Real-Time Bus Information: Users’ Perspectives and Arrival Time Estimations
Abstract
The uncertainty in bus arrival times creates a significant disutility; the provision of real-time bus arrival information is reported as a cost-effective way to reduce this disutility. From a wide viewpoint, this study attempts to contribute on two main topics. One goal is to reveal the transit users’ perspectives regarding real-time bus information by conducting a users’ survey. The other goal is to provide insights on the improvement in reliability of real-time bus arrival estimates. Regarding the users’ perspectives, this study investigated several aspects such as the preferred formats (point or interval estimate) of estimated arrival information, the value of real-time information under different weather condition, and disutility of headway in a scheduled and real-time information system. Results showed that a majority of the respondents preferred the interval estimate over the point estimate. The value of real-time information was $0.59 and $0.41 per trip when the weather was below and above 0°C, respectively. The disutility of bus headway of a real-time information system was found to be around half of the disutility of a scheduled information system. This study investigated the improvement in reliability of bus arrival time estimations in three parts. The first part examined the impact of the time interval of measuring real-time bus location (GPS) data on the accuracy of bus arrival estimations. Incorporating real-time bus location data below a 30-second time interval did not increase the accuracy of the applied estimation models. Estimations based on current and average historical bus speed outperformed other estimations only if the time interval of the location data was short (about two minutes). Estimations that relied on similar historical trips were more accurate when both the estimation horizon and time interval were longer. Horizon is defined here as the distance between the point for which the estimation is made and the location of the bus at the time of estimation. The second part investigated the changes in bus travel time characteristics with horizons. The experimental results showed that a significant change in bus travel time characteristics was observable around a horizon range of 7-8 km. The analysis of changes in probability densities with pseudo horizons showed that bus travel time distribution converges from a rightly skewed distribution to a more symmetrical distribution from a shorter to a longer pseudo horizon in general. Lognormal and normal distributions are found to be the best distributions for before and after a cut-off horizon of 7-8 km, respectively. The third part was about the development of Bayesian models to provide dynamic bus travel time distributions that was updated in light of new observations and include the estimation uncertainty chosen by users or transit service providers. Two models were developed in a Bayesian framework to achieve this objective: Model 1 uses the dynamic linear model (DLM) concept and Model 2 uses the dynamic regression model (DRM) concept. Support vector regression (SVR), which is reported in the literature to outperform several other existing methods is also applied as a benchmark to compare the performance of the developed estimation models. The proposed DLM, DRM, and SVR models were implemented for two different but regular bus routes with diversified service areas in Calgary. The experimental results showed that the proposed DLM and DRM outperformed the SVR model for both routes. It was found that the root mean square error (RMSE) decreased by 13-23% for DLM and DRM compared to the SVR model with the same inputs. The performance of the SVR model was similar to the DLM and DRM for a shorter horizon. However, DLM and DRM yielded much better estimations for a longer horizon. Considering accuracy, computational time, and ability to update the estimation based on the new bus travel time trend, the results of this study suggest that the DLM and DRM are promising real-time bus travel time estimation models, particularly for providing estimations that include the uncertainty of the estimates.
Description
Keywords
Sociology--Transportation
Citation
Rahman, M. M. (2017). Real-Time Bus Information: Users’ Perspectives and Arrival Time Estimations (Doctoral thesis, University of Calgary, Calgary, Canada). Retrieved from https://prism.ucalgary.ca. doi:10.11575/PRISM/27544