IT might look like a normal articulated truck, the same as the millions that carry freight on North America’s highway network. But a silver Freightliner Cascadia lorry currently driving on Virginia’s roads is doing so entirely under the command of autonomous driving software.
A licensed safety driver is in the cab and there is an engineer to monitor the system. However, Freightliner, which is owned by Daimler Trucks, hopes these supervisors won’t have to take the wheel.
The tests began in September in the south of the state and the plan is to expand the deployment of the technology, which delivers US Society of Automotive Engineers (SAE) Level 4 autonomy, to more lorries over the coming months. This strategy reflects the goal of the Daimler’s Autonomous Technology Group, founded in May with a $US 570m investment budget, to deliver SAE Level 4, where drivers remain in a supervisory capacity, to all its vehicles within the next decade.
Daimler isn’t alone in developing automated trucks. US autonomous pioneer Peloton is working with the US Post Office on a two-truck platooning pilot, which does not require a driver in the second lorry. UPS is similarly conducting autonomous tests in Arizona having purchased a stake in TuSimple, another automated driving start-up. Even Tesla is active in this area. The Semi, the company’s electric articulated lorry, is set to feature the second-generation of Tesla’s semi-autonomous technology, offering automatic braking, lane keeping and lane departure warnings.
For railways, the economics of automated trucking are a major concern. Convoying trucks offer significant cost savings and efficiency gains for rubber-tyred vehicles. Electric lorries are also cleaner than diesel locomotives.
The United States Class 1 railways’ response is to improve the efficiency and economics of what they have by introducing their own form of automation. With the rollout of Positive Train Control (PTC) now in the final stretch, the railways are shifting their focus to projects which use this infrastructure. The goal is to improve the efficiency of operations in the short-term with an eye on delivering various stages of autonomous operation in the medium to long-term.
Among these railways is Union Pacific. By the end of the year, and over 11 years since PTC was mandated by the federal government, UP will have invested $US 2.9bn to install PTC equipment on 27,350 route-km and 5515 locomotives. The railway has installed more than 10,000 wayside antennas and according to Mr Michael R Newcomb, the railway’s assistant vice-president for transportation systems development, is currently running 14.5 million service-km every month using PTC.
Newcomb was speaking at the Arema conference at September’s Railway Interchange in Minneapolis. He says that with UP’s network virtually complete, some of the current work in the final stage of systems data operations for PTC is focusing on improving interoperability with other railways.
“Our ability to exchange diagnostics information in support of operations is maturing pretty rapidly,”
Michael Newcomb, assistant vice-president for transportation service at Union Pacific
Newcomb reported that 12 of UP’s 31 tenant railways are now interoperable, equivalent to 85% of total operational kilometres. However, with in-field expertise reducing due to retirements, UP is looking to support its tenants by assisting them with the testing and upgrade process. By issuing release and change management documents, he says the goal is to reduce the replication labour and local expertise required from release-to-release for PTC.
The railway is also looking to improve diagnostics capability for smaller railways. While the Class 1s have advanced diagnostics for themselves, this is often not the case for their tenants. Currently if a small railway has a braking event, UP will request the logs from the locomotive. Initially, this was done via email, but increasingly drivers will use an onboard touchscreen module as a standard way to ask for and log exchanges. The Class 1s are set to have the capability to download this information electronically by the end of the year. In addition, through work with AAR subsidiary, Railinc, UP is aggregating data from this diagnostics information. “Our ability to exchange diagnostics information in support of operations is maturing pretty rapidly,” Newcomb says.
Another area relates to Reliability, Availability, Maintainability and Safety (Rams) of the PTC network.
The Class 1s requested and ultimately paid for vendors to redesign the initial PTC model in 2012. The result is that engineers are now able to write software into the system allowing the railway to model changes and assess their possible impact before they are introduced, which could have significant implications for improved traffic management in high density areas.
“The purpose of this is to get evergreen capability at the industry level for our Rams management exploits,” Newcomb says. “All of our models produce a train delay calculation, that’s the value we added. If there is a train delay, you can calculate the business benefit and the consequences of a change. Our goal is to have the infrastructure in place so as we make changes to the system - maybe we are required at some point to have a new public communications capability - we want to model it, put it in the system and see what the net return will be.”
The railway is also working on improving geo-location of its assets. PTC relies on GPS, which is accurate to around 8m, but makes identifying which track a train is on difficult. Trials with the Positive Train Location (PTL) system developed by Leidos have taken place over the past four years across the United States. The system combines information from multiple sensors and a track database (when available) in the Leidos Embedded Data-fusion Geospatial Engine (Edge) sensor fusion algorithms to create optimal state estimates for position, velocity, and attitude. BNSF has already installed the first-generation product and UP is set to follow suit to assess the product’s performance in real-life operating conditions.
Backbone
Beyond optimising current systems, work is underway to identify, develop and install new technologies that would form the backbone of a future autonomous train control network. Among the early steps in this process is a higher-level study looking at current signalling system design, the approach, and rules, which have been introduced in response to various accidents in the past.
Newcomb also reports that UP is developing the technology and protocol necessary to introduce Quasi Moving Block (QMB) operation.
“We have industry consensus on the concept,” he says. “We have this idea where we might be able to eliminate the intermediate signals as we go towards moving block, but we are not proposing to eliminate track circuits. Roll-on-roll-off protection, presence detection, and broken rail detection, will be maintained, and the information will be transmitted to a back office from where we will issue authorities and establish the train spacing. During normal operation no trains will occupy the same block, as they do with Centralised Traffic Control (CTC), but we will split the block into smaller authorities.
