The understanding and prediction of travellers’ mode choices is crucial not only for the effective management of multimodal transport networks, but also successful implementation of new transport schemes and policies. Traditional studies on mode choices typically treat travellers’ decision-making processes as planned behaviour. However, this approach is now challenged by the widely distributed, multi-sourced, real-time travel information, and smartphone applications especially in the presence of a variety of available mode options in dense urban areas. Within this dynamic context factors like the availability of shared vehicles, real-time passenger information, and unexpected disruptions can affect travellers’ mode choices. This paper presents a novel behaviour concept (here called “hypermode”), which captures the adaptive behaviour in mode choice in response to real-time events. A pilot survey demonstrates the validity of the proposed approach in both recurrent and occasional trip scenarios.
1. Introduction and background
Travellers’ behaviour, in particular in mode choice, has become an important area of interest for not only researchers, but also local government and industry, as it informs sustainable policies and planning.
The underlying assumption of most existing studies on mode choice is that a traveller chooses a specific mode before commencing his/her trip, which is categorized as planned behaviour.
However, due to technology development and in response to travellers’ demand for flexible transport solutions, we are seeing a change in the mobility landscape. For example, travellers can now use a variety of smartphone applications to make informed mode choices based on the time of arrival of the next buses/trains, or the availability of shared bikes at docking stations in real-time. Some earlier research studies have also identified the influence of weather conditions on mode choice, e.g.: Sabir et al., 2009; Saneinejad et al., 2012. These real-time events (weather conditions, real-time information) may lead to a dynamic process where travellers re-consider their mode choice when starting their trip. This study investigates such adaptive behaviour, to accurately predict travellers’ mode choices within this dynamic environment. The proposed approach is called here “hypermode” and is discussed in the next section.
2. The hypermode concept
This section presents the hypermode concept, which entails adaptive decision making processes in the mode choice. The hypermode concept is analogous to the hyperpath concept proposed for the route choice in public transit assignment (Nguyen and Pallottino 1988), which suggests that a traveller first identifies a set of attractive lines that connect the origin-destination (O-D) pairs. Then, he/she chooses a specific service according to a certain strategy, which can be based on the minimization of travel/waiting time, amount of walking, or number of transfers. In an analogous way, the hypermode approach proposes a decision making process articulated at two levels:
- The user identifies a set of feasible travel modes for the trip, which are accessible at the same physical location or nearby. At this level, the decision making is planned (i.e. not real-time), and is affected by static characteristics such as user preferences, socio-economic characteristics, average/historical travel times, and financial costs of using different modes.
- When the trip is about to start, the user evaluates real-time events in order to select a specific mode of transport from the aforementioned feasible set. The real-time events include: availability of vehicles (relevant to shared modes), weather conditions (relevant to walking and biking), vehicle arrival time information (relevant to scheduled or unscheduled public transport), and disruption or crowdedness on public transport.
Such adaptive travel behaviour is suitable for dense urban areas, where plenty of mode options and access points are available to travellers, and walking is always an option especially for short trips. Given that 50% of the trips in urban areas in Europe are shorter than 5 km (The European Commission 1999), the hypermode concept enjoys wide empirical support. The “planned behaviour” (i.e. the situation where travellers do not revise their initial mode choice based on real-time events) becomes one case of this more comprehensive approach and it refers mainly to travellers using their own vehicle, e.g. car or bike.
Figure 1 illustrates, in further detail, the individual components of the decision making process with inputs and outputs of the two levels of choices.
The hypermode concept is illustrated here using a real-world example. The area of interest is part of South Kensington in London (Figure 2). A traveller first identifies all the modes compatible with its destination D (feasible set), which are all accessible in the vicinity of the origin O. These feasible modes are ranked by the user according to his/her own preferences, for example bike-sharing and bus (Level 1 of the choice process).
Then the user checks the availability of shared bikes at the nearest docking station and no bikes are available or it starts raining; so he decides to take the bus instead (Level 2 of the choice process).
It is possible that the repetitive occurring of a negative real-time event on a day-to-day basis may lead to the exclusion of a mode from the feasible set. For example, if a user constantly finds the bike-sharing station empty at a certain time, he/she may exclude bike-sharing as one of the feasible modes in his/her planned behaviour.
To demonstrate the validity of such adaptive behaviour, we conducted a pilot survey, which is presented in the next section.
