To turn theory into reality, autonomous vehicles should have the following capabilities in the future, which will help save lives. It is imperative that Congress and DOT encourage and foster the development of such technologies.

• Autonomous vehicles should be able to recognize when they are approaching grade crossings by identifying the various signs and pavement markings associated with those grade crossings. There should be sufficient technological redundancies in place in order to ensure that autonomous vehicles retain the capability to make these determinations in various types and degrees of weather conditions, as well as if signage were down or misplaced or if road conditions were seriously deteriorated.
• Autonomous vehicles should be able to detect approaching trains, including identifying locomotive headlights, horns, or bells and account for any variables that might obstruct their view.
• Autonomous vehicles should not begin crossing tracks unless they will be able to fully move through them. Stopping on tracks because of traffic queueing or other causes creates a dangerous situation that can be prevented with highly automated vehicle technology.
• It is important for designers of autonomous vehicles to understand that Positive Train Control (PTC) was not deployed across the entire rail network and does not have the capability to communicate train location or speed information to highway vehicles in any event.

Automation promises to significantly enhance other areas of rail safety beyond grade crossings. Automated technologies can detect a wider range of defects, respond faster, and provide a larger window for action than a safety system that is subject to the limitations inherent in human eyes, minds, and hands. Automated track inspections can reduce track defects, leading to fewer accidents. Likewise, automated inspection of locomotives and freight cars has been shown to reduce the occurrence of broken wheels and other mechanical problems.

Unfortunately, due to the current limited regulatory framework, many new technologies can only be used in conjunction with, rather than as a replacement for, manual inspections required by existing Federal Railroad Administration (FRA) regulations. Railroads can sometimes obtain a temporary FRA waiver from existing regulations, but that process is often cumbersome and uncertain. These regulations discourage investment in innovative technologies.

Because automation in the rail industry is new and unfamiliar, regulators will be pressured to identify and resolve every possible risk before allowing testing or early deployment. That pressure must be resisted because hesitation will come at a cost to safety. DOT recognized this in the context of autonomous vehicles in AV 3.0 when it claimed that “delaying or unduly hampering…testing until all specific risks have been identified or eliminated means delaying the realization of global reductions in risk.”

DOT should realize these safety benefits for rail, as well, by encouraging the early deployment of autonomous or highly automated technology on railroads. Unlocking the many potential benefits of automated technology is just as important for railroads as it is for other transportation modes.

In formulating a regulatory framework that ensures a level playing field for all modes of transportation and that encourages the realization of the benefits of emerging technologies, railroads urge Congress and DOT to adhere to several principles.

• Limited short-term waivers from existing regulations do not give the industry sufficient confidence to invest in new technologies. Regulatory barriers must be overcome in ways that are more enduring than waivers. For example, Congress could direct DOT to make permanent long-standing waivers whose value has been proven through successful implementation. Additionally, DOT could issue waivers of indefinite duration and provide procedures for the expedited conversion of time-limited waivers to permanent waivers or final rules if equivalent or improved safety has been demonstrated.
• To the greatest extent possible, carriers and equipment manufacturers should be permitted to continue to create voluntary standards for safety technology. No one has a greater stake in the success of new safety technologies than carriers and their suppliers, and market pressures already incentivize them to create and implement safety technologies that work.
• New regulations governing automated operations in the transportation sector should be performance-based rather than prescriptive. This will focus industry attention and effort on the outcome rather than on how that outcome is achieved. Performance standards would give the industry discretion to experiment with new ways to improve safety while still being subject to DOT oversight, which would oversee goal-setting, ensure that measures and data are accurate, and impose sanctions if carriers failed to meet their safety targets. As such, employees, customers, and the public at large would still be fully protected.
• Regulation of automated operations should occur at the federal level to avoid a patchwork of state and local rules that would create confusion and inhibit the deployment of safety technology. State and local laws governing rail safety and operations are already preempted by federal law and regulation, and it is especially critical to the efficient functioning of the national rail network that the principle of a uniform set of national regulations not be undercut by state or local laws targeting autonomous or highly automated technologies.
• As with any new technology, public fear of the unknown is often unfounded but can prove to be a major obstacle. The public can and will read much into what DOT and FRA say or do not say on the issue of automated technologies. We urge DOT and FRA to be supportive of innovation and work to facilitate the realization of the benefits of these technologies.

The nation’s railroads put technology to work every day. Here are just six innovations railroads have deployed to maximize the safety and efficiency of the nearly 140,000-mile rail network.