“We are heading towards restricted speed protection and train occupancy protection. BNSF is doing a test on this before us at UP, and we see this as a 36-48 month deal to get through all of the safety analysis and to get all of the standards to work. BNSF has their product and we will match the standards. This is not vapour wear, you are going to start seeing it shortly.”
“Rather than relying on wayside infrastructure, trains are set to increasingly communicate with one another and respond accordingly”
Michael Newcomb
Among the key enablers for UP, highlighted by Newcomb, is the update to the end-of-train device (EOT).
Developed during the 1980s to replace the traditional caboose, as well as providing a visible indication of the rear of the train, the devices can monitor brake pipe pressure, relaying information via telemetry to the locomotive. Developing the EOT is necessary to expand PTL capability, which is available for the front of the train, to the rear. Newcomb says this will offer the integrity required to close-up the block distance and enable trains to travel at higher speeds. He says that UP will also add a back-up camera to the next-generation EOT, which is optional, but will be in the specification as it is considered a major enabler of single crew operation in the future.
QMB could offer vast improvements in train reliability. “Rather than relying on wayside infrastructure, trains are set to increasingly communicate with one another and respond accordingly” , significantly improving operating conditions in challenging areas like Chicago.
While QMB removes this restriction within the confines of the existing track circuit model, full moving block will find a replacement for track circuits, with research in this area underway. However, Newcomb concedes that to get there, additional communications capacity is required beyond the 220MHz band currently used by the railways, with his preference the use of 60MHz. “That is part of the thinking and planning moving forward,” he says. “I see communications becoming a large part of the investment cycle over time.”
An enhanced train energy management system (EMS) is also set to play a major role in managing train performance and forces. While EMS was included in the initial specification for PTC, Newcomb says it slipped down the running order as the railways suffered from “interoperability fatigue.”
EMS is now returning as a priority capability that can manage key functions in the train, from stop-start, to air and dynamic brakes. Newcomb says PTC provides the situational awareness for EMS, meaning that it does not have to be turned off when the signal is not there, thus unlocking huge potential to improve the efficiency of operation. “It used to be a clear signal product, now it is a railroad product,” he says.
A key milestone for UP’s use of EMS took place at the beginning of September when the railway operated a PTC-enabled train with EMS “zero-to-zero.” Instead of the EMS activating when the EMD SD70AC locomotive was stationary and up to 16km/h, it remained in use for the entire 69km trip. Newcomb says integration with PTC has taken a 65% opportunity to apply EMS on a trip, given the non-clear signals and other things that occur, to a 90+%, and ultimately a 98-99% application.
In the short-term this offers fuel savings - Newcomb projects that UP will double fuel efficiency using EMS and other associated efficiency initiatives in the next 10 years. In the long-term the benefits include management of the entire consist and the movement away from the distributed power model to autonomous control of individual elements.
As a railway subscribing to the Precision Scheduled Railroading (PSR) operating model, where trains are increasingly made up of varying consists, managing in-trains forces has become a concern.
Trains are braking more, increasing the risk of derailment with the weaker elements of the train. “What we are doing with EMS is potentially managing those in-train forces and reducing the risk,” Newcomb says, adding that one EMS product which has been deployed has exceeded expectations in terms of managing train consists. “We have been running in our most critical territory for the past three months and we have had zero breakages.”
ATO
Inevitably the ultimate objective for UP is Autonomous Train Operation (ATO), particularly if the railway is to see off what Newcomb describes as the “clear and present danger” from autonomous trucking.
There are significant obstacles to get to this point; the future of human resources and the inevitable push back from labour unions - a situation shared by the trucking industry - as well as the technical and regulatory challenges to prove that unmanned vehicles are safe.
Newcomb refused to put a timeframe on the ATO project. However, he says that UP has a high-level architecture for ATO which is subject to change. The writing of ATO specifications, which is taking place in committee and is led by the railway’s Technology Advisory Group, also got underway earlier this year.
Technically he says the most challenging issue is to solve encroachment on the line - whether this is from road vehicles at level crossings, livestock, trees or boulders. While the human brain can see into the distance and make decisions based on this, it is also restricted by poor visibility, so there is an opportunity to deliver a more sophisticated system which enables trains to run at higher speeds than they otherwise would by reducing this risk. There is also the opportunity to solve restrictive operations in areas with limited coverage.
“We have been talking to Waymo, Apple and trucking companies which are interested in rubber-tyred automation,” Newcomb says. “The challenge we see is that they do not have to see as far.
“It ends up being part of the military-industry complex, not only in the US, but actors from outside are starting to solve some of these issues. However, it is still years in the making and it will be a long capital cycle before we get there.”
UP and the Class 1s are of course not the only railways looking at ATO. Rio Tinto’s Pilbara Autohaul project, which is now in full operation, is the flag bearer for such applications.
Newcomb says that he has been closely engaged with the project, meeting with Rio Tinto on several occasions. He also says the zero-to-zero test was based on technology in use by BHP Billiton, another Australian heavy-haul railway pursuing ATO.
However, the scale of what the Class 1s are trying to achieve in the United States compared with the Pilbara emphasises the task they will face. Rio Tinto has a single locomotive fleet running the same 240-wagon consist on a closed 1500km network. UP’s 51,000km network is a whole different ball game.
“We have shown with zero-to-zero that we can do these things, but it is about doing it safely,” Newcomb says. “They [Rio Tinto] have 44 crossings. We have 31,000. That gives you a sense of the challenges ahead.”
