In an interview, we asked Gregory Addes, from our team dedicated to video security and onboard solutions about his work developing new digital technology.

How can journeys by public transport be made more seamless and personalised? Gregory reveals the surprising ways in which artificial intelligence (AI) is being used to customise passenger experience.

What is your career background?

To begin, I studied Electrical Engineering in France. Since then, my career in rail has spanned over many different departments and countries! Initially, I developed site security for an oil & gas Company in Middle East. Then, I went on to work on video security for the Louvre in Paris and the Caracas Metro in Venezuela.

Understanding technology has always been a critical part of my work. Understanding the market is also equally important, which is why I did an MBA. Co-innovation is now a huge part of what people in my sector do. It’s about design thinking – we work with our customers to capture their pain points and we work rapidly to develop solutions.

Tell us more about your current role.

I am a Product Line Manager for Video Security and OnBoard solutions at Hitachi Rail. My team’s focus is on making journeys easier, and more comfortable. One of the main ways we achieve this is by deploying artificial intelligence technology onboard.

What are the big technology trends that are shaping the rail sector?

I would say that the first trend is artificial intelligence. Tesla is already using AI to transform the car industry. My work involves developing this this AI technology for trains.

5G is the other major trend in the rail sector. It is revolutionising communications, and it opens the door to easier train autonomy. We are also seeing the rise of onboard intelligence, as more and more functions move from the wayside to the train itself.

How will passengers benefit from AI?

Personalization is one of the biggest benefits, made possible by AI-powered video analytics. Our solution, DIVA – Distributed Intelligent Video Analytics, enhances the passenger experience in several ways. For instance, it can detect wheelchair users and guide them to the most accessible part of the train, while adjusting the door closing sequence to give them more time to board.

There are many other examples. We can tailor passenger announcements and advertisements based on occupancy and time of day. Additionally, we can perform an “empty train check” at the last station of the rail network. In the event of abandoned luggage on board the train, the system can display the related belongings on platform screens. Hopefully, the concerned passenger will promptly retrieve their belongings, thus preventing the need to call security and halt all trains on the railway network.

What about operator benefits?

DIVA is transformational from an operational perspective. People on tracks detection is an example: with DIVA, we could detect unauthorised access to the track and give the train driver an early warning with the associated video streams coming from the platform. It’s able to respond to what is happening not only on the train, but on the platform as well.

We can also adapt train scheduling. Traditionally, timetables were set in stone. Today, we are working on dynamic scheduling – the ability to match the flow of trains to real-time demand, based on passenger density data from trains and platforms. All of this is derived by processing existing video data with AI.

Have any of these AI functions been deployed yet?

Yes. This passenger density solution has been deployed on 100 trains in Middle East and has proved its efficiency during operation while an important public event took place.

What are the challenges of deploying AI onboard trains?

AI requires a lot of power and processing capacity – and deploying this onboard is a challenge. We manage this in three ways. First, we utilise GPUs – ultra-fast graphics processors. Second, we use frugal algorithms to minimise power demand. Third, we focus on the type of information required: what level of processing capacity is needed to satisfy a specific objective? We focus on doing more with less.

How far has this technology come?

When I started doing video analytics 20 years ago, we didn’t have AI. You needed to configure every single camera – it used to take hours. What is amazing about DIVA is that it works straight out of the box: I connect the camera, and it will detect people automatically. It is a game changer!

What background do people need to work in this field?

Aspiring professionals in this sector need engineering talent, particularly in electronics and IT. But equally important, interested candidates need a talent for collaboration. That means the ability to communicate, to listen, to share, and to win as a team.

They also need to be ready to learn, people in this sector need to adapt rapidly to new technology like AI.