Full implementation of driverless cars is still a long ways away. But what about other autonomous vehicles?
Australian mining company Rio Tinto has taken the leap to develop completely autonomous vehicles that enhance their mining operation.
The AutoHaul train. Image from Rio Tinto
Rio Tinto has taken the concept of autonomous vehicles and used it to streamline rail operation. In the process, they've created what they call the world’s largest robots.
What Is AutoHaul?
At the end of December, Rio Tinto announced that the company had successfully completed the rollout of the first autonomous heavy-haul rail operation. The $940 million dollar project titled AutoHaul comes after six years of development by Rio Tinto and technology supplier Ansaldo STS, joining a suite of other automated advancements.
Rio Tinto first announced they would begin conducting tests back in February of last year, and began testing the autonomous rail in July. Rio Tinto says since completing their first test, they have traveled over 600,000 miles (1 million kilometers) autonomously.
The AutoHaul program uses 200 locomotives and runs them on more than 1,050 miles (1,700 kilometers) of track. Rio Tinto uses the rail system to transport ore from 16 different mines to four port terminals in the Pilbara region of Australia.
A Problem of Sensor Systems
As with any form of autonomous system, data and the ability to gather it quickly are key to successful implementation. In this case, however, the scale is truly massive.
Along with the 1000+ miles of track and the 16 mines, Rio Tinto also has three different ports and three power stations that it owns and operates. This means the company is capable of generating millions of terabytes of data a minute from various mobile equipment, sensors, and CCTVs, all of which provide real-time data to the control room operating over 900 miles away from their physical sites.
According to former CEO Sam Walsh, who spoke at the Data & Analytics Summit in Sydney last year, “Today, much like an air traffic controller, it has many moving parts, and looks like NASA's control center.”
The control center for AutoHaul. Image from Rio Tinto
If the technology and software used for the train operation is anything like that used in the running of autonomous vehicles, then the three main components that would be key to this success are:
- Software/control algorithms
The sensors include features like GPS monitoring, radar, cameras, and ultrasonic technology. While these are most important for allowing the train to sense its environment, it’s also crucial for sensors to be stationed at depots and at places along the track. The trains are also fitted with onboard cameras, allowing the company to monitor the trains from the inside. Public crossings are also fitted with CCTVs along the route, providing further and continuous monitoring.
The connectivity aspect relies on the data gathered from the sensors and other monitoring systems and using it to communicate. In the case of Rio Tinto, this would mean getting the data back to the control center.
Software and Algorithms
Last are the algorithms and software used decision making, such as machine learning software, which plays a major role in the automation of this new rail system. The specifications of this system are, unsurprisingly, a tightly kept secret. However, Rio Tinto did comment that the major challenges to the project lie within software and communication, but their technology supplier Ansaldo STS is working on it.
Urban Applications: Mixing Cars and Rail
Rio Tinto isn’t the only company using autonomous rail systems, Siemens has begun rolling out their newest autonomous technology using trams to drive their research. The company partnered up with ViP Verkehrsbetrieb, and in 2018 began testing the world’s first autonomous tram against real-life road traffic in Potsdam, Germany.
Siemens's autonomous rail system in Potsdam. Image from Siemens
Like other autonomous technology, Siemens uses radars, cameras, and machine-learning software. These allow the system to interact with pedestrians, cars, and other obstacles that a tram would come across in an urban environment.
As of September of last year, Siemens had conducted the test 450 times without passengers. Unlike Rio Tinto, who implemented their autonomous rail system for commercial use, Siemens says “The current project aims at identifying the technological challenges of autonomous driving under real-life conditions, then developing and testing solutions for them.” The goal is to understand how this technology interacts with obstacles in real life urban scenarios which will help them to better understand and develop autonomous technology in the future.
This project thus combines the large-scale concerns of transportation systems on static rail lines with the high-pressure challenges of operating in an urban environment full of changing obstacles.
Autonomous Railways vs. Autonomous Cars
While autonomous cars share many similarities with autonomous rail systems, they’re in many ways also worlds apart. Siemens’s urban tram system bridges the differences between autonomous rail and cars in many ways but autonomous rails and cars are inherently different.
Most notably, the track in the rail system is predetermined. This removes some of the obstacles that a car would have to deal with such as staying in a lane, predicting the movement of the car in front or behind it, making decisions on stopping, etc.
On the other hand, Rio Tinto’s AutoHaul program is designed for heavy-load ore hauling and not passenger hauling, meaning there is less risk regarding passenger safety. By the same coin, however, the heavy loads and high speeds of ore-carrying trains make them more dangerous by some measures. For example, a major concern that divides cars from rail is stopping ability as it is virtually impossible for a heavy-load train to stop at a moment’s notice.
Also of note is the rail’s seclusion through hundreds of miles of largely uninhabited Australian wilderness. Where an autonomous car would ideally have connectivity to a sensor system integrated into a smart city, the rail is more isolated in terms of external sensors that can inform it of issues on the rail. For example, a traffic system in a smart city may be able to gather data and alert autonomous cars to environmental factors. Embedding the same level of sensor intelligence into over a thousand miles of rail is unrealistic, putting much more pressure on sensors on the trains, themselves, to sense obstacles and react to them appropriately.
Depending on who you ask, autonomous vehicle technology is right on the cusp of domination or still has a long way to go before it can be fully autonomous—or will never work at scale.
However, Rio Tinto is applying this technology and companies like Siemens are working to understand how autonomous vehicles interact with obstacles. Regardless of where you stand on the big-picture issue of autonomous vehicles, that’s progress.