Making Integrated Mobility a Reality
GIRO has close to 40 years’ experience in bringing to market new software solutions for optimizing public transportation, most recently in the fields of planning and mobility on demand. There are always new challenges to keep us stimulated, but there is one new technological opportunity that has got us truly excited: artificial intelligence (AI).
We have the advantage of being based in Montréal, which is expanding rapidly as a global hub for AI research. GIRO recently entered into a partnership with the Institute for Data Valorization (IVADO) to conduct research projects based on operations research and deep learning. IVADO brings together professionals from various industries with academic researchers to develop leading-edge expertise in operations research, data science and AI.
But what does AI mean for public transportation in real terms?
Analyzing and understanding big data will be key to the progression and improvement of public transportation. Vast amounts of data are being generated by automatic vehicle location, automatic passenger counting and automatic fare collection systems. Data also flows from newer sources, such as mobile devices or electric vehicles’ battery state-of-charge monitoring systems, and this also has great potential. The use of sophisticated big data algorithms could provide remarkable gains in efficiency, cost reduction and robustness, further improving punctuality and aligning service better with demand.
Deep learning is a subset of AI that can allow software to train itself to perform tasks—thereby enhancing its optimization capabilities. Research projects are already underway, and others are being planned to incorporate components of deep learning into optimization tools developed by GIRO.
Deep learning shows immense promise as a way of realizing benefits, not only for service planning, but also for transit scheduling and operations. We have started with research on a crew-scheduling optimizer, exploring how deep learning can improve the optimizer’s performance. In effect, the optimizer will teach itself which strategies perform best with particular datasets and produce the lowest-cost solutions.
Public transportation agencies have also identified forecasting run times and ridership volumes as key to optimizing on-time performance and meeting service demand. There is huge scope to apply deep learning to be able to predict travel times and travel demand so agencies can adapt scheduling to those predictions. Capabilities of reacting in real time to disruptions in planning, such as driver no-shows or traffic congestion, may also be greatly improved by applying deep learning to historical data on the events that cause those disruptions.
Overall, AI has the potential to accelerate transforming the mobility landscape and making integrated mobility a reality, opening up truly exciting perspectives in public transportation.
Source: Passenger Transport (http://passengertransport.apta.com/)