AWS Automotive Executive Conversations: Torc Robotics

Torc Robotics, an independent subsidiary of Daimler Truck AG, has selected Amazon Web Services (AWS) as its preferred cloud provider to handle the scale and speed needed for data transfer, storage, and compute capacity as the company prepares to deploy its fleet of next-generation self-driving test trucks in New Mexico and Virginia.

Torc will leverage AWS managed services such as Amazon Elastic Kubernetes Service for running simulation software at scale, Amazon Managed Streaming for Apache Kafka, Amazon Managed Workflows for Apache Airflow, and Amazon Simple Storage Service Intelligent Tiering to efficiently manage test data and support regulatory compliance. This integration on AWS will allow Torc Robotics to transfer massive amounts of data for log analysis of real-world tests, while also providing computational power for simulation and deep learning.

Torc Robotics CEO Michael Fleming and AWS Automotive Director Wendy Bauer explain the scope and background of this collaboration:

_

The long-haul trucking industry is undergoing major challenges – how is Torc Robotics involved?

Michael Fleming: Torc Robotics’ mission is to save lives, and we are working to make the trucking industry safer. Trucking is the backbone of our economy, handling 70 percent of all freight transported in the U.S. each year. Trucks deliver products and food to every community from across the country and around the world. However, trucking can be dangerous, involving 415,000 crashes in the United States alone. Autonomous vehicles can play a huge role in reducing the number of crashes on the road. We are developing self-driving technology that we believe will help bring everyone home to their families safely.

Self-driving, long-haul trucks bring advantages other than safety. The technology that Torc and Daimler Trucks is developing can help fill the nearly 60,000 driver shortage gap in the U.S. that carriers rank as the number 1 issue in a recent survey by the American Transportation Research Institute. Self-driving trucks will enable freight haulers to increase overall efficiency. Additionally, self-driving trucks may provide greater fuel efficiency, which is a point of focus in the industry today.

While Torc Robotics’ business is recognized publicly by many, what’s one unique characteristic or feature that is either lesser known or understood about your company?

MF: Torc Robotics is not a startup, but a 16-year-old company with a long history of profitability. Torc has successfully commercialized our technology on numerous platforms in industries including defense, mining, and transit. Now, as an independent subsidiary of Daimler Truck AG, we are working closely with this experienced trucking company to safely commercialize self-driving trucks.

As you are preparing to launch autonomous trucks – what have been major challenges that you needed to overcome?

MF: Testing software for self-driving trucks is a complex and computation-intensive task. Each time we test Torc Robotics’ vehicles on the road, we generate significant and valuable data.  This data must be transferred, stored, and analyzed. Testing scenarios in simulation also creates a large computational load. To address these significant challenges, we chose AWS as our preferred cloud provider because of their industry-leading platform and expertise in handling autonomous driving compute and data needs. As we scale our fleets and move towards production, our need for computation, data storage, and data transfer will continue to grow.

Wendy Bauer: As you know, we start with our customer’s needs and work backwards and 90% of what we build is driven by what our customers tell us matters to them. This is especially true for the autonomous driving space because this area requires lots of new innovation to solve the complex problems in the development of autonomous driving solutions. AWS supports all stages of the autonomous driving development workflow. The foundation of this workflow is highly scalable, highly available, fast, extremely durable, inexpensive data storage infrastructure provided by Amazon Simple Storage Service (S3). With Amazon S3 Intelligent-Tiering, customers benefit from automatic cost savings through intelligent movement of data across frequent access, infrequent access, and archive access storage tiers.

The large data volumes generated at a very high data rate by Torc’s test trucks can be easily ingested over AWS Direct Connect, traversing AWS global network backbone from any source location to any AWS region. Monitoring and other data from connected autonomous driving vehicles can be securely ingested and processed in real time through AWS IoT Core. Then there is the massive computational need for deep learning, building high-definition maps, and running massive simulations. We offer a broad array of CPU and GPU instances based on latest chips from Intel, AMD, and Nvidia. We also offer extremely cost-effective AWS ARM based Graviton2 CPU processors for running many of the autonomous driving workflow tasks, and Inferentia chips for accelerating deep learning inference during simulations. To help autonomous driving companies get the most out of AWS cloud services, we have a team of autonomous vehicle specialist supporting them along the way.

 MF: Along with technical innovation, successful commercialization requires forming the right relationships. For example, we believe that the best way to bring this technology to market in the safest manner possible is through deep collaboration with an OEM. By working with Daimler Trucks, we’re combining Torc’s expertise integrating and testing autonomous systems with their proven vehicles and industry insight.

Additionally, the truck chassis must be transformed to allow for redundancies in safety-critical components such as braking and steering to maintain the vehicle’s maneuverability without human intervention.  Our work with Daimler Trucks includes developing a chassis that is seamlessly integrated with the software.

We can only be successful if we also have strong relationships with our technical providers – such as AWS — where they can anticipate our needs and help us scale. One of the reasons we selected AWS was the soft skills – this is a team that fits our work style.

In the face of the current disruption to the trucking industry, we’ve observed incredible innovations coming from across the industry. How has your company innovated through these challenging times and what are you most proud of?