A pilot survey has been undertaken to explore the validity of the proposed hypermode concept. 50 respondents have been interviewed at Imperial College London. The sample includes academic, technicians and administrative staff as well as students, to ensure that behaviour in different user categories is captured. The respondents have been interviewed face-to-face to ensure a comprehensive and in-depth understanding of their decision-making processes. They were presented with two different scenarios:
SCENARIO 1) The regular commuting trip home from the College at the end of the day, which is a Revealed Preference scenario. The origin is the same for all respondents, but the destinations vary in a wide range, with some at walking distance and others outside of London (in this case we used the railway station in London as destination).
SCENARIO 2) A hypothetical trip from the College to Sloane Square (a shopping destination 2.1 km away from the origin, shown in Figure 2) at the end of the working day. This is a Stated Preference scenario.
In the first scenario we asked the respondent to describe the decision making process that shortlists the possible mode options or leads to a specific mode choice. Afterwards we asked if any of the following real-time events may affect their final mode choice:
- Real-time bus arrival time
- Bike availability at docking stations for bike-sharing service
- Disruptions on the tube
- Other, specify.
If the respondent’s explanation of the decision making process at the open question is in line with the adaptive behaviour (i.e. answering “yes” to any of the above real-time events), then this user behaviour is associated to hypermode.
In the Stated Preference scenario (Scenario 2) the respondents are presented with the mode options in Figure 2 with given average costs and travel times. Two different types of trips are considered, since trip purpose is likely to be an influencing factor of mode choice:
- Leisure (e.g. shopping, visiting friends)
- Important appointment (on-time arrival is needed).
The user is asked what his preferred mode option would be in the described scenario. Then, depending on the preferred mode, a range of real-time events are presented to the respondent, (e.g. overcrowding at the tube station) and we asked if he would choose alternative modes.
Table 1 shows the percentages of respondents associated with the hypermode behaviour with 95% confidence interval. Here, the category “Either scenario” accounts for those who show the hypermode behaviour in either Scenario 1 or 2.
The results of the survey show that the vast majority of the respondents follow an adaptive behaviour, which is in line with the hypermode concept. This is more evident in Scenario 2 and could be explained by the fact that the respondents were less likely to abandon their preferred mode for their regular commuting trip in the first scenario due to extensive learning based on past experience. The hypermode behaviour is also more evident for trips with important appointment due to the urge to reach the destination on time.
A more detailed analysis of Scenario 2 identifies the pattern of mode switches under the two different trip purposes. The results are reported in Table 2.
Table 2 partially illustrates the relevance of users’ adaptive behaviour to the management of multimodal transport networks. For example, when there is a interruption/delay of tube service, 39% of travellers will switch to other modes, possibly at nearby access points. Such information is crucial for planning service interruption at tube stations: the transport operators need to take into account the increase in demand for other modes in the vicinity of the tube station and set up measures to control overcrowding.
This study is relevant to a wide audience within the transport community:
- Researchers: to explore different modelling approaches able to provide an analytical formulation of the hypermode concept;
- Industry: to include the adaptive behaviour in modelling software for more accurate prediction and to inform requirements of smartphone applications;
- Local Government: to support accurate mode choice predictions, hence inform policies, manage effectively multimodal transport networks in case of disruptions and encourage switching towards sustainable modes via mobility platforms (Mobility as a Service).
This study discusses an innovative approach to capture the adaptive behaviour of travellers in mode choice in response to real-time events such as real-time information (e.g. next bus arrival time or availability of a shared bike) and weather conditions.
This study also presents a survey results to demonstrate the validity of the approach and outlines the implications of this research.
We would like to thank the staff at Imperial College London for their participation to the survey.
Nguyen, S., and S. Pallottino. 1988. “Equilibrium traffic assignment for large scale transit networks.” European Journal of Operational Research 37 (2):176-86. doi: http://dx.doi.org/10.1016/0377-2217(88)90327-X.
Sabir M., Koetse M. J., and P. Rietveld. 2009. “The Impact of Weather Conditions on Mode Choice: Empirical Evidence for the Netherlands.” In Proceedings of the 17th Annual conference EAERE, June 2009, Amsterdam.
Saneinejad, Sheyda, Matthew J. Roorda, and Christopher Kennedy. 2012. “Modelling the impact of weather conditions on active transportation travel behaviour.” Transportation Research Part D: Transport and Environment 17 (2):129-37. doi: http://dx.doi.org/10.1016/j.trd.2011.09.005.
The European Commission, DG XI — Environment. 1999. “Cycling: the way ahead for towns and cities.”