Fig 1.
Surface level of the train station Münchner Freiheit in Munich [30].
We took the snapshot from the Vadere simulator’s graphical user interface. We used OpenStreetMap to understand the topography of the train station and then recreated this for the simulation in the Vadere simulator’s graphical user interface. Three main routes lead to the trains of which the vast majority of people use the short route. In our simulation study we assume that up to 300 fans per minute arrive at the station which makes them experience congestion on their way to the train.
Table 1.
Age distribution of two populations: Football fans (1), and students and faculty associates (2).
Fig 2.
Experimental between-subject design.
We obtained eight message levels by combining three optional message components: congestion information, top down view, and team spirit. For a detailed view of the message components see Fig 3.
Fig 3.
Message design with all possible message components.
Right: each message design contains the sentence ‘Please use this route […]’ and a picture with an arrow pointing to the entrance of the recommended route. The components top down view and team spirit (‘Let’s support […]’) are optional. For message designs without congestion information, only the screen on the right is displayed to the participant. For designs with congestion information included, both screens are displayed. We imagine that in a real-life application, the left screen is what the app user sees first, followed by the second screen that appears when clicking on “Which route should I take?” (left). In the survey, we placed the screens next to each other.
Fig 4.
Attractiveness of routes expressed through the Likert scale value assigned by participants.
The error bars around the mean values (dots) represent the standard deviations. Prior to information, students and faculty associates as well as fans had a clear route preference: they favored the short route (top right) over the medium route (top center) followed by the long route (top left). With information provision, the short route became less attractive (bottom right) compared to the setting without information (top right), while the long route became more attractive (left).
Table 2.
Attractiveness of long route expressed through scores on a 5 point Likert scale.
Table 3.
Effect of adding congestion information to message designs on the route attractiveness for students and faculty associates.
Table 4.
Effect of adding congestion information to message designs on the route attractiveness for football fans.
Table 5.
Effect of adding a top down view or team spirit to message designs for football fans.
Table 6.
Route choice proportions after receiving information.
Fig 5.
Simplified scenario with three routes of different lengths.
They correspond to the three most important routes to the train platform at the train station Münchner Freiheit. Virtual pedestrians are generated in the (green) source on the left and walk to the (orange) destinations on the end of each corridor where they disappear. We imagine that they would enter the platform and then board trains. Without information about half of the virtual pedestrians take the short route. When the recommendation system is switched on it steers the crowd to the longest route within the information area whenever the density is the lowest within the long route.
Fig 6.
Overview of simulators and models that we use for the traffic simulation of the Münchner Freiheit scenario (see Fig 5).
We test 18 (= 2+2x8) parameter combinations: for each of the two groups, we simulate the crowd flow prior to information and for the 8 message designs. Every 2s, the re-routing logic receives density measurements from the crowd simulation. If the long route is less congested than the others, a route recommendation is sent to the crowd model (Vadere simulator). The route recommendation is perceived, processed and finally changes the route distribution. Therefore the mobility behavior (Optimal Steps Model) changes.
Fig 7.
Crowd density over time along the short route for the fans prior to information.
A steady state is reached after 250s. In this study, we looked at the steady state only, thus, neglecting any measurement data taken < 250s.
Fig 8.
Boxplots of densities along the short route for the different message designs.
The 25% and 75%-quartiles are represented by the boxes. The thick black line in the box represents the median value. Whiskers do not extend up and down from the box more than 1.5 times the interquartile range (75%-quartile—25%-quartile). Values outside the whiskers are considered as outliers (gray dots). Prior to information, the densities were the highest: the medians were above 1.2ped/m2 (fans: ▪) and some outliers were even above 1.5ped/m2. When a route recommendation was provided, the congestion was alleviated: the densities were lower compared to the setting prior to information. If the message depicts congestion information (boxplots 6–9), the densities were smaller than for message designs in which this component was missing (boxplots 2–5). We observed the largest difference for the fans when we provided congestion information in addition to the arrow and team spirit (compare ♦ and ★).
Fig 9.
Screenshot of a simulation scenario with virtual fans to showcase the impact of route recommendations on pedestrian traffic.
The virtual pedestrians distribute among the routes according to probabilities derived from the route preferences that football fans reported in the survey. The snapshot is taken after 300s. Prior to receiving route suggestions, most football fans take the short route which leads to higher densities along that route (▪). With the guidance system switched on, the virtual fans receive a message recommending the long route when the other routes are traveled more frequently. The message designs for screenshots ♦ and ★ include the components arrow pointing to the long route and appeal to the team spirit. In ★, information on the congestion situation is added. Both designs resolve the congestion in the short route, but congestion information (★) makes even more people choose the long route.