1

Track geometry cars measure every inch of track

Freight railroads use track geometry cars to identify anomalies in train tracks. The cars travel along the tracks measuring every foot for rail-wear, track alignment, elevation in curves, gauge (distance between rails) and many other track geometry measurements.

After the initial survey is complete, railroads quickly respond to any issues, keeping small problems from becoming big ones. Asim Ghanchi, AVP of Enterprise Services s at BNSF, says that track geometry cars — including its autonomous fleet — allow BNSF to measure over 300,000 miles of track each year. “Inspecting that much track, that often, would be impossible without technology,” Ghanchi says.

BNSF also uses machine learning technology to analyze data collected by track geometry cars, which allows them to predict track problems that may occur over the next 30 days. This flexibility means the railroad can complete maintenance at an ideal time, which maximizes safety and network efficiency.

2

Sonar helps protect bridge piers.

Railroads use sonar to assess the stability of bridge piers. Sonar identifies increased erosion around the piers, which can compromise a bridge’s integrity.

The sonar technology railroads use is similar to the echolocation whales use to understand their surroundings. Sonar sends out sound waves, which bounce off the bridge piers and the ground surface below the water. Then, based on the nature of the echo, railroads determine whether there are any concerns with the stability of the bridge piers.

Railroads use sonar to assess bridges when significant water flow occurs following big storms or flooding events.  This is especially useful when the water is extremely murky, making it difficult for a human diver to evaluate — or even see — the piers.

“With sonar, we have a better sense of what’s going on underwater,” says Kevin Day, former General Manager of Field Technology and Data Analytics at Canadian National. “And divers don’t have to enter challenging and unsafe water conditions.”

3

Smart sensors keep wheels turning.

Smart sensors placed alongside track use a host of technologies — such as infrared and lasers — to assess the strength and health of wheels and bearings as they travel across the nation’s rail network.

Overheated wheel bearings can lead to train derailments, so railroads use hotbox detectors to measure the temperature of bearings. When a bearing gets too hot, it can only safely travel another five to 100 miles.

CPKC uses acoustic bearing detectors to predict when bearings will overheat three months in advance of them actually overheating. These sensors use acoustic signatures to evaluate the sound bearings make.

By combining this data with the data collected by hotbox detectors, CPKC can find patterns in the acoustic bearing detector data that indicate when bearings could overheat. “By paying attention to the acoustic bearing detectors, we can make sure our bearings never get too hot,” says Kyle Mulligan, an AVP of Operations Technology at CPKC. “Technology makes us safer and more efficient. It protects our infrastructure.”

4

Locomotive simulators safely train engineers.

As a supplement to required field training, CSX uses tabletop locomotive simulators to train engineers on Positive Train Control (PTC), a set of innovative technologies that automatically monitor the safe operation of a train and prevent certain kinds of human-factor incidents.

Engineers virtually learn train handling procedures on different parts of the track, seeing, for instance, how going down an incline will increase speed. They can also experience how the PTC system initiates by constantly assessing a large number of variables to guarantee the train has the necessary time and space to come to a stop safely where necessary to do so along the route.

To create the simulator, CSX used a helicopter equipped with advanced imaging technology to capture detail about routes, including curves, elevation, track speed and even the location of buildings and overpasses along the track. “It’s very realistic,” says Patrick Barnett, Technical Director of CSX. “Our engineers really feel like they’re in the train’s cab, on the route.”

5

Big Data makes railroads safer and more efficient.

Railroads use machine learning to predict a number of maintenance issues — such as track wear and tear — based on patterns and trends found in huge amounts of data.

“It’s a lot like when your phone guesses what word you’re going to use next in a text message,” says Mabby Amouie, Norfolk Southern’s AVP of Enterprise Data and Analytics. “By analyzing patterns, we can predict the rail’s wear and tear over time.”

Using models and algorithms powered by machine learning and artificial intelligence, Norfolk Southern can predict the wear and tear of track over a five-year period with a high degree of confidence. This five-year look-ahead window allows the railroad to proactively plan repairs and maintenance, helping make its network safer and more efficient.

Track wear is only one of several different use cases for which Norfolk Southern has developed models to proactively leverage big data. And with so much data at its disposal, Norfolk Southern foresees many more exciting big-data applications on the horizon.

6

Machine Visioning

With machine visioning technology, Union Pacific collects 40,000 images per second of trains passing on tracks. A series of algorithms then analyze the images to identify any anomalies, allowing Union Pacific to address issues faster than they could with manual inspections alone.

The technology helps railroads look at many elements all at once, providing a comprehensive view of locomotives, trains and their components. Union Pacific uses machine visioning in Nebraska, Iowa and Arkansas and is exploring where to deploy it next.