Adopt or develop?

The mining industry is not dissimilar to other industries. In a bid to boost both productivity and safety, companies are actively moving towards process and software automation and applying robotic technology to mining vehicles and equipment. While some advances in automation will be unique to mining, the rise of self-driving cars presents substantial opportunities (and challenges) in achieving this reality.

Automakers, technology corporates and startups are investing on a large scale in developing self-driving car technology. This investment is driving rapid improvement in hardware – in terms of sensing technology and computational equipment – as well as in the algorithms and intelligence required to automate these platforms. These advances are already having an impact on technology development in mining, however, it is not a simple case of blindly transferring technologies from one domain – the road network – to mine sites.

So, what are the similarities and differences between the mining industry and related domains such as self-driving cars? In this article, we’ll dive a little deeper into understanding key concepts within the innovation roadmap for full automation in mining vehicles and equipment.

It’s hard to ignore what is going on in the self-driving car industry with participants such as Tesla, Google, Uber, Microsoft – to name a few – heavily competing. Tens of billions of dollars are being spent on R&D in a bid to be the first producer of a fully autonomous vehicle. Although self-driving cars are already navigating their way on our roads and have been since at least the 1980s, there are still significant technology and social challenges to be addressed before they become a widespread reality.

Some of the top challenges facing self-driving cars include the highly variable nature of the environment. Weather conditions such as snow, tropical storms and fog occur regularly and pose a severe challenge to a fully autonomous vehicle. It may struggle to sense and hence understand what is happening in the world around it. Seeing other cars, pedestrians, traffic lights and signs is problematic when falling snow or rain obscures the high-tech laser and camera sensors. Radar sees through some of this but does not reveal sufficient details about the scene (such as the text on a sign) to enable autonomous driving on its own.

Unexpected “rare” events, which occur all too frequently, also pose a major challenge. Many of the big players in self-driving cars are staking their development on technologies that learn how to drive a car from large amounts of data. However, this data does not encapsulate all possible scenarios. Most self-driving cars have never seen or been trained to deal with a person in a chicken suit running out in front of the car, yet this and similar situations occur surprisingly frequently.

Interacting with humans is a major challenge across all areas of automation. It is particularly problematic for self-driving cars. Driving at high speed down a road flanked by pedestrians only metres away involves great contextual interpretation by a human driver, who will, where appropriate, assume that the pedestrian won’t suddenly jump out in front of the car, except perhaps at a crossing. Teaching self-driving cars to reliably make these sorts of inferences is difficult, and is currently not a solved problem. In addition, in the interim period where human and autonomous drivers mix it up on roads, trust and communication are also unsolved challenges that require great thought and testing.

Finally, and especially in light of recent events around the world, cybersecurity is a concern. When vehicles can become deadly weapons in mass casualty events, the prospect of them being hacked and controlled remotely is terrifying. Although the simplification that a self-driving car is simply a “smartphone on wheels” is perhaps a little reductionist. Nevertheless, it illustrates the key point that these systems will be highly computerised and potentially linked into the cloud, making them vulnerable like all other similar systems.

Automation in mining

In many aspects, the mining industry is a leader in the automation space. The first fully automated surface haul trucks rolled out relatively recently, and are already having a major impact on productivity and safety. Mining sites are mostly removed from an urban setting and are highly regulated which makes them a stable training ground for automation and robots. The control of the number and type of people working on a mine site also clearly differentiates mining from an urban road environment. Everyone in a mine site is trained and typically well versed in safety and operational procedures.

Current mining equipment automation examples include surface haul trucks, underground load haul dump trucks (LHDs), semi-autonomous bulldozers and autonomous blast-hole drilling equipment. In other areas, only certain components have been automated such as automatic swing controls for draglines, shovel assist, proximity warning and safety systems for trucks.

