This company turns massive dump trucks into self-driving heavy vehicles

The company retrofits heavy equipment with both autonomous hardware and AI-powered software to ensure safe and productive operations. The retrofit model is rather clever; it doesn't limit companies to purchasing new AVs, which are costly and difficult to scale. Instead, companies can upgrade existing equipment, regardless of the manufacturer or vehicle type. Existing fleets can easily be converted into self-driving assets.

"When it comes to off-road industries, we believe there's both a greater need and a more immediate opportunity for autonomy compared to on-road," Sudhanshu Singh, senior vice president - Global Operations, SafeAI, told Interesting Engineering.

"Industries like construction and mining grapple with unsafe conditions, labor shortages, and rampant inefficiencies – and they take place in highly controlled and repetitive workflow environments, which means there are fewer surprises and edge cases to prepare for. Under these circumstances, autonomy can be put to use right away," Singh explains.

Earlier this year, SafeAI's collaboration with Obayashi Corp resulted in the conversion of a 45-ton Caterpillar 725 articulated dump truck into the world's first electric and autonomous heavy truck. Another partnership with mining contractor MACA is retrofitting a fleet of 100 very large vehicles.

We spoke to Singh about converting dump trucks into AVs and the future of autonomous solutions in the heavy vehicle industry. He had a lot to say.

The interview has been lightly edited.

Interesting Engineering: The mining industry has had autonomous equipment for more than 20 years. What does the next generation of autonomy look like? How has SafeAI contributed to it? 

Sudhanshu Singh: For decades, mining companies have looked to autonomous technology to develop safer, more efficient mining sites. But, despite high demand, widespread adoption of autonomy has stalled because the industry has been dominated by a closed system, original equipment manufacturer (OEM)-specific solutions. This approach forced companies to invest in new autonomous equipment and made mixed-fleet collaboration near impossible, making it difficult to scale autonomous deployments. As a result, less than five percent of mining vehicles worldwide are leveraging autonomy today. 

SafeAI is ushering in the next generation of autonomy—we call it Autonomy 2.0. Autonomy 2.0 is an OEM- and vehicle-agnostic retrofit solution that utilizes an advanced suite of sensors (LiDAR, radar, and cameras) and employs much more robust onboard computing power, relying less on network connections. This means companies can upgrade their existing fleets, regardless of manufacturer – with a more technologically advanced autonomous solution while also greatly reducing the barriers to entry and enabling scale at a much faster rate. 

IE: What does it take to turn a massive 180-ton dump truck into a self-driving vehicle? Could you give us an insight into how the technology works?

SS: SafeAI uses a retrofit approach. This means companies can convert—not replace—vehicles and fleets with autonomous technology, regardless of manufacturer or vehicle type. SafeAI's autonomous solution is powered by aftermarket hardware and proprietary autonomy software and utilizes onboard processing power to enable real-time decision-making. From the latest multimodal sensors to detect surroundings to perception-based localization with GNSS (Global Navigation Satellite System) to emerging AI technologies like deep neural networks, SafeAI leverages the most advanced technologies on the market to bring more effective, efficient, and scalable autonomy to heavy industry. 

In addition, SafeAI has developed the first operating system for autonomous heavy equipment for companies that want to build their autonomous applications but don't have the resources to start completely from scratch. This operating system allows any company to leverage SafeAI's proven infrastructure to accelerate their autonomous application development, saving about four years of development time and millions of dollars in R&D.

IE: How different is the algorithm from the one in use for passenger vehicles? 

SS: The proprietary autonomous software that we use is comparable to the technology used in passenger vehicles. The big difference between the two is the environment where the vehicles are deployed—self-driving cars have to be prepared for thousands of high-stakes edge cases, while self-driving heavy equipment operates in a highly controlled area.

IE: Tell us about your work with Obayashi Corp and MACA.

SS: We began working with Obayashi back in late 2020 to create an autonomous Caterpillar 725. Since then, we've completed that project and demoed it for hundreds of industry professionals and government representatives at a construction site in Japan, marking a critical first step toward bringing autonomy to the Japanese construction ecosystem. This year we announced plans to expand our work with Obayashi, along with a new collaboration partner, Siemens Corporation. We are working with Siemens to retrofit vehicles for autonomous, zero emission and connected applications that will be scaled across a fleet of construction trucks—ranging from 45 - 65 tons and operated by Obayashi Corporation—over three years.

In late 2021, we deployed our first autonomous vehicle in Australia – a 180-ton Hitachi haul truck – with MACA. After completing a successful six-month trial at the Karlawinda gold mine, we signed a partnership to retrofit a fleet of 100 mixed vehicles, creating one of the world's largest autonomous heavy equipment fleets. This is an exciting opportunity because it's a major bet on the next generation of autonomy – which relies on a versatile, retrofit approach – by one of the world's leading diversified contracting groups.

IE: To what extent do SafeAI's autonomous solutions help trucks respond? Can SafeAI completely replace human control? What can be improved?

SS: SafeAI's autonomous technology allows equipment like haul trucks, dozers, and articulated dump trucks to safely and independently navigate a project site and complete important functions. Autonomous sites will always still need people, but their roles will be different. Instead of being directly in harm's way at a dangerous working site, like a mine, employees can be upskilled and take on more strategic planning roles or deploy and monitor the equipment from a safe distance. 

IE: What is SafeAI's impact on the environment? How do its solutions help lower the carbon footprint?

SS: Overall, autonomous vehicles enable more streamlined, efficient operations that reduce emissions. But we're taking it one step further to reduce heavy industry's impact on the environment. In fact, SafeAI is collaborating with Siemens Corporation to create an autonomous, zero-emission heavy vehicle solution. While SafeAI's retrofit approach is already a form of reuse or recycle, this collaboration will help fast track adoption of sustainable technologies for heavy vehicles. With the improving maturity of technology, the total cost of ownership, government incentives, and regulations, there will be over four million zero-emission heavy vehicles deployed by 2030; SafeAI is playing a key role in making this happen. 

IE: The main advantage of a fully autonomous vehicle is to remove people from dangerous jobs or working conditions. How does SafeAI convince workers who worry about autonomous heavy vehicles replacing them? 

SS: Our autonomous technology is designed to reshape worksites, not replace human workers. Our goal is to remove employees from harm's way in dangerous industries like construction and mining and enable them to manage the site from a safe distance. Employees will always be required for strategic, planning, and equipment management roles. So far, the reception we've seen from teams on the ground has been overwhelmingly positive as they look to self-driving vehicles to automate repetitive, mundane, and often dangerous tasks and, more importantly, remove humans from harm's way. 

IE: What is the regulatory mechanism for autonomous heavy vehicles like? Is it more lenient?

SS: SafeAI operates in a number of countries (and growing!), and the regulatory specifics vary from country to country. That said, given our vehicles run in tightly controlled environments, on private land, often far from the public, we can deploy them significantly faster and at a greater scale than on-road applications. For perspective, autonomous vehicles have already been successfully operating in mine sites across the globe for almost 20 years.

IE: What are your predictions for the future of autonomy in the heavy vehicle industry?

SS: Off-road autonomy is at an inflection point. The industry knows that the technology works, and companies have seen the impact it can have on safety, productivity, and project costs. The initial experimental phase for our retrofit Autonomy 2.0 solution is over. We are now on the brink of mass adoption. With increased investment and innovation in this field, there are more options available, and at a lower capital investment than ever before. Spurred by these factors, we're about to see this technology take off in a meaningful way. Automation is already gaining serious traction to boost efficiency – the smart mining industry is expected to be worth about $28 billion by 2027 as automation continues to rise in this industry.