Synthetic Mobility Data: The Future of Smart Urban Planning

Smart cities, optimized traffic, and autonomous vehicles all depend on one crucial resource: mobility data. But how can we harness this power while protecting privacy? Synthetic mobility data offers a compelling solution by creating artificial yet statistically accurate movement patterns without tracking real individuals. Our research team at Bern University of Applied Sciences explores the potential and challenges of this transformative technology.

What are synthetic mobility data?

Synthetic data are artificially generated data, not obtained through the direct collection of real events, processes, mobility, or locations of people. These produced data are created based on algorithms, simulation of partially generative models. Why do we need them? Because the use of synthetic data is very useful when dealing with sensitive data of subjects due to the need to protect privacy and other security requirements.

Use Cases around the world

Synthetic mobility data have already proven their value worldwide.

• Singapore: For urban planning, understanding traffic patterns and pedestrian flows
• Amsterdam and Copenhagen: For optimizing bicycle route networks
• London and New York: For optimizing public transport schedules and routes
• Barcelona and Seoul: For implementing the ‘Smart City’ concept, and by companies like Waymo and Tesla for the development and testing of autonomous vehicles.
• Swiss Cities: In Zurich and Geneva, synthetic mobility data can be used for traffic management, helping to reduce congestion and improve the flow of vehicles, optimize traffic light sequences, design and plan transport junctions, etc.

The balancing challenge

Synthetic data should not contain information about real people and their behavior, but at the same time, they should be informative enough according to the specified parameters of the request. This is the main challenge and advantage in producing synthetic data: high-quality synthetic data should be indistinguishable from real data in terms of their statistical characteristics. Therefore, special attention in the generation of synthetic data is paid to preventing biases in the algorithms and models of their creation.

Avoiding potential biases

Such data are a simplification of real data, so it is important to preserve the informativeness of key indicators. There can be systematic, algorithmic, historical, demographic, geographic, technical, and temporal errors in the generation of synthetic mobility data. To avoid potential biases in the generation of synthetic data, it is necessary to ensure the relevance of the represented age and ethnic groups, geographic regions, consideration of different movement patterns and various means of transportation, seasonal and daily fluctuations in traffic intensity, and the characteristics of urban and rural areas.

Synthetic mobility data provides information on locations, direction, speed, and timing of population movement without any connection to real individuals. The algorithms or simulation models used to create synthetic data take into account factors such as typical movement patterns, transport junctions, public transport schedules, population density, etc.

Diveres generation techniques

Using synthetic mobility data provides the opportunity to test controlled scenarios of planned changes, such as the ability to forecast changes in urban traffic and people’s mobility in connection with the optimization of public transport. Undoubtedly, achieving high accuracy in such a forecast is extremely difficult, as human behavior is determined by many factors that are difficult to foresee and structure into an algorithmic scheme.

Currently known techniques for creating synthetic data include:

• Agent-Based Modeling (ABM) – simulating the actions of individual agents as representatives of an environment or community,
• Monte Carlo Simulations and Markov Models – attempts to forecast potential probabilities of system dynamics based on known indicators,
• Machine Learning Algorithms – creating synthetic data that mimic key indicators of real data,
• Network Analysis – modeling system changes based on the impact of a new influential factor,
• Microsimulation Models – modeling individual units (for example, cyclists’ movements) to predict their interaction on a small scale (district, city),
• Cellular Automata – modeling microdata taking into account a larger context of possible analysis,
• Statistical Extrapolation – using statistical methods to develop scenarios of potential changes.

Challenges and limitations

In the creation of synthetic mobility data, there are a number of difficulties. They are based on statistical factors and patterns and are able to include force majeure circumstances such as weather, special events, road works, and various social factors. Their artificial nature requires verification with real data to affirm their quality and informativeness. Generating quality synthetic mobility data is very resource-intensive, especially in agent-based modeling models. Furthermore, synthetic data can quickly lose relevance due to the dynamics of the real world: population changes, the introduction of innovations, new laws, urban planning, epidemics, use of different types of transport, etc.

Potential applications

In addition to protecting privacy and efficiency in creating scenarios for changes, synthetic mobility data are important for decision-making about resource allocation, for example, where to build new roads, locate services, or optimize public transportation. These data are important for technologies like autonomous vehicles, as they provide a safe environment for development and testing, as well as for conducting various types of research and safe innovation implementation. Potentially, the use of synthetic mobility data will save resources by reducing the need for extensive campaigns to collect real data.

Linked Project

https://www.bfh.ch/de/forschung/forschungsprojekte/2022-470-842-714/