Steamrail Weekender to Maldon Victoria (July 31st to August 2nd)
Vietnam Tour - Travelling by private train on the legendary Reunification Express
QPSR Troop Train
Stunning views on a retro rail trip
Garratt coming to Southern States in 2015
The Outer Circle Line comes to ACMI Melbourne
Australasian Rail Industry Awards Website launched & Dates announced
Geelong & Ballarat Rail 150 – April 2012
Rail Revival Alliance to meet with Louise Staley Member for Ripon
4Tel CEO Joanne Wust tells Rail Express about how the company is employing AI to develop its advanced driver advisory system, to help train drivers be safer and more efficient.
Ensuring safety within the rail corridor is essential to the effective delivery of train services. Various hazards, such as animals and people wandering into the rail corridor, or vehicles standing stationary across level crossings, can cause serious injury and death and lead to serious service disruptions on a network.
Safe and efficient operations require a train’s primary and secondary drivers to be aware of the precise location of the locomotive and the presence of potentially hazardous situations and objects along the route. In a high-consequence environment, train operators must keep an eye out for potential safety risks at all times, all while watching for safety-critical signals, and operating the train in as efficient a manner as possible.
This is no easy feat, and while train drivers in Australia and New Zealand consistently demonstrate a high level of competency, they are not infallible to human error. Fatigue, distraction and loss of concentration can affect anyone during a long journey. Complacency is a human trait that can set in for drivers operating the same route on a repeated basis, and a newer driver may not be as familiar with the route and its complexities.
Newcastle-based digital rail specialist 4Tel is working towards a solution which helps drivers perform their jobs in a safer and more efficient way.
In late September this year, 4Tel carried out a test run of its latest solution, HORUS. HORUS is an Artificial Intelligence (AI) Machine Learning system for an Advanced Driver Advisory System (ADAS) – ultimately a machine-human interface that assists train drivers in the safe operation of locomotives.
The company has been working on the AI system since 2016. With progress speeding up, the project is now moving into the final stages of development of an initial version for operational use.
“HORUS is the next technological enhancement for improving the safety and efficiency of train operations,” 4Tel CEO Joanne Wust tells Rail Express. “With this system, the onboard computers are able to sense a train’s surroundings and detect abnormal objects within the corridor and even beyond the corridor.
“Digital technology can do things that humans cannot do. For instance, we can use cameras that have superb visibility at night or in fog – humans often struggle to see much in these conditions. Also, different kinds of advanced sensors can be used to feed information into the train’s computers. Just as we humans use our different senses to detect whether or not we are in danger, sensor technology enables the same thing for AI computers, but more effectively.”
The HORUS system integrates the sensor data gathered from cameras, sensors and GPS in real-time. Using neural network processing in an on-board computer, the system carries out an ongoing and continuous comparison with previous data records of a given section of track. Advanced algorithms within the software then carry out processes for detection, localisation, awareness, dynamics and route monitoring.
“HORUS can detect the approaching signal and classify the illuminated signal aspect. It can also incorporate signalling telemetry data from the control system, where available, utilise AI and GPS for locational assurance (currently set to 50cm accuracy), and identify temporary and permanent speed boards to ensure the train going at the right speed,” Wust explains.
“The system can also carry out real-time calculations of the braking profile of the train. So if the train is approaching a signal at stop, HORUS can warn the driver and provide a braking profile to assist the driver in stoppint the train before the signal.”
HORUS features a central data centre that collects as-run video that is used to update the system’s track reference record, or the route “master sequence”. This process involves machine learning techniques, which assesses changes to the route on the basis of data collection, assimilating alterations and updating the master sequence. HORUS can therefore use AI to detect both normal and abnormal train operations at a given location.
“A route master sequence is the sum knowledge of what the AI system has learnt based on all the trains that have operated on that route. The more trains that operate a route, the more things are seen and processed, the more weather conditions are experienced, and the more intelligent the AI system will become in assessing hazards from normal route operations, it has been, the more things it has seen, the more situations it has been in, the more intelligent the AI will become,” Wust explains.
The process of developing the algorithms enabling the machine learning techniques took 4Tel three years with the research assistance of the University of Newcastle Robotics Laboratory.
The technology is now at a stage where the data gathered from a sensor array on a moving train can be integrated and analysed onboard to provide real-time information to a driver. “It took us some time to develop the mathematics and optimise them because the AI industry has specific requirements,” Wust says.
“The AI has to be able to interpret the different datasets coming from the various sensors and provide an integrated analysis of this information in real time. It is not to be understated how complex it is to do something like this.”
HORUS is designed to support a variety of sensors, which would be selected in consultation with the train operator to achieve their stated operational outcomes. HORUS collects as-run data that is subsequently processed by the data centre to update the system’s track reference record, or the route ‘master sequence’.
This process involves machine learning techniques, which assess changes to the route on the basis of data collection, assimilating alterations and updating the master sequence. HORUS can therefore use AI to detect both normal and abnormal train operations at a given location.
The updated master sequence is then shared with all other HORUS equipped locomotives to enable continuous learning of all HORUS equipped locomotives. Following the recent successful test run on a route through the Hunter Valley, 4Tel is planning to carry out additional train tests in the coming months. “We’re really happy with the data output that has been achieved. It is now just a case of ensuring the algorithms are not presenting false positives, and that we are processing the information in an efficient way,” Wust says.
According to Wust, early adopters of AI technology will reap the most rewards: they will get to shape the outcome of the DAS system and use the safety and efficiency benefits of HORUS to grow their market share of rail haul contracts.
HORUS also offers the opportunity to improve the competitiveness of intermodal rail freight against road freight.
“Long distance trucks are continuing to increase their capacity and efficiency with B-Triple combinations now appearing on main interstate roads, and vehicle manufacturers are competing to develop the first autonomous and driverless trucks. A truck driver also has few limitations on where they can drive their truck in Australia.
“By comparison, a train has two drivers and they are limited to operating in the territory of their route qualification,” Wust says.
“We see this as an opportunity to assist train drivers with better informed technology, to allow the drivers to focus on the tasks a computer can’t perform.
“Our HORUS technology is designed to work seamlessly across the various rail networks and contain the route master sequence data for all networks in one onboard database, which is continuously improved each time a HORUS equipped train runs on the network,” Wust explains. “Australia has some unique challenges – we have vast distances and the overall complexity of operating trains is quite high. So we have many reasons to adopt innovative technology to improve the safety and efficiency of rail transport.
“We’re excited about the technology we’ve been developing. It offers a lot of potential to the industry.”
Visit 4Tel at AusRAIL PLUS at Stand 180.
The post AusRAIL: Enhancing train performance with Artificial Intelligence appeared first on Rail Express.
This article first appeared on www.railexpress.com.au
About this website
Railpage version 3.10.0.0037
All logos and trademarks in this site are property of their respective owner. The comments are property of their posters, all the rest is © 2003-2020 Interactive Omnimedia Pty Ltd.
You can syndicate our news using one of the RSS feeds.