Automated driving is reshaping the mobility industry, putting software at the heart of vehicle design and development. Just as there are different levels of automated driving, so too are there different levels of the software-defined vehicle (SDV). While the industry continues to hammer out an official consensus on these, there appears no doubt that advanced driver assistance systems (ADAS) are among the defining characteristics of early applications.
“Ten years ago, premium vehicles were defined by horsepower and acceleration,” says Shawn Kerrigan, Co-Founder and Chief Operating Officer of Silicon Valley start-up Plus. “Today, competition is switching to the ADAS offering: how advanced is it, how much safer or more comfortable does it make your vehicle? These are among the aspects that will define automakers in the move towards SDV.”
Software for ADAS and autonomy
Founded by AI experts from Stanford University, Plus started life as an autonomous driving software developer targeting the trucking segment. It quickly gained fame as one of the first companies to test self-driving trucks in California. Since then it has expanded the focus and today provides the core software that underpins not only autonomous driving but also ADAS. “Some people have described it as software-defined ADAS,” Kerrigan comments. Its Open Platform for Autonomy (OPA) is powered by AI models and supports automated capabilities across different sensors, processors, vehicle platforms, and powertrains. Today the OPA is offered as a suite of software solutions.
“ADAS is one of the most important parts of the overall transition to SDVs, and one of the more challenging aspects of these features is the deep AI required to understand the world around the vehicle,” Kerrigan tells Automotive World. “That’s where we come in with transformer-based neural networks.” PlusVision is the company’s AI perception software for passenger cars and commercial vehicles with SAE Level 2+ to Level 4 ADAS. “This gives a really robust 3D view of the world around the vehicle.”
The idea is that automakers and Tier 1s can integrate these perception capabilities with their own automated driving platforms, tailored however they see fit for different trim levels and price points. The solution was unveiled in April 2024, and Plus is currently in implementation talks with various manufacturers. “The timeline on these programmes tends to be quite long, from selection through to start of production,” he points out. “It’s usually a couple of years.”
Software complexity
PlusVision modules can fuse together data produced by cameras, radars and LiDARs, and are representative of the sort of software and compute requirements of the SDV. Notably, the solution has been designed to run on central ECUs. “In the shift toward SDVs, software is growing more complex and needs to be decoupled from hardware in order to accelerate development efforts and feature over-the-air (OTA) updates,” explains Kerrigan. “We developed PlusVision as a flexible software solution with a future-proof architecture that can adapt to the evolving autonomy needs.”
As vehicles advance up the SDV levels, there is widespread agreement that they will require a shift in E/E architecture away from numerous individual ECUs coupled to single functions and towards a zonal approach. Zonal architecture consolidates electronics and functions into centralised zones, dramatically simplifying design and paving the way for OTA updates, more effective data management, and improved cyber security. In fact, some consider zonal architecture the cornerstone of the SDV.
“Vehicles might have 100 different ECUs controlling all the various functions,” Kerrigan says. “That makes it hard to reduce cost and complexity, and to update and test new functionality. With the SDV, you want to consolidate that into centralised compute, ideally having one large compute that allows you to do everything. It offers much more potential from a personalisation standpoint, as well as a good way to drive cost reduction. These are among the top reasons for the emergence of SDVs.”
While the end game is simplification, the journey is not straightforward. “There have been some high-profile reports in the news about automakers postponing vehicle launches because of delays in software,” Kerrigan says. Not only does this new approach introduce a slightly different software coding paradigm, but there are also new challenges around a single computer running software from a number of different providers. “Making this transition is certainly not without complexity,” he concedes.
Ten years ago, premium vehicles were defined by horsepower and acceleration. Today, competition is switching to the ADAS offering
Computing power is also a challenge. “AI that runs in the cloud these days is powered by immense amounts of computing. The computing capabilities on a vehicle have advanced tremendously in recent years but are still relatively limited and constrained compared to the processing power used server side.” Power and heat limits are particularly relevant when running such applications on a vehicle. Plus is working with edge AI semiconductor company Ambarella on this front, offering PlusVision on Ambarella’s CV3-AD AI domain controller family of systems-on-chip and boasting industry leading AI performance per watt. “Offering our transformer-based deep neural network to OEMs and Tier 1s on really power-efficient computer hardware enables them to build in flexible, high-performance and power efficient models,” says Kerrigan.
Up next
Perception systems like Plus’ are finding a home in all sorts of software-defined applications, and the company is equally bullish on the prospects for ADAS and autonomy. With roots in Level 4 autonomous systems for trucks, it counts big-name players like Traton and Iveco among its partners. At the same time, it anticipates strong demand for driver-in assistance features as well.
“We started out building end-to-end software for Level 4, in particular for trucking, but since then, we’ve found applications for that same advanced AI technology in vehicles with human drivers,” says Kerrigan. “It’s a great opportunity to have a really positive impact in passenger cars as well, given the big drive towards SDV and more advanced perception systems to improve safety and comfort. With solutions like PlusVision, we saw a huge opportunity to help make a positive difference and bring our technology to a broader market.”
Plus has expanded its technology offering and partnership network considerably since the company was founded in 2016, but its own journey is far from complete. “We have a bunch of exciting announcements coming up over the next several months,” teases Kerrigan.
