Comparing Fixed-Route and Demand-Responsive Feeder Transit Systems in Real-World Settings

This research presents a method of comparing fixed-route transportation systems and demand-responsive feeder transit systems using passenger survey data, published transit schedules, and optimal routing techniques. Demand-responsive transportation can be utilized to improve transit service levels in low demand areas. Since cities can vary significantly in demand across the region and time of day, it is imperative that an effective means of determing when demand-responsive services can out perform fixed-route services and vice versa. This research builds upon existing comparison techniques, that are focused on gridded street systems, and expands the techniques to includes all types of street networks, transit schedules, and passenger demand levels. The generic techniques are presented and a case study is given for the city of Atlanta to determine where demand-responsive feeder systems might be implemented to improve customer satisfaction and reduce operating costs. TRB 2013 Annual Meeting Paper revised from original submittal. Derek Edwards, Kari Watkins 2 TRB 2013 Annual Meeting Paper revised from original submittal. Derek Edwards, Kari Watkins 3 INTRODUCTION Extending public transit outside of dense urban areas is a major challenge for transportation operators. In medium and low density areas, it may not be economically viable for transit to run frequent service within easy walking distance to every location (1). As a result, ridership in these areas remains low (2). One method of effectively reaching areas of low density is to use demand-responsive transportation. In a demand-responsive transportation system, buses do not run on a fixed schedule or a fixed route. Instead, the buses or vans respond to real-time passenger demand and create dynamic schedules and routes. Since cities can vary signficantly in density from one area to another, determining which parts of the city that are better suited to demand-responsive transit than fixed-route transit is key to building an efficient system. To this end, this paper proposes a method of utilizing open source data and software to compare the performance of demand-responsive transportation (DRT) and traditional fixed-route transportation (FRT) in a variety of street and transit layouts. This paper will briefly review the state of the art of demand-responsive transportion and current methods for comparing DRT and FRT. Existing comparison techniques are best suited for homogeneous city layouts i.e., gridded street systems with a uniform dispersion of passenger locations. Those comparison methods will then be extended to compare heterogeneous layouts with random passenger locations and complex street patterns. Two case studies are presented for comparing DRT and FRT. Both of these case studies operate as feeder systems where demand-responsive transportation is used in conjuction with existing static transit infrastructure to better solve the last mile problem. The first method is applied to a simple city with uniform street layouts and passenger locations. The second method is applied to the city of Atlanta as a case study for comparing DRT and FRT in a broad range of street layouts. BACKGROUND AND LITERATURE REVIEW While the idea of using dynamic vehicles to meet real-time passenger ride requests is not new, interest in demand-responsive transportation has increased in recent years (3). The increase in demand-responsive transportation research is due to the proliferation of mobile phone and mobile internet devices that make the process of requesting and organizing trips faster and simpler. Mobile internet-enabled devices allow passengers to instantly submit trip requests to a transit dispatcher. The transit dispatcher collects these trip requests and dispatches one or more vehicles to transport the passengers to their destinations along an optimal route. Review of the Dial-a-Ride Problem The mathematical problem of optimizing a route to visit a set of passenger locations is known as the dial-a-ride problem (4). The objective of the dial-a-ride problem is to create optimal dynamic bus routes to service a set of passengers, curb-to-curb, with a priori information of the passengers' origins and destinations. A thorough mathematical model of the dial-a-ride problem is presented by Cordeau and Laport in (4)(5). In addition to purely demand-responsive or purely fixed-route systems, a set of hybrid systems exist known as integrated demand-response systems (6). In an integrated demand-response system, a fixed-route transit network is leveraged to decrease operating costs for the demand-responsive vehicles (7)(8). Often the role of the demand-responsive vehicles in such a system is to provide first-mile transportation to a high-speed rail or bus line (9)(10). This type TRB 2013 Annual Meeting Paper revised from original submittal. Derek Edwards, Kari Watkins 4 of system is known as a transit feeder system. These integrated transit and feeder systems