How an AI start-up is helping diesel trains to save fuel

RailVision Analytics develops software to help locomotive engineers make small adjustments in train driving that could lead to big savings in diesel fuel

If the AI solution was used on all Via Rail's trains, the anticipated fuel saving could cut annual carbon emissions by 20,000 tonnes. Reuters

Transport is responsible for one quarter of the world’s greenhouse gas emissions. While companies have touted electric vehicles and sustainable aviation fuel as ways to cut emissions from air and road travel, reducing the climate effects of rail transport has received relatively little attention.

A Canadian start-up wants to reduce that impact. RailVision Analytics in Montreal has developed artificial intelligence-enabled software to help locomotive engineers make small adjustments in train driving that could lead to big savings in diesel fuel.

That could help freight and passenger trains cut into the estimated 100 million tonnes of planet-warming gases released into the atmosphere every year.

“It is just like Google Maps,” says Dev Jain, founder of RailVision Analytics, of his AI application that can be downloaded on a tablet and run offline. While Google Maps tells automobile drivers to turn right or left, RailVision’s app guides locomotive engineers to “stay idle” in the next mile or “increase speed”.

Essentially, the idea is to enable locomotive engineers to eliminate wasteful practices and use invisible forces that could assist their driving. If you’ve driven a car, you are probably familiar with the law of inertia discovered by Isaac Newton in 1686, which states a moving object will continue in the same direction unless a force affects it.

That means a vehicle can stay in motion – in other words, coast – without the use of manmade propulsion until friction or other forces stop it.

But “driving a train is like driving a roller coaster”, Mr Jain says. The length of trains means that when trains are travelling on tracks where the elevation changes, often, some rail cars will begin climbing uphill while the rest are still going downhill.

That poses a tough question for locomotive engineers – should they coast or pull the throttle for more horsepower?

To solve those puzzles, RailVision collects data from rail operations to simulate train dynamics and uses an algorithm to determine the most fuel-efficient driving pattern. While the AI solution is still in its infancy, Mr Jain says it helped Metrolinx, a government agency that serves millions of passengers in the province of Ontario, to save more than 1.5 million litres of diesel fuel during a one-year experiment.

RailVision also counts Via Rail Canada, the country’s main intercity passenger train operator, as an early adopter.

Trains are arguably the most climate-friendly way of getting around. According to the International Energy Agency, rail travel releases 22 grams of carbon dioxide on average per passenger-kilometre globally, which is dwarfed by the 123g generated by air travel, and the 145g per passenger-kilometre from journeys in cars.

Diesel, though, is a particularly dirty fuel. Unlike Europe and Asia, where electrification has made headway in passenger rail service, most long-haul trains in the US and Canada rely on diesel fuel – a status that is unlikely to change any time soon, according to Bruno Idini, a transport analyst with the IEA.

That’s because running trains on electricity requires a costly infrastructure upgrade, and the region’s relatively sparse passenger service can’t justify that investment, Mr Idini says. It’s also expensive to build electric rail networks for freight trains to bring goods to cities and towns scattered across vast distances.

Although some railway companies have begun replacing diesel with lower-emission biofuel, that replacement remains “a drop in the bucket”, he adds, citing its high price and limited supply alongside other constraints.

AI could offer a relatively affordable alternative to reducing rail’s carbon footprint.

“It was the first time that we have seen an innovation that was easily installed and applicable to our reality,” says Francoise Granda Desjardins, a senior adviser of sustainability at Via Rail, which has been testing RailVision's AI solution since 2021.

Via Rail started its experiment with a simulation trial, during which engineers sat in a replica of a locomotive control cab to drive trains in virtual reality. After six months, the engineers used 15 per cent less diesel by following AI-generated guidance than driving without it.

The company has since tested the technology on passenger trains running between Toronto and Ottawa. While Via Rail has yet to finalise the findings, “it has been a positive experience so far”, Ms Desjardins says.

Founded in 2019, RailVision raised $4 million in a seed round last year from investors including Trucks Venture Capital and Blackhorn Ventures. Since fuel is the second-largest operating expense after labour costs for North America's railroads, the start-up’s solution is turning heads.

Genesee and Wyoming, a multinational railroad operator whose business includes freight trains in Canada, has recently launched a pilot with RailVision for its cargo service between Montreal and Quebec City.

RailVision charges railway companies a subscription fee for its AI guidance. While Mr Jain declined to disclose the pricing, he says the service “usually pays for itself in a few months” in saved fuel costs.

There could be notable climate gains, too. Take Via Rail. If the AI solution were used on all of its trains, the anticipated 15 per cent of fuel saving would translate to an annual reduction of more than 20,000 tonnes of carbon emissions, the company estimates. That is equivalent to taking more than 4,000 cars off the road each year.

Still, some wrinkles have to be ironed out to tap that potential. Since it is more complicated to drive in the real world than in a controlled environment, there is a gap in performance between simulated and actual train operations, according to Ms Desjardins.

Additionally, saving fuel will happen only if human beings take the AI advice. As safety regulations prevent locomotive engineers from looking at the app or listening to audio alerts while the train is in motion, they are asked to memorise all the instructions before departure.

It remains to be seen whether locomotive engineers would be willing to improve fuel efficiency in their everyday work or put their full trust in AI in the first place.

To reach net zero emissions, rail transport ultimately will have to move away from fossil fuels. But until that happens, “technologies like the one from RailVision are really helpful”, Ms Desjardins says. “If we go up to 15 per cent fuel reduction, that is significant.”