RAILWAY AGE, SEPTEMBER 2025 ISSUE: Imagine a future where autonomous freight trains in North America operate across long distances. This future is one the industry must consider if it wants to compete against trucking and its futuristic vision of autonomous truck platoons, according to those working to bring autonomous operations to freight rail. “If it can help the industry be more efficient and productive, it can bring more volume to the network,” Philip Moslener, Wabtec Corporate Vice President Advanced Technologies, told Railway Age.
Automation Exploration
Autonomous rail already exists in many places around the world, such as airport people-movers or driverless rapid transit like Honolulu’s Skyline system. But taking rail to the next level and incorporating autonomous technology into the freight rail space, like that used in Rio Tinto’s Australian operations, requires the ability to handle complex situations. Unlike airport people-movers and rapid transit, which operate in a captive, self-contained environment, autonomous freight rail must have systems that can anticipate, analyze and respond to various situations occurring along an open right-of-way, in real time.
“You need to have what’s called a perception system,” Moslener said. He and his team research new technologies, such as those incorporating AI (artificial intelligence), and see how they can be leveraged in rail operations to support market share growth and reduce carbon emissions. “It could be cameras, LiDAR (Light Detection and Ranging), heat sensors, acoustic sensors, you name it. That’s one of the fundamental building blocks of any autonomous system—rail, truck, cars, whatever.”
MxV Rail has been working to develop increased levels of automation in North American freight rail. Researchers there have been benchmarking efforts against what’s been done in Australia, and they’ve also studied automation advancements in the automotive space, according to Joe Brosseau, Assistant Vice President of Communications and Train Control. “Overall, the approach to automation in North America has been an incremental development roadmap, essentially adding technologies piece by piece to attain near-term benefits that will increase safety, operational consistency and efficiency, but building those things on a pathway to higher levels of automation,” he said.
When the freight rail industry initially explored automation several years ago, it looked first at designing the full specifications of what would be required for autonomous operations, according to Brosseau. However, as time went on, the industry realized that one huge system would lead to unreasonable costs with no benefit for a long period of time. “In 2021, we redesigned the program to focus more on an incremental approach where we can gain benefits along the way but still be on a pathway to what we call high automation,” he said.
Now the rail industry uses a multi-pronged approach to develop automation technology. This approach seeks to incorporate Positive Train Control (PTC), the existing overlay system that enforces speed restrictions and train separation, among other safety-critical tasks.
Researchers have been exploring how PTC can interact with sensors that would enable autonomous operation, including those that detect and identify right-of-way obstacles, for example. Information from those sensors would interact with PTC to set up stop or speed targets. Some in the industry call this “PTC 2.0.”
“PTC enforces the maximum speed limit, but it’s still incumbent on the crew to look out for obstacles in front of the train,” Brosseau said. “With the restricted speed enforcement enhancement, we essentially add some of these sensing capabilities to detect obstacles.” The PTC system gets the information on the obstacle and automatically forces the train to slow down. MxV Rail has also explored how energy management can promote automation in manual control zones, according to Brosseau.
MxV Rail doesn’t directly investigate use of rail AI applications, but it does indirectly review its use. The various sensing capabilities that MxV Rail is testing with the supply community also involve machine vision capabilities, for example; “AI that’s interpreting data and then providing that information to automation systems, energy management or PTC, to detect obstacles,” Brosseau said.
Indeed, suppliers such as Wabtec use AI to develop sensors that take in data from the environment. The AI assists in situational analysis and trajectory planning, according to Philip Moslener.
Research Vs. Real World
MxV Rail spends a lot of its time in autonomous rail researching sensing capabilities. One aspect of that is evaluating the radar technology being used in some of the automotive automation space and developing proof-of-concept testing. “It’s a low-cost technology, and from what we’ve seen, it may be capable of doing what’s needed for restricted speed enforcement operations, our current focus on the incremental roadmap,” Brosseau said.
Another aspect is working with suppliers with technology they believe can meet requirements for the automation space. They involve several different sensor types, among them cameras, video, radar and LiDARs. “We’re not focused on one specific technology. We want to see what’s out there,” Brosseau said. “We’re setting up a program for next year where we hope to be equipping locomotives with some of the technology now available in the supply market, essentially demonstrating if it has the potential to meet our requirements for restricted speed enforcement, specifically.”
MxV Rail has been working with suppliers to conduct testing at MxV Rail’s Pueblo, Colo., campus, but Brosseau hopes next year will see deployment in revenue service operations so that MxV Rail and others can collect information in different operating environments. “There’s a lot we can do at our facility in terms of testing out fundamental functions of the technology,” Brosseau said. “But for this type of sensing technology, we need to see how it operates when there’s dense foliage, for instance, which we don’t have here in Colorado. We want to collect information from a more diverse environment.”
Like MxV Rail, Wabtec approached the concept of constructing an autonomous rail system by thinking in terms of multiple building blocks. These include perception, supervisory and PTC systems. This approach has helped Wabtec develop Teleops, its solution for when an autonomous freight train breaks down in a remote location and there is no one on board to take care of the problem.
