AI-Powered Solution to Optimize Passenger Experience at Bus Stops
Addressing digitalization in public transport will undoubtedly improve the service given to users by helping manage peak hours and avoid delays. It is known that crowd congestion is one of the main challenges in public transport, and more concretely when referring to bus services, congestion at bus stops. Currently, there is limited insight into the activity at these critical points. By addressing this crucial aspect, bus services could be further optimized, especially on lines with a high volume of users.
Unlike inside buses, where passenger activity is tracked through ticket validation systems and drivers can provide observational insights, bus stops typically lack the obtention of such structured information. These critical points of transit are left without data on the number of waiting passengers, their waiting times, or crowd dynamics. This absence of data makes it challenging to manage congestion, especially during peak hours, and to optimize bus arrival schedules effectively. By addressing this gap, we could bring much-needed visibility to bus stop activity, enhancing overall public transport efficiency.
Obtaining information on what is happening in urban contexts is not straightforward due to legally binding confidentiality laws. For this reason, a new project developed through collaboration of TMB and CARNET aims to obtain relevant information at bus stops, equipping them with sensors, all of them freely usable in public spaces. The main insights extracted will be: monitoring the number of people, their waiting time, and bus arrival times. With this in mind, a depth camera is installed at selected bus stops. This kind of camera registers distance/depth (giving a similar output to thermal cameras but without temperature information) and provides anonymized data (therefore, this technology is not subject to GDPR regulations). Using computer vision and AI, the data obtained with this camera enables the detection and tracking of waiting passengers and bus arrivals, providing valuable insights to optimize public transport services.
The main key tasks of the project are: testing the best position to install the camera at the bus stop and recording different setups with varying conditions such as sunlight, time of the day, number of people, or bus arrivals. The data will be annotated, counting how many people are present every N frames and ensuring no person is counted twice. For this, it is necessary to research and adapt available object detection and tracking models, which also apply to bus arrival detection.
This project was originated in the innovation department (TMBinnova) of the Transports Metropolitans de Barcelona (TMB), and implemented together with the Universitat Politècnica de Catalunya (UPC) and CARNET, reflecting the strong partnership between CARNET and Transports Metropolitans de Barcelona (TMB), Barcelona’s primary public transport operator and a key player in urban mobility. Together, they are committed to improving mobility by implementing innovative solutions into public transport systems.
