Data from two corona years shows our walking behavior is still changing
Since the outbreak of the COVID-19 pandemic, people’s lives have changed. During the past two years, several measures were implemented to slow down the transmission and to prevent the virus from spreading. One of these measures is physical distancing, which has been proven to be an effective way to reduce the risk of transmission.
From the beginning, the implementation of physical distancing organizations raised many relevant questions. One of the main questions was: ‘How to optimize the capacity of an infrastructure while guaranteeing people can keep the required physical distance?’ InControl analyzed the impact of physical distancing on capacity during the pandemic in Pedestrian Dynamics®, by implementing a new parameter called ‘Physical Distance’. This parameter is designed for conducting capacity analyses based on the assumption that pedestrians strictly keep the prescribed physical distance.
To analyze the impact of physical distancing, it is important to investigate the changes in the walking behavior of people as a result of the introduction of prescribed physical distance. These insights can be used to develop a crowd model to gain insights into the transmission risk and the efficiency of the measures to reduce the spread of the virus. From observations, it is clear that people will not keep strict distances when this is not possible. However, further research is necessary to acquire more knowledge on the interactions between pedestrians, particularly in situations in which people are not able to comply with physical distance rules.
To investigate the interactions of pedestrians, Roberto Gonzales conducted his Master Thesis Research at InControl to complete his studies Transport and Planning at the TU Delft. His research aimed to get a better understanding of how the walking behavior of pedestrians changed as a result of physical distancing. He also investigated the capability of Pedestrian Dynamics® to predict the behavior observed in reality by conducting a calibration of the model. The research focused on bidirectional flows, which makes it difficult to keep distance even at very low flow rates.
Figure 1: Physical distancing in Pedestrian Dynamics®
Did the walking behavior of pedestrians change as a result of physical distancing?
To investigate the walking behavior an assessment of bidirectional flows was conducted by analyzing the trajectory data from a corridor at Utrecht Central station. The data was collected by embedded sensors with a new technology that allows tracking pedestrians anonymously within a predefined area. The data correspond to the years 2019, 2020, and 2021 to determine the differences in behavior before and during the pandemic and the differences throughout the pandemic.
The data showed that the demand declined in 2020 compared to 2019. In 2021, although it slightly recovered, it was still at levels below 2019, see Figure 2. Moreover, the density levels during the studied period of each year were below the critical density, which entails that congestion did not occur and thus pedestrians could walk continuously and naturally (free-flow). The analysis concludes that the relation between the flow-density and free-flow conditions has remained the same during the pandemic. Which was contrary to what was expected. As a result of the physical distancing, you would expect to see lower speeds for different densities.
Figure 2: Demand in bidirectional flow in June of 2019 2020, and 2021
Regarding the assessment of the local behavior, different metrics were used to determine the impact of physical distancing. These metrics are the minimum distance headway, the effort distribution which describes changes in pace, and the speed distribution. The most relevant conclusions from the analyses are:
- Minimum distance
Analysis of the minimum distance headway showed pedestrians kept a distance larger than 1.5m on average. It was also observed that fewer people accepted a distance lower than 1m compared to the pre-pandemic situation. In general, the prescribed physical distance of 1.5m was not fully complied with. The results of the minimum distance headway over the years are shown in Figure 3.
- Effort distribution
Analysis of the effort distribution showed different results based on the studied year. In 2020, this metric suggests there was an increase in pedestrians’ awareness and a larger variance of their pace, which implies that pedestrians deviated more often from their most direct path to keep a larger distance. In 2021, the opposite was found as the effort results showed that pedestrians would deviate less and thus a decrease in their awareness and variance of velocity could be expected. Similarly, the speed distribution showed that the speed was less evenly distributed in 2020 than in 2021 compared to the corresponding mean speed of each year. These results also suggest an increased awareness of pedestrians during the first year of the pandemic and a decrease in 2021.
- Walking behavior
The assessment of the walking behavior using all three metrics showed that the physical distancing rule has changed the interactions between pedestrians. Furthermore, this impact has also changed throughout the pandemic.
Figure 3: Changes in minimum distance headway-metric before and during the pandemic
The capability of Pedestrian Dynamics to reproduce the observed behavior
To determine the capability of reproducing the observed walking behavior during the pandemic in Pedestrian Dynamics®, the model needed to be calibrated. The calibration included parameters that describe the local behavior in Pedestrian Dynamics® and are relevant to cope with the changes caused by physical distancing, these are:
- relaxation time
- viewing angle
- personal distance.
The calibration results showed that the model better replicates the behavior during the pandemic when pedestrians are modeled as reacting faster, and changes in their movement direction occur more abruptly. This finding was in accordance with the expected greater awareness and larger variance of pedestrians’ velocity. These changes in the walking behavior were introduced in the model by decreasing the value of the relaxation time, which was found to be the most relevant parameter to be calibrated to improve the accuracy of the model.
Furthermore, the calibration showed that the optimal values of the parameters differed according to the combination of scenarios (i.e. 2020 and 2021) and metrics analyzed. The scenarios and metrics used in the calibration influenced the accuracy of the model since a decrease of the goodness of fit (GoF) of a specific combination was observed when using the optimal values of parameters corresponding to another combination. However, this decrease was very small and therefore a model calibrated for all scenarios and metrics could be used when the calibrated model is intended for general usage in COVID-19 situations.
Conclusion
In conclusion, the results show that relaxation time is the most important parameter. The error significantly changes based on its value, see figure 4. The optimal values of the parameters for a combination of all scenarios and metrics could be used for general usage in COVID-19 situations. However, provided that the accuracy of the model mostly depends on a single parameter, the model calibrated for a specific situation during the pandemic would best replicate the behavior observed in reality. This is mainly due to the significant changes in the walking behavior throughout the pandemic.
Figure 4: Variation of model’s accuracy according to changes in the relaxation time
Future developments of Crowd Management simulation
Innovation is core to our business and therefore at InControl we are continuously improving and developing our software products. This research is part of our continuous efforts to increase our knowledge of people’s walking behavior and to validate and improve the walking behavior algorithms in Pedestrian Dynamics®.
This particular research has provided insight into the changes in walking behavior during the pandemic and the impact of physical distancing. Moreover, the findings show that pedestrians have changed the way they behave throughout the pandemic. This raises the question of whether the walking behavior would be different in other periods of the pandemic were studied and how significant these changes would be in comparison to the ones found in this research.
It is important to further study the impact of the physical distancing on the choices of pedestrians in upper levels. These levels consist of activity choice, activity scheduling, and route choice. This would allow improving the model, allowing it to reproduce the behavior of pedestrians at all levels during the pandemic more accurately.
Are your students interested in a thesis project using Pedestrian Dynamics®?
There are many interesting research topics in the area of crowd management simulation. We offer students that are working on their Bachelor’s or Master’s Thesis support and a Student Pro license for an affordable price. This license comes without limitations in model size and offers the complete functionalities of our software.
Find more information here or contact our research and education team.