What type of infrastructures do e-scooter riders prefer? A route choice model
Introduction
E-scooter is an innovative transportation mode that meets the travel demand of many travelers by providing flexible, affordable, and accessible mobility services. E-scooters are well-suited to replace trips shorter than 2 miles, which are 36% of all trips in the U.S. (U.S. Department of Transportation & Federal Highway Administration, 2019). Currently, the majority of these short trips, i.e., approximately 66% (i.e., calculated using weighted 2017 National Households Travel Survey [NHTS] data [U.S. Department of Transportation & Federal Highway Administration, 2019]), are made by private automobiles. E-scooters have proven popular with the public, with usage increasing at a much faster rate than other forms of shared mobility (i.e., cars, bicycles) (Clewlow, 2019). In 2018, there were 85,000 e-scooters available for public use in 100 U.S. cities, resulting in 38.5 million trips (NACTO, 2018).
In response to growing demand and e-scooter usage, many states and cities have passed laws and regulations to govern the deployment of shared fleets and use of e-scooters. Since the launch of the first shared e-scooter fleet program in 2017, the majority of U.S. states have enacted laws to allow the operation of e-scooters on streets, with some restrictions, such as speed limits, no e-scooters on sidewalks, operation time, driver’s license requirements, age constraints, etc. (Unagi Scooters, 2019). Some states, such as Kentucky and Hawaii, treat e-scooters as bicycles. Many city governments, such as Arlington County, VA, Washington D.C., and most recently, New York City (Toll, 2020), have passed ordinances to locally regulate the use of e-scooters in public spaces, and launched structured pilots or begun permit programs for private sector operators to offer shared fleets of e-scooters to the public.
Local policymakers, however, face the challenge of providing safe and convenient environments to accommodate this emerging travel mode. City transportation planners seeking to adapt the current transportation system for e-scooters must develop an understanding of usage patterns and user needs. Importantly, there is a limited understanding regarding infrastructure preferences of e-scooter riders due to data limitations, especially the lack of revealed preference data. To the authors’ best knowledge, no study, to date, has been devoted to systematically modeling e-scooter user infrastructure preferences using empirical data.
E-scooter riders may have different infrastructure preferences compared with bicyclists. Bicycles are more stable and perform better on rougher roads, due to their geometry and tire size. With smaller tires and their unique geometry designs, e-scooters require a more upright steering angle, which makes them less stable on roads with rough surfaces, such as pebbles and bumps. Meanwhile, e-scooter riders may find roads with steep slopes less intimidating, as riding an e-scooter does not require physical effort. Additionally, e-scooters are also smaller and lighter than bikes, making them easier to switch between sidewalk and street. However, no prior studies have explored the e-scooter riders’ infrastructure preferences to provide insights for infrastructure planning for e-scooters.
This study develops an e-scooter route choice model using revealed preference data collected using Global Position System (GPS) units installed on e-scooters on Virginia Tech Blacksburg Campus. The model results show relative preferences for different transportation infrastructures, as well as the effect of route complexity and topography on route choice. The remainder of the paper is organized as follows. In Section 2, we review existing literature on e-scooter and bicycle route choice preferences to highlight research gaps. In Section 3, we describe the modeling techniques and data used in this study. We provide model results and interpret e-scooter riders’ preferences in transportation infrastructure in Section 4. Finally, we conclude the major contribution of this paper and discuss future research directions in Section 5.
Section snippets
Prior studies on e-scooter and bicycle route preferences
Research about e-scooters is still nascent, given this innovative travel mode has just emerged recently. Most of the existing e-scooter studies are purely descriptive (BCDOT, 2019, Chowdhury et al., 2019, Jiao and Bai, 2020, Liu et al., 2019, Noland, 2019, PBOT, 2018). E-scooter demand and mode choice studies are biased by small sample size (Degele et al., 2018, Smith and Schwieterman, 2018) or hypothetical assumptions in travel behavior (Lee et al., 2019). Zou et al. (2020) descriptively
Methodology and data
In this section, we first present the Recursive Logit (RL) model and its advantage over conventional discrete route choice models. We then discuss the revealed e-scooter preference data and network attributes used to develop e-scooter route choice models.
