Analytics and optimization in public transport

T-Systems offers an innovative solution for more accurate and stable arrival and departure predictions. We leverage diverse data sources and employ advanced supervised and unsupervised learning techniques on scalable open-source platforms, driven by artificial intelligence and machine learning. Our central concept involves accumulating incremental delay forecasts along a train’s journey using various models. This solution guarantees 99.9% availability, extensible forecasting algorithms, and high performance, thanks to specialized nodes (GPU) and ample bandwidth. With an on-demand SaaS model and a Fast Lambda Architecture, we’ve developed a highly scalable platform that balances real-time requirements and extensive dataset modeling. Our vision is to deliver value, prioritize customers and quality, and enhance the forecasting process for the benefit of travelers.

Unexpected events can cause delayed trains. But the stress associated with getting connections and finding alternative travel options can be avoided for both passengers and rail operators. The forecasting machine (“Prognoseautomat”) calculates the arrival and departure times of trains 90 minutes in advance and updates them every minute. Machine learning and artificial intelligence work together in automated forecasting. Data from a range of sources are used in the calculations. The result? Accurate predictions which were unattainable earlier. With T-Systems and AI, gain a competitive advantage in the industry. 

T-Systems embarked on an endeavor to optimize connections between various German cities through a comprehensive catchment area analysis (the process of using location data and catchment areas to analyze where passengers or users are coming from). Our solution involved a meticulous examination of the pertinent catchment areas for bus connections, ultimately identifying the top 20 areas of origin for the Stuttgart–Hamburg route during the summer evaluation. To accomplish this, we leveraged data on source-destination relationships for communities, departure times categorized into 5-hour intervals, and assessments for five different day types. Furthermore, we conducted analyses based on two distinct periods, encompassing both summer and winter data, and factored in travel times between cities. This data-driven approach not only allowed us to quantify the real catchment areas but also provided invaluable insights for tailoring targeted marketing strategies, thereby enhancing the overall travel experience and efficiency for passengers.

Camera image analytics has various applications in public transport. Our machine learning algorithms are capable of detecting situations captured by cameras in vehicles and stations, e.g. passenger count, seat utilization, hazard detection, vandalism, counting objects, and capturing and analyzing movement patterns. The evaluation is carried out anonymously and directly at the source, so video files are not stored or forwarded. Furthermore, existing camera infrastructure can be used. Hence, our solution is cost-efficient and compliant with the requirements of our customers and their respective governments.

In the ever-evolving landscape of transportation, the integration of live dashboards has been a game-changer. These dashboards deliver real-time insights into traffic conditions, aiding in the analysis of route optimization, particularly for public transit systems like buses. Employees within the public transit sector, both on the bus and off-duty, can now receive mobile notifications regarding current issues, enabling more efficient transit operations. What's truly remarkable is that these live dashboards can be accessed via mobile devices, even outside the central hub. Our IT solutions empower other companies to implement similar systems, not only enhancing transit operations but also contributing to reduced emissions by optimizing routes and fostering more sustainable mobility practices.