How can AI and generative artificial intelligence serve transportation?

What are AI and generative artificial intelligence? 

Artificial intelligence may have only recently gained the media spotlight, but it has long been a part of the tech landscape. As early as the 1950s, researchers were exploring the possibility of replacing human intelligence with machines. Increased processor capacity, advances in algorithm research and the explosion of data volumes have enabled the development of machine learning and deep learning and, more recently, generative artificial intelligence. These techniques allow machines to learn by themselves and improve their performance in fields such as image recognition, text translation and obstacle detection. After just a few years, these techniques are opening up new horizons in mobility, notably with the rise of autonomous vehicles.

More recently, generative artificial intelligence began enabling machines to create content such as text, images, music, video or computer code. These generative AI systems are based on enhancing existing data to produce new content. They have now become an integral part of our daily lives, for example through ChatGPT, Deepl or Mid Journey, and they also open up new possibilities in the field of mobility, particularly in customer relations.

Whether it is predicting ridership in public transit, optimizing preventive vehicle maintenance, calculating the CO2 trajectory of mobility across a given perimeter or improving safety, AI and generative artificial intelligence are now part of Keolis’ operations and services to improve the performance and quality of its networks. Let's take a closer look at three use cases.

Generative artificial intelligence in the service of passenger information  

How do you guarantee reliable passenger information during public transit disruptions? This was the issue facing the Keolis Nederland network. Unexpected disruptions in public transit are a source of dissatisfaction and stress for passengers and employees alike. The network thus wanted to set up an easy-to-use tool that would provide passengers with fast, reliable information. The adopted solution uses generative artificial intelligence. Based on observed events, the tool generates and adapts passenger information messages, which are then distributed via the network's communication channels: social networks, website, apps, on-board screens, etc. Thanks to this AI, information processing time has been cut from around 20 to less than 5minutes, with low investment and operating costs. The network's employees have already received training for this new tool.

Anticipating ridershipin real time thanks to artificial intelligence 

The Department of Transport and Planning in the State of Victoria, Australia, wanted to find an innovative approach to anticipating ridership on Melbourne's tramway network and on public transit in general. To achieve this goal, Keolis Downer and the University of Melbourne joined forces to develop IMove. This tool uses artificial intelligence to analyze external data (weather, census, calendars, etc.) and data collected by ticket validation or counting systems installed on board some vehicles. The algorithmic models they have developed make it possible to estimate tramway capacity in real time and more accurately forecast ridership. This innovation prevents the network from having to equip all trams with costly counting systems and makes up for the imperfect system of validation data, which does not take into account fraud and free travel in specific zones. In addition, by estimating ridership, we can adjust supply to meet demand and better coordinate the various modes of transport.

Artificial intelligence to reduce wait times 

How do you ensure a waiting time of less than 12 minutes for passengers on the shuttle bus to Mont-Saint-Michel? To meet this commitment, Keolis developed a tool to anticipate peak ridership as accurately as possible. Keolis' Northwest Regional Division has teamed up with French start-up Cross Data to implement artificial intelligence algorithms that forecast passenger flows and adapt the transport offer accordingly. The tool will draw on a wide range of data to ensure its reliability, including weather forecasts, calendars, information on the opening of nearby parking lot barriers and site promotion campaigns. This project will help to regulate the influx of visitors to this major tourist site. Forecasts can be shared with other local players, such as restaurant owners, whose business depends on visitor numbers. The first beneficiaries will of course be passengers, with improved passenger experience.

Next steps: continue to explore the possibilities of AI and generative artificial intelligence 

As with many new technologies, Keolis is attentive to the ethical and security issues raised by the use of AI and generative artificial intelligence, such as the protection of personal data, intellectual property and confidentiality. 

Keolis is now using artificial intelligence to improve its mobility operations in France and abroad. These first practical applications will enable us to assess the benefits of these new technologies and capitalize on feedback.