Logging Data From E-Scooters To Improve Traffic Safety

Abstract

The transport sector has seen a major overhaul in recent years with the emergence of several new forms of transport. The increasing environmental concerns, fuel prices and traffic congestion have had people recon substitute means of transportation such as electric cars, e-bikes, e-mopeds etc. E-scooters due to the rapid growth of the scooter-sharing system have emerged as an attractive solution. These motorized scooters also help mitigate the first/last mile issue associated with the use of public transport. However, the introduction of any new mode of transport leads to new conflicts which thereby can result in new types of crashes. While the regulations in several parts of the world treat the e-scooters similar to bicycles, numerous studies have proven several dissimilarities in the overall dynamics between them. To understand the causes of conflicts, data about the usage of e-scooters is essential. Naturalistic data is data collected using instrumented vehicles in road traffic, by users performing their day-to-day activities. These datasets are less likely to suffer from bias as compared to data collected in a regulated environment such as a test track. Naturalistic data are a widely recognized source to analyze and model the behavior of road users, to improve traffic safety. Naturalistic data collected on e-scooters will provide a unique understanding of the details of e-scooterist riding behavior, interactions with the surrounding road users, and reactions in different situations. This thesis aims at developing a ride data logger which will enable naturalistic data collection when extended to a large fleet of e-scooters. In addition, preliminary data collection and analysis are conducted to test the effectiveness of the logger and identify the potential issues. Data corresponding to 15 variables is logged at a frequency of 10 Hz from a plethora of sensors on the e-scooter, providing detailed information about the kinematics of the ride. In addition, video data is captured which provides a visual information of the riding environment and the rider. When the logger is connected to the internet the data stored in the local memory storage of the data logger is automatically transferred to the cloud storage. This not only automates the process but also minimizes human dependencies and interventions. To facilitate the analysis of the datasets collected as a part of the process, a Python-based graphical user interface, which allows visualization of each datapoint, has been developed. This help in analyzing the kinematics and video data simultaneously. A pilot data collection involving participants has been carried out based on which a preliminary data analysis has been conducted to indicate the potential of the new data logger to serve for a large-scale naturalistic riding study. Exposing the prototype to different road surfaces, lighting conditions, riding styles, and ride durations over the span of 350 km has facilitated the identification of points of improvement for a furture large-fleet data collection.