“Teleops is basically the ability for somebody to go on board the train virtually,” Moslener said, likening the program to situations where someone in an IT department assesses a work computer remotely. “Teleops means ‘operation from a distance.’ So, it’s somebody at a desk, with the controls, patching themselves into the train and controlling it.” Wabtec has demonstrated the technology on its locomotives at its Erie, Pa., test track.
Although Teleops is still in the research and development phase, Wabtec is using some of its elements now, according to Moslener. “We can use it today, for example, to prepare trains for departure,” Moslener said. “In a yard, it takes a long time to initialize a train and prepare it for operation. Why can’t we do it from an air-conditioned office?”
Another Wabtec R&D project is the Maverick, a small autonomous locomotive that would haul short trains traveling distances less than 750 miles, such as between a port and a distribution center. Designed to haul 10 to 15 railcars, a Maverick would be equipped with perception and teleoperation technologies, enabling it to be operated from a remote dispatch center. Its purpose is to capture a short-haul, just-in-time market that tends to be underserved by the rail industry and where transloading from truck to rail is cost-prohibitive.
Development of Wabtec’s R&D projects depends in part on market demand and interest in these offerings, according to Moslener. However, “we’ve found that as we develop these technologies, we come up with other ideas,” he said. “Teleops is going to be needed for a large, fully autonomous train, but we can use it today in certain applications.” For example, a mining customer in Brazil is interested in using Teleops for unmanned zone operations where it’s too dangerous for humans because of landslides.
Through its work on Teleops, Wabtec has developed Pathfinder, a technology that can be implemented on older, low-horsepower locomotives to allow them to use new digital equipment and sensors that one day may allow for autonomous operations. “If you want Netflix on your older TV, what do you do? You change the TV, or get a Roku Streaming Stick,” Moslener said. “Think of Pathfinder as a Roku Stick for an older locomotive. Pathfinder, with its digital technologies and PTC, can be plugged in, allowing older low- and medium-horsepower locomotives to be outfitted with new technology.”
Wabtec describes Pathfinder as “a small digital slug designed to automate the train, not necessarily the locomotive. It detaches the brain from the brawn, where the low-horsepower unit still relies on its existing propulsion technology, but all the intelligence and automation in how to run the train resides in Pathfinder.”
Meanwhile, the R&D team at Progress Rail has been developing Talos, which the company describes as “leveraging machine learning to analyze and optimize train routes, as well as develop customized strategies for individual routes that factor in operational conditions. Talos incorporates track topography, train consist information, route data and historical analysis to build a unique and optimized operating strategy resulting in significant improvements in fuel and/or time efficiencies, The technology will also enable the locomotive engineer to focus on safety and situational awareness.”
Talos has been designed “for use in all locomotive makes and models,” and Progress Rail engineers have been testing the technology at MxV Rail. The company says the train, controlled by AI, operated more than 200 miles “without human intervention, operating flawlessly with independent power.”
“Talos is more than an energy management system—it’s a bridge to the future of autonomous freight rail,” Sammy Akif, Progress Rail Vice President of Advanced Rail Technology, told Railway Age. “By integrating an AI-based controller and machine vision safety systems, Talos transforms traditional rail operations into an intelligent, sustainable network. As the industry moves toward automation, Talos stands out as a scalable, interoperable platform designed to optimize fuel, reduce emissions and seamlessly connect legacy infrastructure with next-generation locomotive technologies.”
Overcoming Hurdles
Discussing autonomous freight rail operations in North America may sound glamorous or futuristic, but for some, the elephant in the room is whether autonomous trains will replace workers. Moslener said he believes that wouldn’t be so, and Teleops is being designed with the understanding that automation will still require some level of human involvement in the loop.
“Everybody’s saying, it’s going to replace jobs and all that. The way we look at it is, there’s a lot of opportunity in rail. Rail is underutilized,” Moslener said, pointing to how the North American rail network, despite some choke points, has a lot of capacity. To him, addressing that underutilization “really should be our motivation in the industry to say, if we can grow rail traffic, it means more jobs, not fewer jobs. It’s not about eliminating people. It’s about maximizing what we have. We’re working for the purpose of increasing the amount of volume on the network, which increases the amount of jobs, and we’ll need more maintenance and more operators. The message we’re trying to convey: Don’t be afraid of AI in this space. A good utilization of AI is one where we can improve efficiency and bring more volume onto rail, and that just benefits the industry.”
Besides the issue of jobs, bringing autonomous technology to freight will also require discussion and interaction with the public, just as the debates over autonomous trucking lanes have surfaced. The major benefit in developing intermediate steps to autonomous freight rail technology is that “it builds trust with the public,” Moslener said. “Autonomous technologies are now used in other applications. We need to build trust in the systems we’re developing. The building block approach helps gain trust in these systems. We’re going to be able to get to autonomous freight rail at some point, but we need to make sure we can trust these systems. And we’ve got to put them out there and try them.”
