Providing real-time information about how many people are on our trains across the fleet empowers customers with information to better plan their trip.
If customers know ahead of time how full train carriages are, they can make an informed decision to board a different carriage or take another service. It can also be used by employees to help move customers along the platform as a dwell management tactic.
While we are currently providing carriage load information to our customers for Waratah train sets via travel apps and indicator boards, providing the same information for other sets has been challenging.
A project led by Director, Customer Information Channels Anna Pockley and Senior Manager, Planning Analysis & Modelling Ruimin Li is having success in solving the curly problem and will allow us to provide carriage load data for our entire current fleet.
The project is a collaborative effort between Customer Operations, Strategy, Planning & Investment, Sydney Trains IT and Transport for New South Wales and started as a proof of concept in early 2020.
Anna explains that the challenge is the technology – such as weight sensors and special cameras – is not available on the Millennium, Tangara and C and K sets to tell us how many customers are travelling in each carriage.
Ruimin and her team partnered with the University of Technology, to create a machine and predictive learning model to predict carriage load for non-Waratah sets.
By taking into consideration factors such as carriage loads of Waratah trains travelling the same route, around the same time and carriage loads of Waratah trains that have run the same course in the past, predictions can be made on the carriage occupancy of non-Waratah trains.
This project is the first of its kind for Transport for NSW and has led the team to find solutions to problems that the project was not intending to solve.
“With the emergence of COVID-19 the dataset became even more important. We needed to know how many people were catching services and how to help encourage customers to physically distance. We also had Transport for NSW eager to leverage the work for the smart notification project. Our data is now seen as the most accurate measure for occupancy levels by train carriage,” said Anna.
Ruimin and her team extended the original proof of concept to develop a solution for the platform crowding.
The team found access to more accurate train carriage load data across all sets can also help understand how many people may be on station platforms. Data on platform occupancy can then help station employees, where necessary, assist with crowd management.
“The great thing about this project is that it can throw up wonderful surprises. Connecting the data and joining the dots in one area may lead to a data set that can help solve problem in a related area,” said Ruimin.
Anna agrees with this sentiment: “The project is really exciting for Sydney Trains. It’s creating a new data foundation that will continue to be extended and facilitate real-time, data-led decision making.”
The team is currently planning the delivery of the project and is using an agile development methodology to be flexible in addressing problems and solutions as they arise.