Light Insight Trial (LiT) Research Report

There is a strong and immediate need to investigate innovative methods to ensure the safety of vulnerable road users, particularly cyclists. To address the gap of collecting useful data from cyclists and deriving insights pertaining to road safety, a collaboration project of Transport Accident Commission (TAC), See.Sense, and Deakin University has been conducted, with the support of the iMOVE CRC. The research project embarks on an Australia-first trial of the smart light technology from See.Sense. This technology automatically captures ride data from cyclists through a smartphone application software, aiming to understanding cycling behaviour and deriving road safety insights. As the research partner in this project (denoted as LiT trial), Deakin University has been entrusted to lead research activities, which are focused on three key objectives, namely: i. to develop a data storage and management system suitable for the LiT trial, ii. to examine approaches for data analysis to consider individual privacy and security while still providing specific road safety insights, and iii. to develop methods for data analysis as well as a reporting tool for effective communication of information and insights derived from the collected data. This report presents a comprehensive description of the activities conducted by the research team at Deakin University. The overall methodology undertaken during the course of this project is described. Various data sets and data types used for analysis, with the storage and management processes are explained. Specifically, the research leverages two major sources of data, i.e., data obtained from See.Sense through the bicycle lights and software App, as well as datasets from the Victorian Government open data platform (data.vic.gov.au) that provide contextual information to those obtained from the See.Sense lights. These include Principal Bicycle Network, Strategic Cycling Corridors, road networks, speed zones, and clearways. Topographical data set (i.e., elevation) on road networks as well as police-reported crashes, and Super Tuesday and Super Sunday data of cycle counts are also utilised. With respect to data privacy and security, a number of methods have been investigated to ascertain what is required to ensure data privacy without significantly losing the ability to derive road safety insights from the collected data. A five-step privacy protection methodology was developed and applied to the project data which cover privacy zone set by a user, pseudonymisation (removing and replacing obvious identifiers), trajectory analysis procedures for anonymising data, grouping cycling trajectories and aggregating data by using the k-anonymity technique for visualisation purposes. To further protect privacy and security of the data, a four-level role-based data access mechanism has also been implemented. For conducting analysis and visualisation, a dashboard (denoted as LiT dashboard) has been designed and developed, which incorporates data analytics functionalities for obtaining travel behaviour and road safety insights. The LiT dashboard allows users to analyse the project data (e.g., number of trips, trip timing, speed, braking, swerving, road surface quality) using statistical measures and distributions, visualised on the dashboard as well as by exporting summarised data and results for further analysis. Users have various flexibility in selection of data and reporting of analysis results in the dashboard, such as analysing data from a specific time period, for a large or small area (e.g., an LGA or multiple LGAs or a small area within an LGA), for a road segment or multiple segments, for types of roads and bicycle infrastructures, etc. Importantly, to demonstrate how the data and the dashboard can help answer various research questions, a total of eight use cases co-designed with the project partners have been presented. The associated data and results presented for the use-case demonstrations are useful for understanding the functionality of the dashboard. As the data and results presented in the use-case demonstrations come from a short period of data collection (2-3 months) and from a selected group of cyclists (fewer than 900 cyclists), the obtained data and results should be carefully interpreted for understanding safety trends or making any decisions pertaining to safety policy-making. As the dashboard and project data showed strong capabilities to generate road safety insights, future use of the project data obtained from more cyclists and a longer time-period would allow deriving conclusions about safety trends and insights.