Agent-based modeling is a powerful tool that allows autonomous agents to interact with one another within a model, and has already been used in a variety of fields for various applications such as control systems, e-commerce, information management, health care, entertainment, and of course, transportation research (Chen & Cheng, 2010). As modern transportation and traffic applications expand in scope, their complexity requires more accurate and dynamic models than purely analytical ones (Daamen, Buisson, & Hoogendoorn, 2014). Within the scope of Transportation Research, agent-based models have been used to model both vehicle and pedestrian movement, as it allows for more nuanced movements and actions that make the simulation more realistic and accurate.
Coxton, Chandler, & Wilson (2015) used agent-based simulation to model passengers boarding and alighting suburban trains. From their simulation, they determined that this type of modeling and software creates authentic representations of crowd behavior and can be used to test the efficacy of future infrastructure designs before they are built. However, one discrepancy with their model was that when their modeled passengers were waiting for trains, they populated the waiting area with a random distribution, when in real life it had been observed that the waiting passengers tended to congregate in one area. Many models have incorporated algorithms that cause the pedestrian agents choose paths that will be shortest, quickest, and avoid colliding with other agents, (Dijkstra, Timmermans, & Jessurun, 2001) ( Kukla, Kerridge, Willis, & Hine, 2001) but few take into consideration the real life observed crowding patterns such as where pedestrians are more inclined to wait, if such a desired place exists.
Xu, Jin, and Deng (2014) used agent-based models to simulate a variety of different types of crowding and make them look as realistic as possible for use in video games. They model human-like agents for crowd evacuation, crowd formation, and pedestrian crowding, insects for swarm simulation, and cars and other vehicles for traffic simulation. They are able to mimic the complexities of real life traffic by implementing a variety of different types of vehicles and using data from actual video footage to assign driving “personalities” to the agents. The pedestrian agents used to simulate the pedestrian crowding also had individual “personalities” that helped to make the model more realistic – they had differing speeds and movements. Though these simulations were done with the focus of making video games appear more realistic, they are the most nuanced in terms of appearance and movements that the author has found, and could translate well into the field of modeling for research purposes.
Mcdonnell and Zeller (2011) used agent-based modeling to see how the implementation of BRT stations could benefit car-drivers and bus-riders alike. Factors such as exclusive bus lanes, pre-boarding ticket machines, express stops, and bus frequency are incorporated to analyze the effects of the BRT system under different conditions. They suggest that their model can aid in determining the efficacy of implementing different BRT schemes.
Fare Card Data
As more of the transit payment systems around the world become more automated with fare-cards, data is more abundant than ever on passengers’ individual routes, since their origin, transfers, and final destinations can all be stored. With this increase in available data come opportunities for analysis and application to prior models.
Studies have been conducted to analyze reliability based on automated fare-card data from systems that require passengers to validate their fare cards on both entry and exit of the system. Lee, Sun, & Erath (2012) used fare card data from the bus system in Singapore to measure the average velocity and headway distribution of the buses. This information was then used to not only determine the reliability of the system, but also to develop a tool to reorganize and optimize the bus routes to avoid bus bunching, which would decrease the waiting time and overall traveling time for passengers. In London, Uniman, Attanucci, Mishalani, & Wilson (2010) did a similar study to determine the reliability of the London Underground rail transit system. They also used a method to differentiate between recurring and nonrecurring performance, so the reliability could be measured under typical and atypical conditions.
However, even this fare-card information has its limitations, as the actual purpose of each passenger’s trip cannot be determined from this data alone. The data must be used in conjunction with surveys to figure out the purposes of the trips (Hon & Sohn, 2016). Hon and Sohn used the smart-card data in Seoul, Korea, along with a continuous hidden Markov model to impute plausible activity patterns for the passengers. They found that their estimations of these activity patterns aligned with the observed patterns.
Fare card data can provide essential information on Origin-Destination demands and flows as well, and multiple studies have used this data to show that passengers do not necessarily use a fixed route for their regular commutes (Kurauchi et al. 2014) (Nassier, Hickman, & Ma, 2015). Nassier, Hickman, & Ma used fare card data from Brisbane’s public transit network to study route choice decisions made by the passengers and develop a model to predict passengers’ decisions. It was found from this data that passengers are averse to taking routes with transfers, long initial wait times, and infrequent or unreliable timetables.
Real Time Passenger Information System
Real Time Passenger Information (RTPI) systems have become more and more prevalent in public transportation – so much so that by 2005, RTPI displays were becoming an expected part of the service (Panter, 2005). RTPI systems allow passengers to track their buses and trains in a variety of forms – website, mobile apps, and display monitors at stations and on the public transport vehicles. This allows for better understanding of the routes and options available to the passengers.
A recent survey in Edinburgh found that only 6.0% of the 613 respondents did not consult any information source for their trip, despite the fact that the majority of the respondents were familiar with the city – 86% lived or worked in the city (Fonzone, 2015). Of the different sources of information, the displays at the bus stops that showed the bus arrival times and/or routes, were the most commonly consulted. Of course, from this information it cannot be determined if the passengers actually needed the bus trackers at the stops, or if they only looked at them out of convenience and proximity to the displays. However, it has been shown that passengers tend to overestimate their wait time when they don’t have access to RTPI (Watkins et al. 2011) (Bowman & Turnquist, 1981), and that when they do have access to RTPI, their perceived wait time is equal to their actual wait time. So their access of RTPI causes them to be more satisfied with the overall service (Dziekan & Kottenhoff, 2007).
While studies have been done on how RTPI systems increase passenger satisfaction and aid passengers in making route decisions (Hickman & Wilson, 1995), not much has been done to study the effects of having a display screen at to show which buses are present, as compared to stations without these screens.
