Dynamic timetables - a smart way to cut costs in the aftermath of Covid-19

Important results from the project

The project´s objective was to show new opportunities of, and calculate the effect of, dynamic timetables in public transport. Du to the pandemic, the project was refocues when the same input data could be used to develop an AI technology to forecast ridership and real onboard capacity with the aim of preventing onboard crowding. Results: A new AI engine that predicts on- and offboarding. It is integrated with The Train Brain-platform, is completely self-learning, and can be used on all bus traffic systems in the world that have some form of positioning and some form of customer billing.

Expected long term effects

The self-learning intelligence developed in this project gives public transport the ability to provide end customers with forecasts, the ability to prevent crowding in planning and traffic control, to understand and to forecast ridership and the opportunity to become more resource efficient by enabling balancing of supply and demand through forecasting. After only 30 days of training, the service is approximately 80% fully trained and ready to run - provided that there is passenger counting equipment on board and 12 months of historical on/offboarding data.

Approach and implementation

The project began with us securing input data in the form of timetables, vehicle positions and passenger on/offboarding (from optical sensors at the bus doors. This data has been provided by the bus operator Nobina. The input data has not been in real time, but batchwise every night. To enable real-time computation, we built a function that with that batch data simulated a real-time data stream. Based on our technology for forecasting driving times, we were able to develop, validate and test on-board forecasts of travel with the above input data.