Like some of their civilian self-driving car counterparts, autonomous trucks are fitted with radars, lasers, communication antennas, inertial navigation systems (INS) and high precision GPS to operate communications, guidance and collision avoidance systems. These technologies enable trucks to use pre-defined GPS courses to automatically:

• Navigate haul roads and intersections
• Move within the loading and dumping areas
• Enter the tie-down area for refuelling
• Interact with manned equipment such as excavators, graders, bulldozers and light vehicles

Future automation initiatives in mining will include real-time monitoring, better integration into a full mining ecosystem, enhanced information systems and a framework for managing the interactions with other equipment (both manned and autonomous).

Further to these requirements would be the monitoring of the health of the truck including detecting, isolation and reporting faults and monitoring changes in the environment i.e. quality of the road surface, dust, adverse weather conditions etc. Most of these activities are still carried out by the truck driver. In addition, automation in underground mines requires a system which operates without GPS and predominately in the dark. Unlike surface mining, it isn’t affected by rain and fog.

Similarities and differences

While the overall objective remains the same, successfully guiding a driverless vehicle from A to B, there are similarities and differences between an urban landscape and a mine site. The city and suburbs are more densely populated, offer multiple routes over an expansive area and have a higher level of unpredictability. A mine site, on the other hand, is a far more controlled environment, with defined routes, a relatively structured environment, is less populated and has a higher level of safety consciousness and training.

The more constrained domain of a mine site means the technologies brought to bear for enabling widespread autonomous vehicles may be somewhat different than in urban environments. It may be possible to train an autonomous mine vehicle navigation system with every possible scenario that it may encounter in a mine site, however, this goal is near impossible in a civilian setting where there is such a wide range of scenarios.

Autonomous self-driving car development is also facing a crossroads in terms of whether the technology goes to full autonomy straightaway, or to semi-autonomous where a human driver may be required to intervene in certain situations. The latter option is still in question as humans are poor at maintaining vigilance and having to suddenly intervene, as an example, to prevent a fatal collision may induce a delayed response. On a mine site, the vehicle occasionally stopping and flagging a human operator will be more feasible for many mine site processes, so this approach may gain additional traction than on the open roads.

Major differences exist in the business model as well. One particular area of difference to civilian autonomous cars is capital cost. Civilian car production cost per unit is much lower than large mobile mining equipment which can easily be in the millions of dollars. The larger capital cost means it’s more feasible to solve some of the automation problems with a higher capability sensor suite in mining than in self-driving cars.

Although it’s not desirable from a flexibility point of view, adding limited infrastructure to make automation easier is perhaps more feasible in some mine site operations, versus instrumenting up an entire city.

In the long term, a mine site offers the potential to be completely autonomous, with human personnel involved in remote operations. In that scenario, the vehicles do not have to be “perfect” from a safety perspective, they just have to be sufficiently so to make operations feasible. This is unlikely to be an acceptable situation in self-driving cars, in part because urban environments will never be free from large numbers of pedestrians moving around. Even if autonomous driving becomes ubiquitous and humans are banned from driving, they will still exist in the environments around the cars.

There is currently more money and hype surrounding autonomous vehicles in civilian environments than in mining (or other traditional big industries like defence). Consequently, most of the money is being spent in that domain and is drawing much of the top research and development talent from the field (as well as away from universities). Even as a “large” industry, it’s likely the mining domain will need to pick and choose from which technologies it tries to develop in-house, and which others it simply applies a fast follower and adoption mindset.

Conclusion

The mining industry faces a range of challenges in automating an entire mine. Some are shared entirely with other domains such as self-driving cars, and hence it is logical that mining will heavily draw upon technology and developments there to inform its own advances. However, a naïve transfer of technology will be far from sufficient, given the unique challenges that mine sites pose that do not exist in urban environments. Perhaps most importantly, the mine site offers a range of unique development opportunities. Autonomous technologies that are not easily developed in the self-driving car domain may well see major advances in autonomous systems coming from mining and propagating back into other industries.

Michael Milford
Professor of Robotics at QUT and Technology Leader at Mining3