One issue with RTPI systems is that in order to implement them, they require a lot of technologies and systems that may be costly – both in monetary terms, and in terms of data storage. Hounsell, Shrestha, and Wong (2011) present a case study of London’s iBus system, one of the largest automatic vehicle location systems in the world. It has been implemented on over 8000 buses and 90 bus garages in London. Not only does this system track the buses using GPS, but it also records when the buses open and close their doors, and can interact with the street lights to decrease wait times. Not all of this information is used for the RTPI system, but RTPI is one of the main applications. As Hounsell, Shrestha, and Wong point out, systems such as the iBus in London present challenges in terms of data storage and processing requirements. Careful thought has to go into how to effectively process and store all that data, and how long to store it for.
Much has been done in terms of agent-based modeling, fare card data analysis, and analysis of the effectiveness of RTPI systems; however, when it comes to combining all three of these things to model a Bus Rapid Transit station, prior work is lacking. It can be seen that these methods work well together from prior work that combines RTPI systems with simulations (Hickman & Wilson, 1995), those that utilize both fare card data and simulations (Lee, Sun & Erath, 2012), and studies that have arrived at similar conclusions regarding route variability even though one uses fare card data (Kurauchi et al. 2014) and the other utilizes information from RTPI systems (Hickman & Wilson, 1995). Combining these three aspects will be a challenge, but it is the next step in transportation research.
Bowman, L. A., & Turnquist, M. A. (1981). Service frequency, schedule reliability and passenger wait times at transit stops. Transportation Research Part A: General, 15(6), 465-471.
Chen, B., & Cheng, H. H. (2010). A Review of the Applications of Agent Technology in Traffic and Transportation Systems. IEEE Trans. Intell. Transport. Syst. IEEE Transactions on Intelligent Transportation Systems, 11(2), 485-497.
Coxon, S., Dr., Chandler, T., Dr., & Wilson, E., Mr. (2015, September 30). Investigating commuter train boarding and alighting dispersal by contemporary agent based modelling techniques. In Australasian Transport Research Forum 2015. Retrieved from http://www.atrf.info/papers/index.aspx
Daamen, W., Buisson, C., & Hoogendoorn, S. P. (2014). Traffic simulation and data: Validation methods and applications. CRC Press.
Dijkstra, J., Timmermans, H. J., & Jessurun, A. J. (2001). A Multi-Agent Cellular Automata System for Visualising Simulated Pedestrian Activity. Theory and Practical Issues on Cellular Automata, 29-36.
Dziekan, K., & Kottenhoff, K. (2007). Dynamic at-stop real-time information displays for public transport: effects on customers. Transportation Research Part A: Policy and Practice, 41(6), 489-501.
Eboli, L., & Mazzulla, G. (2011). A methodology for evaluating transit service quality based on subjective and objective measures from the passenger’s point of view. Transport Policy, 18(1), 172-181.
Fonzone, A. (2015). What Do You Do With Your App? A Study of Bus Rider Decision-making With Real-time Passenger Information. In Transportation Research Board 94th Annual Meeting (No. 15-1100).
Han, G., & Sohn, K. (2016). Activity imputation for trip-chains elicited from smart-card data using a continuous Markov model. Transportation Research Part B, 83, 121-135. Retrieved from http://www.sciencedirect.com/science/article/pii/S0191261515002593
Hickman, M. D., & Wilson, N. H. (1995). Passenger travel time and path choice implications of real-time transit information. Transportation Research Part C: Emerging Technologies, 3(4), 211-226.
Hounsell, N., Shrestha, B., & Wong, A. (2012). Data management and applications in a world-leading bus fleet. Transportation Research Part C, 22, 76-87. Retrieved from www.elsevier.com/locate/trc.
Kukla, R., Kerridge, J., Willis, A., & Hine, J. (2001). PEDFLOW: Development of an Autonomous Agent Model of Pedestrian Flow. Transportation Research Record: Journal of the Transportation Research Board, 1774, 11-17.
Kurauchi, F., Schmöcker, J., Shimamoto, H., & Hassan, S. M. (2013). Variability of commuters’ bus line choice: An analysis of oyster card data. Public Transport Public Transp, 6(1-2), 21-34.
Lee, D. H., Sun, L., & Erath, A. (2012, April). Study of bus service reliability in Singapore using fare card data. In 12th Asia-Pacific Intelligent Transportation Forum.
Mcdonnell, S., & Zellner, M. (2011). Exploring the effectiveness of bus rapid transit a prototype agent-based model of commuting behavior. Transport Policy, 18(6), 825-835.
Nassir, N., Hickman, M., & Ma, Z. (2015, September 30). Behavioral findings from observed transit route choice strategies in the farecard data of Brisbane. In Australasian Transport Research Forum 2015. Retrieved from http://www.atrf.info/papers/index.aspx
Panter, D. (2005). Real time passenger information: where to from here?. In Smart Urban Transport Conference, 4TH, 2005, Brisbane, Queensland, Australia.
Uniman, D., Attanucci, J., Mishalani, R., & Wilson, N. (2010). Service reliability measurement using automated fare card data: application to the London Underground. Transportation Research Record: Journal of the Transportation Research Board, (2143), 92-99.
Watkins, K. E., Ferris, B., Borning, A., Rutherford, G. S., & Layton, D. (2011). Where Is My Bus? Impact of mobile real-time information on the perceived and actual wait time of transit riders. Transportation Research Part A: Policy and Practice, 45(8), 839-848.
Xu, M. -., Jiang, H., Jin, X. -., & Deng, Z. (2014). Crowd simulation and its applications: Recent advances. Journal of Computer Science and Technology, 29(5), 799-811. doi:10.1007/s11390-014-1469-y