have shown promising results for balancing the high cost and flexibility of demand-responsive systems with the highly efficient but more rigid fixed-route transit systems (11)(12). For this reason, the systems studied in this research operate as feeder transit systems. Review of Demand-Responsive Transportation Comparison Techniques Since this research seeks to appropriately apply fixed-route and demand-responsive transportation systems where they will be most effective, an objective method of comparing these two systems is required. Current methods of comparing DRT and FRT transportation have focused on homogeneous street layouts. Methods by Diana et. al (13), compare FRT and DRT in gridded street systems as well as ring-radial systems, common in European cities. Li and Quadrifoglio perform similar analysis of feeder systems (3) (14). The feeder systems analyzed by Quadriofoglio and Li divide a city into multiple feeder pools where each pool is rectangular in shape with a transit station at one end of the pool (15). A feeder bus then traverses the length of the feeder pool, collecting passengers and dropping them at the transit station. The techniques developed by Li, Quadrifoglio and Diana provide an excellent foundation for comparing DRT and FRT systems and determining when a DRT will outperform an FRT system. A key contribution of their research is identifying the costs of travel for the passenger and the transit operator (16). The passenger costs are broken down into walking time for the passenger, time spent waiting for a vehicle, and the riding time for the passenger. These costs do not represent all the costs levied on the passenger, notably the fare is missing from this equation. These costs are intended to represent the variable costs to the customer. Theoretically, a transit fare is a flat rate, but the time it takes to complete a trip can be controlled by the operator in real time. The transit operator’s cost is expressed as vehicle miles traveled (VMT). Once again this is not intended to represent all the costs of operating a vehicle. It is merely an approximation of the costs that can be controlled by the driver, specifically how many miles the vehicle drives. These “costs” are what the dial-a-ride algorithms attempt to minimize when selecting optimal routes. The main drawback of the existing techniques is that the cities and service areas studied are homogeneous, meaning that there is no randomness in their layout. Each system assumes a perfect grid or circular city layout. These techniques also work best for transit systems that arrive at unchanging intervals and with passenger arrival rates that are uniform. The next logical extension of this research is to apply these comparison techniques to heterogeneous systems that include random passenger arrival rates, non-uniform street layouts, and irregular transit schedules. ANALYSIS OF GRIDDED STREET SYSTEM In the first of two case studies, a decision must be made to expand the existing transit infrastructure or adopt a demand-responsive transportation policy in a homognous city using a feeder system. This work is a direct extension of the work by Li and Quadifoglio where instead of analyzing individual feeder areas, the system as a whole is analyzed. Layout and Behavior of the System The bus layout of the city under examination is a grid consisting of NxN bus stops evenly separated by a distance H, as seen in Figure 1. The grid of bus stops are serviced by a set of N north/south bus lines and a set of N east/west bus lines traveling at a speed of with headway . TRB 2013 Annual Meeting Paper revised from original submittal. Derek Edwards, Kari Watkins 5 Figure 1: Bus stop layout in gridded street simulation. In the fixed-route system, passengers walk to the nearest bus stop, wait on the bus, possibly transfer between buses, ride a second bus, and then walk to their final destinations. In the demand-responsive system, a dynamic-route vehicle carries the passenger from his/her origin to the nearest fixed-route bus stop and dynamic-route vehicle also handles transporting the passenger between the passenger’s final bus stop and his/her final destination. The set of dynamic vehicles act as a feeder system for the fixed-route bus network. The purpose of this examination is to determine whether a dynamic, feedersystem with sparse fixed-bus stops is preferable to a purely fixed-route system with frequent stops. These two systems will be compared for various densities of bus stops as well as varying levels of passenger demand. As with Li and Quadrifoglio, a combined cost of passenger walk time, wait time, and ride time will be considered as the passenger costs, and vehicles miles traveled will represent operator costs. Passenger Costs To create an efficient transit system, passenger costs and operator costs must be minimized. In this example, passenger costs consists of walking time , waiting time , and riding time . The combined passenger costs, , is represented as a weighted combination of these individual costs.