Aptiv’s Satellite Route to ‘Democratizing’ Autonomous Tech

One central controller is the key to simpler, more compact and higher-performance ADAS and AV systems, argues CTO Glen De Vos.

Aptiv’s tech trajectory from SAE Level 1 through Level 5. (Aptiv)

Despite the myriad challenges the world faced in 2020, it was a "great year of growth” for Aptiv, the technology company formerly known as Delphi Automotive, according to senior VP and CTO Glen De Vos. He outlined Aptiv’s next steps toward “democratizing” both automated driver-assist systems (ADAS) and autonomous-vehicle (AV) systems in a recent online media conference.

Aptiv’s Glen De Vos predicts the future convergence of ADAS and AV systems architectures. (Aptiv)

Underpinning Aptiv’s growth are its technology pillars – the Active Safety Domain Controller (ASDC) and the Smart Vehicle Architecture (SVA). Three major OEMs use Aptiv’s "satellite” SVA for their AV technology today, De Vos noted, and two more are launching in 2021. He projected that SVA will be deployed on 10 million vehicles by 2025. In some cases, the OEMs will use the SVA across every vehicle in their lineups, over time, while others will target SVA use by a brand or by region.

Shift from ‘smart’ sensors

De Vos described the SVA as the “brain and nervous system” of its AV technology, with the ASDC serving as the brain and its sensor array the nerves. The “satellite” approach enables radar sensors to be up to 70% smaller and 30% lighter, while Aptiv’s software centralization simplifies over-the-air updates (OTA) – one central controller with all the software is easier to manage than a set-up where the intelligence is in the sensors themselves, De Vos argues.

“[The ASDC is] where all of the software regarding the perception system and the environmental models, the sensor fusion, planning and all of the ADAS features reside,” he explained. Under a distributed or a smart sensor-based architecture, the sensors themselves have intelligence to make some decisions. For entry-level systems like RACam (Aptiv’s integrated radar and camera device), that works just fine, De Vos said. But increasing the number of inputs or repositioning sensors across various vehicle models requires a different system.

“It's not a scalable or a flexible architecture,” he said. “What we've done with ‘satellite’ architecture is we've stripped all of the intelligence out of the sensor, anything except the most rudimentary image processing, and whatever it takes to get it back to the central controller.” This enables the use of smaller, cheaper, lighter and easier-to-package sensors.

Aptiv’s Scalable Sensing and Perception configuration enables ADAS systems scaling regardless of vehicle type or levels of automation. (Aptiv)

Shifting away from ‘smart’ sensors required tying together the SVA and the ASDC, De Vos said, both for low- and high-voltage systems. Integrating brain and nervous system, and thinking of the vehicle holistically, also helped Aptiv’s OEM customers optimize those architectures to better manage complexity, maintain and deliver required performance levels at the lowest total operating cost over the full lifecycle of the vehicle, he claimed.

“SVA, for us, remains a central theme as part of how we think about our portfolio and how we think about our technical strategies going forward,” he noted. And it’s those future technologies where Aptiv’s SVA could prove most valuable. Currently, for the five OEMs that currently use or are about to launch vehicles with the SVA, the technology offers autonomous capabilities somewhere between SAE Level 1 and the unofficial (and increasingly popular) Level 2+.

Capabilities converging

To configure a capable L2 system, Aptiv uses up to five radar sensors, a camera and potentially a connection to an HD map, De Vos said. This relatively simple set-up keeps the cost for an L2 system somewhere between $450 and $550, while L2+ is closer to between $750 and $1200, depending on the system and connectivity, he said. To get to SAE Level 3 and above, lidar will likely be needed. That ramps up cost significantly.

“When you get to Level 3, we think [lidar] is required, in particular in the forward-looking direction,” he said. “So, as we scale these systems to L3, you need to plug in that forward-looking lidar component and redundancy.” The good news from Aptiv’s perspective, De Vos said, is that it can add lidar to the L2 and L2+ systems in use today. That system architecture serves effectively as the foundation for an L3 system, he said. The step-change from Level 2+ to Level 3 could add as much as $5,000 to the total cost, at current prices.

“When you make that switch, even though significant portions of the architecture remain unchanged, you need to add a lot to the system, in particular redundancies and all of those extra measures that are required for the safety of that system when the car is in control,” he said. “That’s part of the reason why the adoption of those [OEM] L2+ programs have been pushed out a little bit, in terms of timing.”

However, Aptiv sees the costs of vehicle autonomy coming down, driven by L4 and L5 systems being introduced on commercial vehicles first. “Today, the end applications are so fundamentally different that Level 1, 2, 2+ or 3 are very differently architected than today's Level 4 and 5,” he said. “The reality is that we see them converging – but the convergence point is up for debate. Is it 2030? 2035? 2028? But ultimately they do converge.”

Aptiv has more than 20,000 engineers globally, in 12 major tech centers, but it’s the five OEMs using SVA that are really giving Aptiv time to prepare for that convergence. “Having done the heavy lifting associated with the development of these systems, we now have the experience, knowledge and understanding of what it takes to actually develop fully satellite architecture systems, integrate those systems, validate and then launch those systems and maintain them,” De Vos said. “There's no substitute for actually doing it.”