To realise the full social and economic benefits of AI, new technologies must be sensitive to the diverse needs of the whole population. This means understanding and reflecting the complexity of individual needs, the variety of perceptions, and the constraints that might guide interaction with AI. This challenge is no more relevant than in building AI systems for older populations, where the role, potential, and outstanding challenges are all highly significant.

The RAIM (Responsible Automation for Inclusive Mobility) project will address how on-demand, electric autonomous vehicles (EAVs) might be integrated within public transport systems in the UK and Canada to meet the complex needs of older populations, resulting in improved social, economic, and health outcomes. The research integrates a multidisciplinary methodology – integrating qualitative perspectives and quantitative data analysis into AI-generated population simulations and supply optimisation. Throughout the project, there is a firm commitment to interdisciplinary interaction and learning, with researchers being drawn from urban geography, ageing population health, transport planning and engineering, and artificial intelligence.

Predicting and Meeting Needs

The RAIM project will produce a diverse set of outputs that are intended to promote change and discussion in transport policymaking and planning. As a primary goal, the project will simulate and evaluate the feasibility of an on-demand EAV system for older populations. This requires advances around the understanding and prediction of the complex interaction of physical and cognitive constraints, preferences, locations, lifestyles and mobility needs within older populations, which differs significantly from other portions of society. With these patterns of demand captured and modelled, new methods for meeting this demand through optimisation of on-demand EAVs will be required. The project will adopt a forward-looking, interdisciplinary approach to the application of AI within these research domains, including using Deep Learning to model human behaviour, Deep Reinforcement Learning to optimise the supply of EAVs, and generative modelling to estimate population distributions.

Working in the UK and Canada

A second component of the research involves exploring the potential adoption of on-demand EAVs for ageing populations within two regions of interest. The two areas of interest – Manitoba, Canada, and the West Midlands, UK – are facing the combined challenge of increasing older populations with service issues and reducing patronage on existing services for older travellers. The RAIM project has established partnerships with key local partners, including local transport authorities – Winnipeg Transit in Canada, and Transport for West Midlands in the UK – in addition to local support groups and industry bodies. These partnerships will provide insights and guidance into the feasibility of new AV-based mobility interventions, and a direct route to influencing future transport policy. As part of this work, the project will propose new approaches for assessing the economic case for transport infrastructure investment, by addressing the wider benefits of improved mobility in older populations.

At the heart of the project is a commitment to enhancing collaboration between academic communities in the UK and Canada. RAIM puts in place opportunities for cross-national learning and collaboration between partner organisations, ensuring that the challenges faced in relation to ageing mobility and AI are shared. RAIM furthermore will support the development of a next generation of researchers, through interdisciplinary mentoring, training, and networking opportunities.

Project Stages

The project will run from February 2020 to August 2023, and involve completion of six work packages:

WP1 – Generating Older Population Lifestyle and Mobility Profiles In the first work package, we will conduct a review and secondary data analysis to chart changing life situation and mobility practices of older people. The primary output will be an evidence base on these issues, which will be used to develop future lifestyle and mobility profiles as inputs for later modelling steps.

WP2 – Constraints in Older Persons Mobility and Use of AVs This work package will focus on understanding different groups’ views of possible on-demand, autonomous vehicle transportation systems. All of the consultations below will have the same primary purpose, but separate focus groups (five per task in each location Canada and the UK) will be conducted, with variations in the questions dependent on the individuals included.

WP3 – Measuring Current and Future Choices via Tracking Data The main goal of WP3 is to better understand the travel choices affecting end-to-end journeys within ageing populations at finer spatial and temporal scales. Through use of mobile phone-based surveying and tracking we will collect continuous, highly detailed and micro-level data on individuals’ movements and preferences.

WP4 – Predicting Demand through Agent-based Modelling The fourth component of RAIM involves the development of AI-based simulations to predict heterogeneous demand for an EAV DRT service given spatial, temporal, and population-level variation. WP4 will build on the data and insights collected through WPs 1-3, and interact iteratively with WP5.

WP5 – Fleet Design and Optimisation In this module, a methodology is developed to design and operate an efficient EAV-based door-to-door DRT fleet to serve ageing populations considering temporal and spatial sparsity and heterogeneity of demand, estimated through WP4.

WP6 – Establishing the Governance, Policy, and Economic Case The sixth work package will focus on establishing the economic case for an EAV DRT service for ageing communities, addressing both the costs of operation and wider benefits gained through increased mobility in ageing populations. This work package will be executed in close collaboration with our project partners, Winnipeg Transit (WT), Transport for West Midlands (TfWM), and Transportation Options for Seniors Network (TONS), with all scenarios and constraints co-designed with partners.