniques.
More than 15 factors for each of the seven components were compiled and examined. Table 2 lists major cost factors affecting component capital costs. Because of the proprietary nature of the PRT and AGT systems, factors associated with high or low costs for the CCC component were difficult to characterize. The engineering and project management component covered a wide array of activities or costs, such as regulatory requirements, certification processes, construction schedules, commissioning, and other miscellaneous items. Because none of the published data revealed such detailed information, the statistics for the 23 AGT and APM systems in North America were used to develop the comparable cost for this component.
Table 2: Major Factors Impacting Component Capital Costs
HIGHER COST
VS.
LOWER COST
Guideway Component
Underground
Field constructed
Curved guideway segments
Linear induction motors
Major bridge spans
Heated guideway
vs.
vs.
vs.
vs
vs.
vs.
Elevated and at-grade
Prefabricated
Straight Guideway segments
Rotary DC motors
Uniform span lengths
Unheated guideway
Station Component
Underground
Free standing
Enclosed platform
Overpasses/Underpasses
ADA/Elevator & Escalator
Fare collection
CCTV, Intercom, PA
vs.
vs.
vs.
vs
vs.
vs.
vs.
Elevated and at grade
Joint use
Open platform
Simple pedestrian access
Stairs or at-grade access
No fare collection
No station communication
Maintenance & Control Facility (MCF)Component
Larger system
Larger vehicle fleet
Large/complex vehicles
Large maintenance facility
Indoor vehicle storage
All-weather operations
vs.
vs.
vs.
vs
vs.
vs.
Small system
Small vehicle fleet
Small/simple vehicles
Small maintenance facility
Outdoor vehicle storage
Warm climate operations
Power & Utility Component
Larger guideway length
Higher voltage electric motors
Larger vehicle fleet
Larger/heavy vehicles
Higher design speed
Heated guideway
vs.
vs.
vs.
vs
vs.
vs.
Small guideway length
Lower voltage electric motors
Small vehicle fleet
Small/Light vehicles
Lower design speed
Unheated guideway
Vehicle Component
Large fleet size
Steel wheel on rail
On-board switch
Large/heavy vehicles
Higher design speed
ADA compliant
vs.
vs.
vs.
vs
vs.
vs.
Small fleet size
Rubber tired vehicles
Side/center guidance surface
Small/light vehicles
Lower design speed
Not ADA compliant
Command, Communication and Control (CCC) Component Major Factors Type of centralized control Guideway and station control Automatic vehicle control technology On-demand service type contro system Security systems
Regression Analysis
Regression results were found to be reasonable and statistically significant for four of the seven components – vehicles, guideways, power systems, and MCF. Vehicle component PRT cost estimates are an excellent example of how the regression analysis was performed. One major difference between the PRT and AGT vehicles is vehicle dimensions or weight. The PRT vehicle is substantially smaller, seating no more than 5 passengers, whereas some AGT vehicles can carry more than 100 riders. Proper scaling of the AGT vehicles to comparable PRT vehicle size was a major issue. To objectively scale down the AGT vehicle costs comparable with PRT vehicle costs, a regression analysis technique was applied. Regression analysis is a statistical method to develop the functional relationships between variables. For example, total vehicle costs is regarded as a function of fleet size and vehicle dimensions. After this relationship is established, the researcher can substitute the PRT fleet size and vehicle size to predict the total PRT vehicle costs. Another unique feature of the PRT design is in onboard switching mechanism. Incorporating small vehicle size and onboard switches into the AGT vehicle regression analysis would be a better method to predict PRT vehicle comparable costs.
In addition to vehicle size, comparable AGT vehicle functionality was also carefully examined. Among 25 AGT systems built in North America, 5 systems were judged to be outliers for the vehicle component regression analysis. Vehicles based on the Wedway People Mover systems developed by the Walt Disney Company were excluded from the regression analysis. These vehicles have limited functions compared with other AGT or PRT vehicles. These systems use small polyurethane tires that nave no motors. Propulsion units in the guideway move at 8 to 20 km/h (5 to 6 mph), and their applications are limited to nonhostile environments. Three AGT systems use this technology – Bush Intercontinental Airport (Houston, Texas), the U.S. Senate subway, and the Disney Wedway People Mover. Vehicles used for Jacksonville, Florida, and Detroit, Michigan, downtown people movers were also identified as outliers. Those vehicles had very high unit costs (more than $2.5 million per vehicle). The pricing structure for the vehicles used in the Jacksonville system reflects a single supply source. In fact, the Detroit system was an outlier for all other component regression analyses as well.
The selection of independent variables was crucial in performing the regression analysis. As mentioned previously, small vehicle size and onboard switches are the unique features of PRT vehicle design. A relationship including these two independent variables would be desirable. Equivalent vehicle seated spaces (EVSS), calculated from exterior vehicle floor area, was selected to measure the impact of vehicle size on vehicle cost. An onboard switching dummy variable was initially included in the regression analysis. Further analysis of the observations, however, indicated that only one AGT system (Morgantown, West Virginia) has onboard switches. Regression results for this independent variable were not as significant as expected. It was concluded that costs to account for PRT onboard switches should be adjusted after the regression results.
