Objective

The passage of the American with Disabilities Act (ADA) has changed the landscape for demand responsive transit (DRT) systems. First, the demand for this type of transit service has experienced tremendous growth. As a result, DRT ridership has nearly doubled and operating costs have increased to $1.2 billion annually, more than 6% of the national budget for public transportation in 2000 (Federal Transit Administration 2000). Second, besides the increase in demand, ADA also set strict guidelines for the providers on trip denials and on-time performance. In essence, transit agencies today are expected to provide better services while experiencing increased usage for demand responsive transit systems. These systems are highly subsidized. The National Transit Summaries and Trends (NTST) report for 2000, the most recent year available, indicates that the average cost per passenger trip for DRT systems is $16.74 with fares ranging from $1.50-$3.00. By way of contrast, the NTST report indicates that the average cost per trip for fixed route lines is $2.19 with fares being roughly 25% of the cost. With this high of a subsidy for DRT systems and the continued growth, the operation of these systems which is not optional for transit services as mandated by ADA, will put a tremendous strain on the budget of transit agencies. We aim to address this national problem by studying innovative methods that can increase the productivity and reduce cost for these types of services. The objective of this study is to investigate the impact of specific operating practices on the productivity and cost of DRT systems.

Methodology

One important component that adds significantly to the cost and productivity is deadhead miles. Many factors influence the amount of deadhead miles including the system design and time window size. In terms of system design, many large transit agencies operate a decentralized control structure as opposed to a centralized structure. For example, Access Services Inc. (ASI), the agency responsible for coordinating paratransit service within Los Angeles County, divides the service area into six regions. In each region, paratransit service is contracted to a private operator who is responsible for picking up customers in that region. If a vehicle drops off a passenger outside their service region, they cannot pickup another passenger in the outside region. Although this decentralized method is easier to manage, it can lead to tremendous inefficiencies due to the increased deadhead miles since the number of passengers traveling outside their region can be as high as 30% in Los Angeles County, as reported by ASI. Hence, the return trip will be done by a different service provider regardless of the dwell time of the customer at their dropoff location. Furthermore, under this situation, the customer is required to make two different reservations, one for each provider.

Clearly, the current decentralized operating practice is easier to manage. However, this may come at the expense of a significant number of deadhead miles. This study will compare the current decentralized approach with a centralized strategy where any vehicle can pickup any passenger regardless of service region. We will quantify the productivity and cost impact of using the centralized strategy. A drawback with the centralized strategy is that the Computer Aided Dispatching (CAD) systems of different providers need to efficiently communicate between them in order to effectively manage such a design. In addition to investigating a completely centralized design, we will investigate the effect of centralizing only part of the service area (in particular those areas that have high travel between regions).

In order to compare the different system designs, a methodology for scheduling the requests needs to be developed. We will develop a fast robust scheduling algorithm based on a regret insertion technique to schedule the requests. We will also compare our algorithm against existing techniques such as classical insertion techniques.

Another important factor impacting productivity and cost is the size of the time window. Customers prefer small time windows. However, in order to maintain small time windows, transit agencies may have to increase their fleet size and the possibility of ridesharing also decreases, contributing to increased cost and less productivity. Therefore, the determination of the best time window size needs to balance customer service with the negative impact on productivity and cost. Currently, ASI uses a 20 minute time window whereas many other agencies use a 30 minute time window. In order reduce costs ASI is interested in increasing their time window to 30 minutes. However, there exist no models that can measure the quantitative impact on total trip miles and fleet size of the increased time windows. Without a means to measure this impact, it would be difficult for transit agencies to justify the time window used. We will develop a model that determines the fleet size and total trip miles including empty trip miles as a function of the time window.

In order test the models and algorithms on a realistic operating environment, ASI will provide data on at least ten days of relevant operations for the different service regions. The data will include information on demand locations (pick-up and drop-off points), requested pickup times, fleet size, and service times. Based on the collected data, we will develop probability distributions of demand information in order to develop simulation and analytical models that represent ASI operations. These models will be used to study the impact of the two above factors.