Biking better: AI-enabled personalised journey planning to enhance the experience and uptake of bike riding

Project supervisors

  • Associate Professor Ben Beck, School of Public Health and Preventive Medicine, Faculty of Medicine, Nursing and Health Sciences (main supervisor)
  • Professor Dana Kulic, Department of Electrical and Computer Systems Engineering, Department of Mechanical and Aerospace Engineering, Faculty of Engineering
  • Dr Joanne Caldwell Odgers, Department of Physiology, Faculty of Medicine, Nursing and Health Sciences

Areas of research

  • Bike riding; active transport; data science; AI; virtual reality

Project description

Cycling is an active mode of transport that confers substantial health, environmental and social benefits, particularly by promoting an active lifestyle that improves population health. Despite the profound benefits, the number of people commuting by bicycle in Australia is low (e.g. 1% of trips in Melbourne) and injury rates are on the rise. Our research has demonstrated that the key barrier to increased bike riding participation is how unsafe people feel when bike riding.

We have demonstrated that 78% of the population are interested in riding a bike, but only if infrastructure enables them to feel safe. We also know that people experience various forms of cycling infrastructure in different ways; the level of infrastructure needed to ensure people feel safe is related to an individual’s experience, perceptions, and cognitive and emotional responses. However, existing mapping tools (e.g. Google Maps) provide routing between an origin and a destination, with limited consideration of the unique needs of individual people. As a result, we lack capabilities to support people to take safe journeys by bike that reflect their specific comfort levels.

The overall objective of this PhD project is to develop an AI-enabled personalised bike journey planner to enable the identification of routes between origins and destinations that meet the unique needs of an individual. To achieve this objective, a three-phase program of research will be undertaken:

  • Phase 1: Collect quantitative and qualitative data capturing user experiences across a diversity of people in virtual reality (VR) environments;
  • Phase 2: Using the collected user experience data, with real-world traffic-level data, build a machine learning model to predict how users will experience bike riding in a given traffic scene;
  • Phase 3: Using findings from Phase 1 and Phase 2, develop an AI-enabled personalised bike journey planner.

This PhD project will leverage from and build upon existing bike riding and data science and AI programs. In Phase 1, we will use a novel on-bike data collection system to capture user experiences as people ride (funded by the Monash Data Futures Institute) and a state-of-the-art bike-specific VR system (funded by the Federal Office of Road Safety). Adult participants will be purposively sampled across age, sex and bike riding ability. Participants will ride in VR environments utilising our novel on-bike data collection system, which combines physiological (heart rate variability, galvanic skin response) and biometric (eye gaze and AI emotion detection) measures. To generate a variety of scenes for the VR dataset, we will leverage real-world bike near-miss data (collected in a related project funded by the Federal Office of Road Safety) and bicycle crash data (collected in a related project funded by the Australian Research Council). Phase 1 results will enable us to understand how people’s experiences and comfort levels vary across key demographic and bike riding ability characteristics.

In Phase 2, we will build a machine learning model to predict user responses to different infrastructure and traffic conditions based on the dataset collected in Phase 1.  The model will take as input short visual image sequences of infrastructure and traffic conditions, and output a prediction of user comfort.  We will investigate a variety of deep learning architectures, including convolutional neural networks (CNNs), recurrent networks (RNNs) and transformer architectures.  We will evaluate different training methodologies, including a population model and individualised user models, as well as considering pre-training and curriculum learning on traffic density data.  Specifically, in this phase we will develop methods to predict how users will experience a variety of infrastructure and traffic conditions across an entire city (Melbourne) based on observations in a limited subset of conditions.

In Phase 3, we will apply findings from Phase 1 and Phase 2 to develop an AI-enabled personalised bike journey planner. Specifically, an algorithm will be developed to identify the optimal route between an origin and a destination to meet the needs of individual users. The proposed approach will search for a path that optimises a combination of user comfort (as estimated from the approach in Phase 2) and route length.  To generate fast solutions, we will use rapidly exploring random trees (RRT), with edges generated to ensure minimum acceptable comfort according to the user-specific model.  A web-based platform will be developed that will firstly identify the user’s level of comfort (assessed through users selecting their perceived comfort across videos of a variety of environmental scenarios), and subsequently provide the user with their optimal route between an origin and a destination based on their specific level of comfort.