The final vehicle component regression analysis was based on 20 North American AGT systems with three independent variables – fleet size, EVSS, and type of motor (rotary DC versus linear induction). All coefficients of the three variables were significant at 99 percent confidence. The multiple regression had an R² value of 0.858, which is shown in Figure 4. The PRT-comparable vehicle unit cost is $0.37 million per vehicle in 1996 dollars. A 10 percent upward adjustment accounting for the PRT onboard switch and more sophisticated vehicle controls was made. The final comparable unit cost is about $0.4 million per vehicle.
Regression analyses for other components was performed following similar procedures, and the results are shown in the supplemental presentation. It includes a graph that shows AGT guideway cost versus guideway length. Both at-grade and underground guideway lane miles were converted to equivalent elevated lane miles. The simple regression analysis was based on 16 North American AGT systems and had an R² value of 0.964. Another graph shows AGT MCF capital cost versus guideway length. The multiple regression analysis was based on 19 North American AGT systems with three independent variables: guideway lane miles, system complexity index, and EVSS. The multiple regression had an R² value of 0.873. A graph that shows AGT power and utility cost versus guideway length is also available. The multiple regression analysis was based on 17 North American AGT systems with three independent variables: guideway lane miles, fleet size, EVSS, and snow and ice removal dummy. The multiple regression had an R² value of 0.943.
Comparable Component Analysis
Both station and CCC component costs were estimated using comparable component analysis techniques. In the analysis of the station components, factors such as station and platform configurations, site environment (urban, suburban, and airport settings), vehicle dimensions, berthing requirements, revenue collection, and communications features were examined for each of the 20 AGT and APM systems with available data. The evaluation showed that only four systems (Jacksonville and Miami, Florida; Las Colinas, Texas; and Morgantown, West Virginia) shares the most PRT-like characteristics – an urban or dense suburban environment; free-standing, elevated stations with an open platform design; a moderate number of stations and platform length; elevators and Americans with Disabilities Act accommodations; and a closed-circuit television and station intercom system. The statistics from the four systems were used to develop a comparable cost for the PRT station.
CCC cost analysis used a similar evaluation process. Five AGT and APM systems – Hartsfield International Airport (Atlanta, Georgia), Denver airport, the Miami Metromover, Morgantown, and Vancouver, British Columbia – were identified as sharing similar characteristics with Raytheon’s PRT system. These systems all have a three-level control architecture, including (1) automatic train operation for speed, stops, and doors, (2) automatic train protection for vehicle direction, spacing, and safety, and (3) automatic vehicle supervision for the system-control interface. Each system has extensive security, including closed-circuit television, intercom, and public address systems.
Analysis Limitations
The intent of the cost analysis was not to replace Raytheon’s cost projections but to provide other data points for RTA’s decision makers. The cost analysis was limited to published data. Some inconsistent component reporting was found. For example, some systems reported maintenance of tracks or sidings as part of a guideway cost, while others categorized the tracks as a maintenance facility cost. Morgantown reported onboard vehicle control equipment as a CCC cost instead of a vehicle cost. Inconsistent reporting was also found on whether or not to include utility or street location, landscaping, and urban design in the capital cost. Another inconsistent analysis area was project status. Cost estimation accuracy varies by level of engineering design. The study compared Raytheon’s preliminary design estimates with actual costs of detailed design and construction of AGT systems. Inclusion of a contingency cost in Raytheon’s cost projections would be more appropriate for the comparative analysis with AGT systems. Despite these limitations, this planning review of PRT costs provided not only PRT-comparable cost estimates, but also revealed the cost relationship between PRT and AGT systems.
Analysis Results
Table 3: PRT-Comparable AGT System Component Cost Summary Results
(Unit Costs in Millions of 1996 Dollars)
Range of Cost
System Component
Unit*
Low High Comparable Costs (in quartile of the cost range) Analysis Technique Guideway Cost per Guideway Lane Mile $5.2
$15.2
2nd/3rd
Range from regression analysis of 16 AGT systems Stations Cost per Station $0.5
$2.9
2nd/3rd
Range from 4 comparable AGT systems Maintenance and Control Facilities Cost per Guideway Lane Mile $1.3
$2.2
2nd/3rd
Range from regression analysis of 19 AGT systems Power systems Cost per Guideway Lane Mile $1.2
$2.4
2nd/3rd
Range from regression analysis of 17 AGT systems Vehicles Cost per Vehicle $0.2
$0.7
2nd
Range from regression analysis of 20 AGT systems Command, Control & Communication Cost per Guideway Lane Mile $1.7
$5.8
2nd
Range from 5 comparable AGT systems Engineering and Proj. Management Cost per Guideway Lane Mile $2.2
$8.9
3rd
Range from 23 AGT systems * 1 mile = 1.61 km
Table 3 summarizes the study results by component. Cost comparisons between Raytheon’s cost projections, PRT-comparable AGT cost estimate, and an average of 17 AGT systems were also made. Selected conclusions from the analysis follow:
The PRT-comparable cost range tends to be on the lower side of the cost ranges for the 17 AGT systems.
