In the research project APT-R, we will develop new methods and tools making it easier for public transport authorities to leverage insights from sensor data. Daily, public transport vehicles collect large volumes of automated sensor data. The collection of such data is what enables real-time information to passengers. However, the automated public transport data are also archived by the public transport authorities, and this opens for the use of machine learning to build knowledge and make predictions.
Public transport plays a key role in a sustainable transport system. To reduce GHG emissions and local air pollution it is a goal that more people travel by public transport. An effect of the Covid-19 pandemic is, on the other hand, that fewer people use public transport, and the road traffic has increased to a higher level than before the pandemic. How will this continue in the future? To make better plans, adapt to change, and deliver attractive and effective services to the public, it is crucial that public transport authorities utilize the potential that lies in large scale automated public transport datasets.
The project will bring together researchers from statistics, machine learning, and transport, with actors from the public transport sector, to develop novel methods giving public transport authorities new and relevant insights into travel patterns and system performance. We will develop a tool, an open-source software library in the programming language R, with methods tailored to automated public transport data. These datasets have both a spatial and temporal dimension. Every stop a bus (or light rail, metro, train, etc.) makes generates a new data record with the number of passengers getting on and off, and the actual arrival and departure times. This yields many opportunities for detailed analyses and using modern statistical methods for space-time data we will model travel patterns, flow, and delays.
The key role of public transport in a sustainable transport system is explicitly acknowledged in UN SDG 11.2 and in the National Climate Plan. The national zero-growth goal state that all growth in urban passenger transport should be by the modes walking, cycling and public transport. Last year, the COVID-19 pandemic put an abrupt stop to the steady increase in public transport trips seen over the past decade. This has been inevitable when the recommendation has been to avoid public transport if possible. The pertinent question, however, is whether people will fully return to the public transport system after the pandemic, and if, and in what ways, mobility demand and travel behaviour have changed. In the years to come, data-driven learning, monitoring, planning, and prediction will be important for the public transport sector to balance supply with demand and operate efficient and attractive services for the citizens.
Today, most public transport authorities (PTAs) collect large-scale automated passenger counts and vehicle location data. APC-AVL data enables real-time information to the passengers, but the detailed sensor data also provide insights into actual travel behaviour and system performance. Archived APC-AVL data has a great and untapped potential to help PTAs build evidence-based grounds for their planning and operation decisions.
The APT-R project will bring together researchers focusing on the development of machine learning methods, with applied transport researchers and practitioners from the domain of public transport to explore new methods and models for leveraging insights from APC-AVL data. The project will develop novel machine learning methods for large-scale space-time network data and build an open-source R-package tailored for APC-AVL data. The R-package will serve as a toolkit for professionals from the public transport sector and be a platform for research and innovation in the fields of transport, statistics, and machine learning.