This project is fundamentally interdisciplinary and brings together world-leading programs of research across Monash University. A/Prof Beck leads an interdisciplinary and internationally-recognised team that focus on advancing sustainable mobility and safety, particularly for people who ride bikes. Prof Kulic is internationally recognised for her work in robotics and human-machine interaction, developing autonomous systems that can operate in concert with humans, developing algorithms for wearable sensor data and has deep expertise in data science and AI. Dr Caldwell leads applied physiology research and is internationally recognised for her work in exercise and thermal physiology. This PhD project brings together this unique interdisciplinary expertise of the supervisory team, and draws upon existing collaborative projects between the supervisory team members.

Collectively, this project will act as a step-change in our ability to provide personalised journey planning based on users’ individual comfort levels; an objective that has not been achieved previously. This has multiple potential benefits. We know that how unsafe someone feels when riding is the biggest barrier to increased uptake of bike riding. Therefore, providing tools to users to ensure they can select routes that meet their needs will be critical in increasing participation. Further, by routing riders away from routes that have high rates of ‘near misses’ and crashes, we may reduce the number of people who are injured. While building a smart-phone based app is likely outside the scope of this PhD, the outcomes of this project have the potential to inform numerous existing platforms and transform the way that we provide journey planning information for people who ride bikes.

PhD student role description

While it is well established that how unsafe someone feels while bike riding is the key barrier to increased participation, how road environments influence perceptions of safety across a diversity of people is not well understood. As a result, we have limited understanding of the needs of individuals, which is particularly problematic when trying to address substantial and persistent inequities in bike riding. This PhD project acts as a transformational change in our ability to support individual users to take routes that align with their comfort levels and needs. Developing an AI-enabled personalised routing planner will fundamentally shift the way that we provide directions and routes to people who ride bikes.

The student in this role will have the opportunity to drive and shape this novel and impactful research domain, the end result of which has the potential to have real world impacts on people not just in Australia, but globally. The student will have the opportunity to develop skills in applied data science and AI, laboratory and real-world experiments (including applied physiology) and geospatial modelling, drawing on the expertise of a highly interdisciplinary team of supervisors and collaborators.

In Phase 1, the student will be responsible for conducting virtual reality (VR) data collection and analysing a diversity of physiological, biometric and geospatial data. In Phase 2, the student will learn and apply novel AI methods to fuse subjective user experience data with objective measures of safety, and subsequently use these data to develop new ways of predicting perceptions of safety across entire cities. In Phase 3, the student will bring together findings from Phases 1 and 2 and develop advanced machine learning approaches to identify optimal routes between origins and destinations that are specific to the individual needs of users.

The supervisory team will foster the student’s development, from methodological, applied and leadership perspectives. The primary supervisor will meet with the student once per week, with meetings with all supervisors occurring at least once per month. The student will spend the majority of their time in A/Prof Beck’s interdisciplinary Sustainable Mobility and Safety Research team, while also spending time in Professor Kulic’s Engineering team and Dr Caldwell’s physiology laboratory. The student will benefit from the diversity of expertise of the supervisory team and their respective groups, engaging across multiple domains to maximise the development of the student and project outcomes.

Increasing the uptake of bike riding as a healthy and sustainable mode of transport is a core priority of federal, state and local governments. The skills acquired through this PhD will open many opportunities for the student wishing to pursue a career in academia or industry, particularly given the interdisciplinary experience and the ability to work with and communicate with a variety of people and organisations. The student will be provided with opportunities to engage with government and non-government organisations relevant to the project, as well as being immersed in a diversity of existing research programs. This project provides the student with a unique opportunity to sit at the forefront of applied data science and AI, with an application that has the potential to shift mobility towards active and sustainable modes of transport, thereby leading to substantial gains in population and environmental health.

Required skills and experience

  • Previous experience with machine learning / artificial intelligence methods
  • Coding experience (e.g. Python)
  • Previous experience with data collection, working with participants and VR would be desirable

Expected start date

  • March 2023

Submit your Expression of Interest