Results and discussion
The model is estimated using Matlab by adjusting an existing open-sourced code developed by Tien Mai (https://github.com/maitien86/RecursiveLogit.Classical.V2). There are 10,856 OSM network links and 7362 nodes on Virginia Tech’s campus. It takes approximately 3 h to finish estimating one configuration of the RL model on a computer with 8 Intel(R) Core i7 @ 3.1 GHz cores and 16G of memory.
Table 4 shows the model estimation results for variables that are significant at 90% significant level. In
Conclusion
In this study, we developed recursive logit (RL) route choice models to examine infrastructure preferences of e-scooter users using GPS trajectory data collected on Virginia Tech campus from September to October 2019. A sample of 2000 e-scooter trips was analyzed. The final model provides rich insights on e-scooter users’ preferences on different transportation infrastructures, which have not been examined before due to a lack of revealed preference data.
The results suggest e-scooter users are
Credit authorship contribution statement
Conceptualization: WZ, RB, AB, TS; Data Curation: WZ, TS; Methdology: WZ; Formal analysis: WZ, RB, AB; Wrting - original draft: WZ; Writing - review & editing: WZ, RB, AB, TS. The results and manuscript have been reviewed and approved by all authors.
Acknowledgements
The authors would like to recognize Ford Motor Company for sponsoring this research and thank Mike Mollenhauer at VTTI for his leadership. We appreciate the instructions and tutorials provided by Dr. Tian Mai, which helped running the recursive logit model in Matlab. We are also grateful that Owain James contributed to the preparation of slope variables used in this study.
References (39)
- et al.
Where do cyclists ride? A route choice model developed with revealed preference GPS data
Transport. Res. Part A: Policy Pract.
(2012) - et al.
A link based network route choice model with unrestricted choice set
Transport. Res. Part B: Methodol.
(2013) - et al.
A decomposition method for estimating recursive logit based route choice models
EURO J. Transport. Logist.
(2018) - et al.
A nested recursive logit model for route choice analysis
Transport. Res. Part B: Methodol.
(2015) - et al.
Route choice of cyclists in Zurich
Transport. Res. Part A: Policy Pract.
(2010) - et al.
Trails, lanes, or traffic: Valuing bicycle facilities with an adaptive stated preference survey
Transport. Res. Part A: Policy Pract.
(2007) - et al.
Bike route choice modeling using GPS data without choice sets of paths
Transport. Res. Part C: Emerging Technol.
(2017) - Abraham, J. E., McMillan, S., Brownlee, A. T., Hunt, J. D., 2002. Investigation of cycling sensitivities....
Environmental and travel preferences of cyclists: Bicycles and bicycle facilities
Transp. Res. Rec.
(1994)- et al.
Analysis of Bicycle Commuter Routes Using Geographic Information Systems: Implications for Bicycle Planning
Transp. Res. Rec.
(1997)
Bicyclist link evaluation: A stated-preference approach
Transport. Res. Record
Off-road but on track: Using bicycle and pedestrian trails for transportation
TR News
A GPS-based bicycle route choice model for San Francisco, California
Transport. Lett.
Cited by (68)
Perceptions of shared e-scooters service among university students in Braga, Portugal
2024, Transportation EngineeringFaster, greener, scooter? An assessment of shared e-scooter usage based on real-world driving data
2024, Transportation Research Part A: Policy and PracticeA simple agent-based model for planning for bicycling: Simulation of bicyclists' movements in urban environments
2024, Computers, Environment and Urban SystemsExploring electric scooter regulations and user perspectives: A comprehensive study in Hungary
2024, Case Studies on Transport PolicyAn E-scooter route assignment framework to improve user safety, comfort and compliance with city rules and regulations
2024, Transportation Research Part A: Policy and Practice