AI in the Railway Ecosystem

Every year billions of people travel by train. Maybe it is still less than air passengers, but rail travel is far from being obsolete. Yet to compete with other travel modes in terms of safety and convenience the rail industry must develop, which today means embracing digitization. Innovative technologies based on artificial intelligence and machine learning offer Railways of vast potential for improvement. So today we look at five ways for AI to make rail travel better.

The most obvious AI implementation in the rail industry is autonomous trains. We already have working prototypes of flying vehicles, self-driving vehicles and even flying self-driving Vehicles. While the world is going into a frenzy about unmanned cars, many rail systems have been successfully operating autonomous trains for decades. In 1968 the Victoria line of the London Underground opened as the first fully automated metro line. Today multiple cities like Paris, Dubai, Tokyo and Barcelona operate modern driverless Urban trains. As for non-metro rail in October 2021 Siemens and Europe's largest railway company Deutsche Bahn revealed the world's first automated driverless train in Hamburg, Germany as part of a 70-million-dollar modernization project. China started the operation of a completely autonomous bullet train for the 2022 Winter Olympics. It can travel 350 km/h moving its passengers from Beijing’s downtown to the Olympic venues in just 50 minutes down from three hours via a regular Express train.

Here is how it works: automatic train operation or ATO controls the train traction system and brakes. It also gets the current traffic information over the radio to adapt operations in real time. So, the increase in efficiency is achieved by optimizing speed and reducing braking. The result is more punctual trains, more stable timetables, greater travel comfort, less energy consumption and reduced mechanical stress on vehicles. On top of all that, the automated operation is reported to reduce accident Risk by eliminating human error which brings us to an important topic of Safety and Security in rail. There are some ways that AI can help here when combined with video surveillance.

We are used to cameras all around us. It's estimated that over 1 billion cameras are installed in cities worldwide. The main goal of video surveillance is security and when enhanced by AI CCTV technology can identify threats and prompt reactions more efficiently. In the rail industry cameras installed on stations can detect people on tracks and immediately inform controllers who can stop the train thus saving lives. One example for this technology is the implementation of such a system by Docklands Light Railway in London. The operator has defined detection zones along the platform edge. When a person enters the zone, an alert sounds, and a live video image is immediately displayed in the control center. Similar computer vision technology also spots trespassers. Moreover, it can even be trained to identify aggressive behavior loitering, theft or unattended baggage and alert security teams.

Another application of AI enhanced video technology is analyzing passenger density in rail cars. Then the system sends this information to passengers waiting at the station so they can be directed to a less crowded car and so we move from safety to convenience and service quality. That's another potential area for AI implementation namely digital assistance. If you have an iPhone, you're probably familiar with Siri. Those who have ever dealt with a smart home system might have used Alexa or Google Assistant to control it and most likely you're one of the 1.5 billion people who have encountered an online chat bot at least once. All of these tools are powered by conversational AI. This is a technology that can understand the user's natural language and provide personalized answers.

Here are some examples:

• Amtrak launched its virtual assistant JULIE back in 2001 to handle passenger queries. Today Julie assists with 5 million questions annually booking tickets and providing other helpful information like train schedules directions refund and baggage policies and much more.
• Indian railways presented their chatbot DISHA in 2018. This smart bilingual assistant processes about 150 000 passenger queries daily improving customer satisfaction by 70 percent. It also reduced the usage of other communication channels by 70 percent, taking heat off the customer support teams.
• Japan Railways launched their multilingual BEBOT to guide visitors that arrive at the Tokyo train station. It can provide information on landmarks, local restaurants, currency exchange offices and sightseeing spots in the area.

One more way AI can enhance rail travel is through route planning and scheduling. Today many trains run on age-old routes that are no longer optimal while missing out on areas with high demand. AI tools analyze both historical and real-time data on ticket sales and travel patterns. This way they identify Trends in passenger flow and find ways to optimize train routes and schedules. On top of that AI can incorporate such information as weather changes, holidays or special events to predict demand and best adjust rail operations to actual passenger needs. Toshiba a technology Provider from Japan created a digital twin simulator model for the UK rail operator Greater Anglia to optimize its operations. Digital twins are essentially virtual replicas that reproduce real world environments. This technology helps better understand the system performance and tryout or simulate various scenarios to find the optimal one. A digital twin made by Toshiba is designed to analyze the Greater Anglia rail network and improve their train schedules. What if delays happen and planned schedules are disrupted? Hitachi’s hybrid Railway traffic management system is one of the solutions that are designed to address operational disruptions. It incorporates intelligent ML-based technology to automatically recalculate and modify timetables and train routes in real time.

Besides passenger-related processes AI can greatly enhance internal operations. It plays a significant role in predictive maintenance, railway inspections drive for behavior analysis and much more. Today rail companies install multiple sensors on tracks and rail cars. That turns trains into a kind of rolling data centers. Powerful onboard computers analyze incoming data and find ways to increase operational efficiency reliability and security. SNCF a French rail operator spent over 10 years installing thousands of sensors and transmitting devices on its rail stock. The results of this arduous project are impressive: today they analyze over 8 000 variables per train including 2 000 in real time from over 1 000 trains at the same time. With remote diagnostics and predictive maintenance in place they can eliminate about two-thirds of breakdowns. Moreover, their maintenance costs were cut by 20 percent.

These were just a few examples from completely different areas of application that impressively demonstrate the great importance of AI for railway technology. Regardless of whether it is in service to increase passenger comfort, operational safety and reliability or predictive maintenance, the applications described and already in use show the great future potential of AI in the railway sector.

Author: Dimitrios Koutrouvis, Managing Director Lütze Transportation GmbH