«  View All Posts

The Aurora Driver Development Program: A structured approach for the creation of Aurora Driver-powered vehicles

Aurora Driver | September 10, 2021 | 3 min. read

By Sterling Anderson

 

In 1978, Intel launched the first of what would become an enormously successful line of microprocessors. While these processors were impressive for their time, the secret to their success lay not in their specs, but in the common family of instruction set architectures that Intel designed for them. In the years that followed, Intel became increasingly involved in its customers’ design programs, leveraging the standard x86 architecture to help dominant computer manufacturers of the day create highly performant systems around a common standard. The result of these partnerships was an explosion in personal computer adoption and Intel’s market share.

In the same way that Intel’s chips required computers designed for them, a safe, scalable self-driving system requires vehicles designed for it. From the integration of physical components, like sensors or the computer, to the introduction of new electrical, thermal, logical systems, and the establishment of the vehicle’s communications protocol, the creation of an Aurora-powered car or truck is a complex process that requires a holistic, structured approach and a deep understanding of both the self-driving system and the vehicle. When designed around a common architecture and developed in close partnership with established manufacturers, the resulting product can be as broad ranging as it is powerful.

Over the last several years, we’ve developed a common architecture for the Aurora Driver and a common set of requirements for vehicles powered by it. Along the way, we’ve refined this process in collaboration with half a dozen OEMs, as we’ve integrated our Driver into eight distinct vehicle platforms. The result is a highly refined Driver-vehicle interface and a structured process for the design, development, and launch of vehicles designed for it that we call the Aurora Driver Development Program.

Just as Intel leveraged the power of a common CPU architecture to the benefit of various partners, the Aurora Driver Development Program leverages the “common core” of the Aurora Driver to facilitate broad, efficient adoption. When we say “common core,” we mean that the Aurora Driver’s hardware, software, infrastructure, and development tools are designed to work across all vehicle types. This commonality ensures that every learning, development, hardware improvement, and cost reduction made to the Aurora Driver benefits every vehicle it powers, which also allows for concurrent vehicle development. 

Years ago, I described the essence and value of the Aurora Driver’s common core. We’ve been less forthcoming about the process we undertake with our OEM partners to create vehicles designed for it. Here’s a glimpse into how that works, roughly segmented into five phases.

Lay the foundation

Our partnerships begin with alignment around a common set of values, purpose, and commitment. Before any development agreement is signed, we establish a common understanding of the goal (product characteristics, launch timelines, and deployment models) and the process we’ll follow to get there. This foundation of trust and common understanding is critical to the success of these programs. In turn, these programs are essential to the creation, scale production, and commercial support of safe, highly-performant vehicles.

Define & build

Early in the process, we align on the vehicle requirements that will guide the design and deployment of Aurora-powered vehicles. We design these requirements to ensure the combined vehicle-driver system meets the rigorous requirements of our Safety Case Framework and the end customer. 

Throughout the vehicle design process, we lean heavily on our platform partners’ understanding of existing vehicle architecture to guide the modifications required to satisfy these requirements. Because autonomous vehicles don’t have a human’s eyes, ears, and muscles to fall back on should something go wrong, electrical architectures and control actuators for safety-critical systems like braking, steering, and propulsion must be designed to eliminate common single point failures. With over 25,000 parts in the average truck and 15,000 parts in the average passenger car, the support and perspective of the manufacturer in incorporating these redundant systems safely, efficiently, and reliably is invaluable.

In the final step of this phase, we build prototypes of the integrated vehicle to check the interface designs, sensor placement, and other requirements and begin to test the vehicle on public roads. By the end of this phase, it’s not unusual for us to have spent tens of thousands of person-hours strategizing and cross-collaborating with the platform partner and end customers to align on these requirements. 

Refine & pilot 

When we’re satisfied with the design and performance of prototypes, we build a handful of vehicles on which we’ll test, refine, and validate the requirements through on-road testing, virtual testing, and commercial pilots with networks, carriers, and private fleets. In our trucking pilots, we pull real loads in commercially representative ways to gain experience and incorporate customer feedback into the final product. Throughout this phase, we work closely with our platform partner to understand the full scope of customer preferences, even as they relate to non-autonomy-specific characteristics. 

Validate

With customer feedback incorporated and design modifications made, we begin the production of a larger fleet of verification vehicles that we run through a complete suite of driver- and vehicle-level validation tests. In this phase, we can run several validation tests in parallel to confirm both the vehicle and the driver meet their respective requirements laid out in the safety case. Meanwhile, we work closely with our platform partner to prepare for the scale production and commercial support of Aurora-powered vehicles. 

Launch & expand

Once the safety case has been closed and production readiness assured, we push the big green button. At this phase, our platform partners kick into high gear on the production, commercialization, and support processes they’ve been perfecting for decades. Aurora, for its part, turns its attention to expansion of the Driver’s operational domain, rapidly unlocking new routes and geographies as we scale the product together with our customers.

In the coming weeks, we’ve got a few exciting updates to share with you on the tremendous progress the team’s made in executing the Aurora Driver Development Process with our platform partners.

Sterling Anderson

Aurora co-founder and Chief Product Officer