Evaluation of real-time bus arrival information systems

This study uses an expected utility approach to evaluate the performance of real-time bus arrival information systems (BAIS). Three BAIS alternatives are considered in terms of the real-time bus location data available to a "broadcaster" predicting bus arrival times for passengers at a bus stop: BAIS1 has no available real-time data; BAIS2 considers real-time data on the time since the most recent bus departure from the bus stop; BAIS3 considers real-time data on the headway at a point upstream from the bus stop in addition to the data available in BAIS2.;Bus and passenger arrivals at a stop are modeled as random variables, and passenger utility is modeled as a function of the actual waiting time and the predicted bus arrival time conveyed to the passenger upon arrival to the stop. This predicted time is assumed to be determined by the broadcaster who maximizes the passengers' expected utilities given the available real-time data. The expected utility across the passenger population is analytically formulated for each BAIS alternative. Since closed form solutions can be intractable, Monte Carlo simulation is used to evaluate the BAIS alternatives under different scenarios reflecting bus operating performance and passenger utility functions.;Simulation results show that the headway mean has an important effect on the expected utility of BAIS1 but little effect on the expected utilities of BAIS2 and BAIS3. The headway standard deviation, on the other hand, has little effect on the expected utility of BAIS1 but a large effect on the expected utility of BAIS2. The effect of the standard deviation on the expected utility of BAIS3 is tempered by the additional real-time data on the approaching bus. The various passenger utility functions examined have little effect on the expected utility patterns of any of the BAIS alternatives. A sensitivity analysis relating to the effect of the broadcaster's knowledge shows great sensitivity to accuracy in the headway mean but little sensitivity to accuracy in headway standard deviation and utility function parameters. An application to three stops on a realistic bus system illustrates the reasonableness, value, and applicability of the developed methodology.