A simulation study of dynamic scheduling of a VTOL airport feeder system

Abstract

Introduction: In considering ultra short haul, high density transportation systems, it may become feasible to use short term, real time decision making in operating the system. Here the dispatch of vehicles would be based upon actual traffic demands, the passenger waiting times for service, with perhaps some consideration given to expected future demands at the originating and downstream stations. This is called dynamic scheduling, or demand scheduling to differentiate it from the scheduling planning process which uses as input the expected or average demands for the system over some extended period. An example of pure dynamic scheduling is present taxi service in most urban areas. A fleet of roving or dispersed taxis is controlled by a centralized dispatcher who receives all demands by phone, and assigns a vehicle to a service using a radio communication system. The other extreme is typified by present domestic airline schedules where services, vehicle and crew assignments, etc. are ordained at least a month in advance, and the schedule is followed as closely as possible. Most transportation systems fall in between these extremes with trains adding extra cars or bus carriers making extra sections available at short notice, etc. The EAL shuttle service is partially dynamic in that the guarantee of a seat may force an unplanned extra section, and is partially planned since a continuing study of the patterns of demand allows planning for most extra sections. The published shuttle timetable remains fixed although departures occur before, on, and perhaps after the scheduled time. By having a fixed operating plan, the job to be performed becomes deterministic, and adequate planning can ensure good operating efficiencies over the system as measured by load factor, vehicle and crew utilization, ground facility utilizations, etc. With an uncertain operating plan, the system must have above average resources in order to be able to call them into service at peak or above average times. This implies lower load factors and lower utilizations on the average. The higher efficiencies mean lower costs, and presumably lower fares. The lower efficiencies of the dynamic system may mean higher costs, but will be accompanied, presumably, by better service for the traveller. A number of questions are thereby raised: How much will the traveller pay for an improved service? What sort of service improvements can we provide by being responsive to actual real time demands, and what will they mean in operating costs? What type of market will allow most effective use of dynamic scheduling? What kinds of dynamic scheduling strategies can be employed? How do we discover efficient strategies, and how do we test them? There does not seem to be any clear or well defined set of answers to such questions. This is a report on some preliminary investigations into the problems of dynamic scheduling in very short haul markets which exist in collecting and distributing passengers from a major transportation center.

(cont.) The decision making process by which the system operates -3- is called a scheduling strategy. Given the present system state in terms of accurate real time information concerning demands, passenger waiting time, vehicle availabilities, etc. and some short term expectations of future system states, a set of operating rules is established which determine the transportation system response. This set of rules, (or strategy always exists, either explicitly in the form of management policy directives, or implicitly in the form of the experience and intelligence used by a taxi dispatcher. Whether complex or simple, there are a wide variety of strategies which can be selected far testing in various markets. Each strategy will use certain information about the system state, which assumes that such information will be made readily available. One of the first problems is to discover strategies which allow efficient operation of the system with an economical use of data about the present and projected system states. This report describes the operation of a final strategy which has evolved from reference 1, and further testing during this study.

(cont.) The classical aim of airline managements is to maximize short term profits. It could easily be minimization of costs, maximization of revenues or aircraft utilization. From the public service point of view or longer term management objectives, it could be minimization of passenger waiting time. It could be some weighted combination of any of these factors. Different situations will dictate deferent objectives, and it is not clear which objectives are preferable, or what type of strategies are most effective in achieving any chosen objective.

(cont.) Simulation models of operations systems have benefited management in the decision making process and in comparing basic alternatives of operating policy. Computer simulation is a technique which provides management with means of testing and evaluating a proposed system under various conditions. In our study the system's behavior is modeled by a computer program which reacts to various scheduling strategies in a manner very similar to the system itself. With the use of the simulation model, management can thus determine the effects of many alternate strategies without tampering with the actual physical system. The result is that we do not risk upsetting the existing physical system without prior assurance to some degree of confidence that the proposed changes in strategies will be beneficial. Computer simulation thus produces a system which is efficient and fulfills the system operational objectives. Use of simulation can save cost and time. In this study for example, five days of airline operation have been simulated in less than three minutes of computer time using the General Purpose System Simulator on IBM 7094. The simulation allows us to follow through the system and observe the effects of blocking caused either by the need of time-share facilities or caused by limited capacity of parts of the system. Outputs of the program give information on: 1. The amount of traffic through the system, or parts of the system. 2. The average time and the time distribution for traffic to pass through the system, or between selective points on the system. 3. The extent to which elements of the system are loaded. 4. Queues in the system. 5. A departure schedule. 6. Miscellaneous parameters of interest in the system. With our simulation model a number of different dynamic scheduling strategies have been examined. The final decision rules are described in chapter 2. The effects of variations in these final decision rules are shown in chapter 3.