Raytheon’s cost projections for five of the seven components are comfortably in the comparable component cost range.
From the system cost perspective, Raytheon’s projection is in the range of PRT-comparable AGT system cost. If initial PRT deployment or nonrecurring costs are excluded, Raytheon’s PRT unit cost projection is significantly lower than the comparable AGT unit cost.
ASSESSMENT OF ROSEMONT PRT RIDERSHIP FORECAST
In addition to the cost study, the RTA planners were asked to evaluate ridership estimates prepared by Wilbur Smith Associates. In reviewing consultant task reports, several positive aspects of the ridership study were identified:
Substantial primary data were collected.
A sound forecasting methodology was applied.
The travel market for the PRT system was segmented.
Data Collection
An intensive land use inventory of the PRT station area was performed to determine the size of the PRT station catchment areas. PRT focus group sessions were conducted to understand potential barriers to PRT usage. Employer/tenant, pedestrian intercept, and state preference surveys collected travel characteristics for predicting total travel market, travel patterns, and travel modes for the PRT service areas.
Forecasting Methodology
The forecasting methodology used for the study is the standard forecasting procedure consisting of trip generation, trip distribution, and modal choice. Total trips generated in the PRT alignment area were based on the land use characteristics and building activities surrounding PRT station areas. A land use inventory, employer/tenant surveys, and trip generation rates published by the Institute of Transportation Engineers were the primary data sources for projecting the total travel market for the proposed PRT system. The total travel market was segmented into different trip categories and was further aggregated into potential origins and destinations using information collected from the pedestrian intercept surveys. The travel patterns were distinguished between days with and days without events at the Rosemont Convention Center. A computerized network of the study area was developed to replicate the existing roadway system. Each building in the study area that could be a potential trip origin or destination was defined as an analysis zone.
Travel distances from these buildings to sidewalks, streets, or parking garages were coded in the computerized network representing actual building-to-building travel. The modal choice models were calibrated based on the stated preference survey and refined using information collected from the pedestrian intercept survey. The model structures were designed to estimate the relative utility of PRT as measured by several attributes, including travel time, walk time, and travel cost, against the attributes of other modes. The major competitive modes to PRT include automobile, walking, shuttle bus, and taxi. Although CTA rail and public bus are travel modes into the PRT services area, they would supplement rather than compete with PRT.
Travel Market Segmentation
The PRT ridership projections were based on the results of substantial market segmentation analyses. Several levels of travel market disaggregation were performed. Nine trip purposes and four trip categories were used to segment the total travel market. Trips in each of the trip categories were further disaggregated into internal and external origins or destinations. The final market segments were as follows:
Rosemont-to-Rosemont (internal) trips – work, business-related, non-work, convention-related;
External-to-Rosemont trips using transit – work, business-related, non-work, convention-related; and
External-to-Rosemont automobile trips – work, business-related, non-work, and convention-related.
Four separate demand forecasting models were developed. The model calibration results show that PRT is the most attractive mode for computer trips now using CTA rail services, convention visitors’ business-related internal trips, and all internal non-work trips.
PRT Ridership Forecasting Results
Weekday ridership – with a range of PRT fares from free to $2.00 and transfer fares from free to $1.00 – were projected. Estimated weekday ridership with a $1.50 PRT fare and $0.50 transfer fare for three types of activities at the Rosemont Convention Center is shown in the supplemental presentation.
It was found that if a trade show were being held at the convention center, approximately 6,650 riders would use the PRT system on an average weekday in its initial year. The station with the most activity is expected the CTA Rosemont station, which would account for 36 to 42 percent of trip origins or destinations depending on convention center events. A diagram was prepared to show the station-to-station movement from CTA's Rosemont station. The largest flows are expected from the CTA station to office buildings or the convention center.
CONCLUSIONS
RTA and Raytheon senior management’s confidence in the PRT capital cost and ridership estimates has increased. The estimated 5,530 to 6,650 passenger trips per weekday is reasonable on the basis of the consultant’s sound forecasting approach, extensive data collection effort, and detailed market segmentation analysis. Raytheon’s PRT capital cost estimate is within the range of 17-23 million/lane-km (27-37 million/lane-mi) of PRT-comparable AGT system cost. The regression techniques worked well for most cost component groups and show promise for use in other cost analyses involving new technologies or application of existing technologies at a scale outside the bounds of previous experience.
REFERENCES
1. Cost Experience of Automated Guideway Transit Systems. Dynatrend, Inc., Woburn, Mass., April 1984.
2. Factors Influencing Future Transit Efficiency: Automated People Mover Cost Study. Florida International University, Miami, Dec. 1994.
3. World People Mover Survey, APM Case Studies. Transportation Systems for the Twenty-First Century (Trans21), Boston, Mass., 1990.
4. Characteristics of Urban Transportation Systems. FTA, U.S. Department of Transportation, 1992.
Last modified: June 25, 2003