Cities and their citizens can reduce urban CO2 emissions through rules and incentives but also by improving user awareness and choice. C02TRAFFICAI uses Artificial Intelligence to transform sustainable urban mobility and enable policy-driven Mobility-as-a-Service (MaaS). It combines a cloud-native platform, IoT sensor/data management, and artificial intelligence to redefine how CO2 emissions are assessed and how city stakeholders use this new knowledge.

Mobility scenarios (what-if and counterfactual) are assessed using probabilistic inference, to decide on, e.g., policies, their implementation and engagement campaigns; automated and continuous machine learning will support system performance refinement over time.
Citizens and Public Authorities will gain awareness about the emission impact of their choices and become empowered with the most effective and sustainable options to experience urban mobility and usher in the carbon-neutral future we all need.

A user story showing C02TRAFFICAI vision and scope, summing up value proposition and main capabilities could be:

Bénédicte is a professional who needs to travel in the city of Paris with non-regular travel behaviour: sometimes she goes to the office but it may happen that, during the day, she must reach different areas of the city, not always well served by Public Transport. For this reason, Bénédicte usually travels by car, according to her daily plan. In addition to fuel and parking costs, she pays the fixed charges by buying the LEZ stickers issued by the municipality.

Thanks to her MaaS subscription, Bénédicte has a monthly 120 € subscription that offers her: 5 taxi rides, 4 daily transfers by subway, unlimited use of buses, 10 carsharing rides and the Crit’Air LEZ sticker according to euro category of her vehicle, granting 10 access to Low Emission Zone.

Nevertheless, the Paris municipality has no control on the daily behaviour of all MaaS users that, in certain days where pollution turns to be higher than expected, may opt, on a larger share, to have single-car trips that are suggested as the most individually convenient by their MaaS journey planners. The C02TRAFFICAI system of the Paris municipality identifies the days when CO2 emissions are higher than the acceptable threshold by estimating the evolution of CO2 according to traffic levels and weather conditions, while also assessing previous similar situations; these simulations lead the municipality to decide for Active Demand Management measures: a message is sent to the MaaS operator to lower the priority of single-car trip and, in certain conditions, even to apply surcharges to Crit’Air LEZ sticker, to decrease convenience to such mode of transport and to incentivise more sustainable behaviours.


Bénédicte receives such notification on her smartphone and, at the next journey planning query, her MaaS app puts in low priority or with a cost surcharge the itineraries including single-car trips; other transport modes are now even quicker due to overall traffic reduction, thus Bénédicte decides to opt for the multimodal trip including bus and bike sharing, contributing in saving CO2 and receiving the green label from her MaaS app to be redeemed in the future with special offers provided by her MaaS operator.

Artificial Intelligence is a major enabler for C02TRAFFICAI. AI will enhance the gathered data with insights, predictions, counterfactuals, acting in an autonomous way without explicit user requests. Additionally, in some interactive scenarios the intelligent system works with one or more human operators to explore a given situation beyond what is expressed by the available data; at a minimum, the solution would present itself as a (probabilistic) Decision Support System, with some possible PoC-level examples of automatic actuation. Among the possible AI approaches, C02TRAFFICAI will leverage probabilistic inference (Bayesian and, more specifically, causal) because of its suitability for time series analysis, base explainability, and effectiveness even in situations where training data size is limited; combination with neural networks and deep learning is also considered within the C02TRAFFICAI design landscape.


Firstly, AI will enhance the measurement and estimation of CO2 (or other greenhouse gases if needed) emissions to approach real-time and hyperlocal monitoring. Secondly, the scenario- and simulation-based capabilities of causal inference will be put to use to assist and empower city stakeholders facing complex and timely decision making in the face of large and dynamically changing data streams. The combination of direct CO2/GHG measurements, model-based emission estimation from other variables, and AI inference and learning will enable quick and effective data-driven decisions.

In turn, such a decision support framework lays the groundwork for the next level of policy-driven MaaS by lowering the barriers to dynamic, situation-based governance that can achieve joint objectives combining sustainability, economic prosperity, and welfare for the city and its citizens. Last, but not least, further initiatives from public services, private operators, and citizens themselves can be empowered through open collaboration around the common data, the transparent policies, and the resulting AI-enabled insights.