MOBILEguide: advanced solution for reliable occupancy predictions

Information about how full a bus or train will be allows passengers to adjust their travel plans accordingly and, if necessary, switch to a less busy connection. This requires reliable real-time information on vehicle occupancy, as provided by INIT’s occupancy rate management and passenger guidance system, MOBILEguide.

Occupancy predictions enable a more even distribution of passengers in bus and rail transport and ultimately ensure faster changeover times, improved punctuality and greater efficiency. The technological challenge lies in providing a solid database to determine a vehicle’s occupancy rate. For this reason, MOBILEguide, INIT’s occupancy rate management and passenger guidance system, is based on state-of-the-art technologies, sophisticated algorithms and on cutting-edge system architecture.

The passenger count data is stored in a Kafka data broker, which was developed to store and process data streams and is characterised by high scalability, availability and IT security. The Kafka data broker acts as a data hub for all subsystems involved, for example INIT’s data evaluation and statistics system MOBILEstatistics.

Integration of additional data sources

Additional data and systems can be easily integrated to improve the data base – for example, if not all vehicles are equipped with counting sensors. Other possible data sources for passenger counting include Wi-Fi and Bluetooth signals from smartphones, passenger information platforms and the Mobilithek (formerly mCloud) of the German Federal Ministry for Digital and Transport, an open data portal that makes mobility, geo and weather data available. The research project Mobile Data Fusion, in which INIT is cooperating with several partners, is currently working on a fusion of these data sources, which have so far only been used separately.

Current occupancy rate and predicted occupancy rate

In order to determine the current occupancy rate, the transmitted boarding and alighting figures are processed in the data broker after each departure from a stop. This means that they are first matched with schedule data and then checked for plausibility. This provides an actual number of passengers in a vehicle – the ideal basis for further calculations.

In order to calculate a future occupancy rate, the INIT solution uses a self-learning algorithm.

In the background system, the current occupancy rate is correlated with typical boarding and alighting behaviour at the next stop which is obtained from historical data. In this way, the number of expected drop-offs is also taken into account. In a unique, patented procedure, the occupancy rate at a specific stop is predicted with the help of AI elements – after deduction of the expected drop-offs. In this way, the INIT solution is vastly superior to conventional systems in terms of reliability.

INIT’s occupancy rate management and passenger guidance system uses real-time occupancy rates in combination with historical data to calculate the expected occupancy rate.

Historical passenger count data based on clusters

The reliability of occupancy rate predictions is also significantly improved by the way INIT processes the count data of boarding and alighting passengers. Since a fleet’s vehicles are not usually all equipped with counting sensors, the gap created by missing data is closed using extrapolations, hence providing "complete" historical data. For this purpose, the collected data is summarised in clusters. Each cluster consists of data from journeys that are similar in terms of type of day and time of day, for example. Hence, only data is summarised that was collected under similar conditions. That allows conclusions about typical boarding and alighting behaviour. Data processing in INIT’s statistics system, MOBILEstatistics is therefore extremely precise and fully complies with the requirements of VDV (Association of German transport companies) publication 457 on Automatic Passenger Counting Systems.

Occupancy level predictions for passenger information

The data broker makes information about the expected occupancy rate available to other systems, for example, to passenger information channels like stop displays, apps or transport companies’ websites. By sharing occupancy predictions, transport companies can offer their passengers a new service that enables them to increase passenger comfort while also improving punctuality and efficiency. INIT offers a solution that is both technically sophisticated and expandable.


Iwan Wiens

Head of Department