MF: The pandemic revealed how truly vital trucking is to every community, as more and more households and businesses moved to online shopping and door-to-door delivery. While the pandemic delayed our on-road testing in New Mexico for a few months, Torc still launched our new test facility in Albuquerque. We used the time to scale future fleet infrastructure and enhance our virtual simulation capabilities. Like many organizations that transitioned to working from home, we innovated on stronger collaboration with remote teams. This, of course, created even more constantly changing computing demand. We are proud of the teams’ continued flexibility, and that we were able to maintain our growth trajectory throughout the year.

Machine Learning is a hot topic, especially in the autonomous driving space – can you tell us a bit about how ML is used?

MF: One area is perception – the way our system “sees” the environment around the vehicle. Machine learning is used to analyze camera data to classify information such as signs and traffic control devices. Additionally, we use deep learning in localization and object detection. However, machine learning is only one tool in our software stack. We do not, for example, use machine learning to directly command the behaviors of the vehicles.

WB: Our goal is to provide the broadest and deepest set of machine learning services to our customers so they can have the right tool for the right job. In addition to working with various deep learning frameworks directly on Amazon EC2, or Amazon EKS, Amazon SageMaker ML platform accelerates deep learning with purpose-built tools for every step of autonomous driving perception system development.

What kind of impact has the cloud, and more specifically AWS on your business?

MF: We appreciate the flexibility we get from AWS to quickly scale compute resources on demand without having to build additional on-premises computing capabilities. Relying on our preferred cloud partner for computing infrastructure allows us to focus on development and helps speed our innovation process. We also appreciate the data and computation security provided by AWS. By utilizing AWS, Torc can rapidly and cost-effectively implement the complex workflows and tooling essential for commercializing our SAE Level 4 self-driving vehicle platform. We believe the AWS capabilities — including secure data transfer, intelligent tiered storage, managed orchestration and analytics tools, and high-performance multi-core CPU and GPU compute capabilities will help us scale our on-road testing program quickly and effectively. It’s the data from the testing that is critical to ensuring a safe, reliable, robust commercial effort.

How do you think about security in the cloud?

MF: First of all, safety will always be our top priority and data security is a critical component of that goal. AWS broad set of security services helps us meet our core security and compliance requirements.

WB: Torc Robotics and AWS operate on a shared responsibility model where AWS manages and controls the components from the host operating system and virtualization layer down to the physical security of the facilities in which the services operate, and Torc Robotics is responsible for securing its application.

What makes you excited about the future of trucking? What should Torc Robotics’ customers be excited about?

MF: We are excited to work with freight haulers on a technological solution that can not only solve business challenges, but also make our roads safer. We believe we can transform the trucking industry to provide safer jobs that are more consistent but still plentiful. We are most excited about being part of a team commercializing a technology that has the potential to save thousands of lives each year.

To learn more about the Torc Robotics and AWS partnership, please click here.

—

Michael Fleming is CEO of Torc Robotics, the company charged with commercializing Level 4 autonomous trucks in partnership with Daimler Trucks, the global leader in the market. After competing as a student for Virginia Tech in 2004 and 2005 DARPA Grand Challenges, Fleming founded Torc to save lives through the commercialization of self-driving technology. In 2007, Torc developed its first-generation self-driving stack for Team Tango which finished in third place. Since then, Fleming has led Torc in deploying its technology across the trucking, automotive, transit, mining and defense industries.

Fleming earned his B.S and M.S. in mechanical engineering from Virginia Tech in 2002 and 2003 respectively. He served as a research associate with Virginia Tech from 2004-2006. In recognition of extraordinary career achievement, he was inducted into Virginia Tech’s Mechanical Engineering Society of Distinguished Alumni and the Junior Achievement Hall of Fame in 2019.

Wendy Bauer joined Amazon Web Services in October 2019 and leads the Global Automotive Sales organization. In this role, Wendy is responsible for supporting the worlds largest Automotive OEMs and Tier suppliers to accelerate their digital transformation journey while maximizing value creation. Prior to AWS, Wendy held leadership roles at Microsoft as Worldwide Manufacturing & Energy Business Strategy Leader and General Manager – US Industry. In these roles, Wendy was responsible for building the Industry GTM Strategy and accountable for accelerating commercial execution in achieving both near-term business growth and strategic customer partnerships. Her management responsibilities included engaging with c-level executives, supporting their digital transformation and innovation strategies, along with maturing industry-relevant alliance partner relationships to accelerate Industry GTM execution success. Prior to Microsoft, Wendy spent over 20 years in the automotive industry in leadership roles across both OEMs and Tier 1 suppliers. Her experiences spanned Sales, Product Strategy & Business Development, Engineering, Purchasing and Quality. Her business success has come from strong focus on market leading innovations, strategic partnerships, customer obsession and operational discipline and execution. Wendy holds a Master of Business Administration degree from Indiana University – Kelley School of Business, a Master of Science in Engineering from Purdue University and Bachelor of Science degree in both Electrical and Mechanical Engineering from Kettering University (formerly GMI Engineering & Management Institute).