The December 2022 issue of IEEE Spectrum is here!

Close bar

Mentor Graphics Moves Into Automated Driving

Electronic design automation company launches real-time sensor fusion platform for fully self-driving cars

3 min read
A black button with a glowing blue halo that reads 'Level 5 Self-Driving.'
Photo: Mentor Graphics

It’s been a busy week for Mentor Graphics, the electronic design automation and embedded systems company. Last Thursday, Siemens completed a $4.5bn acquisition of the Oregon-based firm and today it is launching into the crowded field of automated driving, with a Level 5 self-driving platform that it says is faster, simpler and up to ten times less power-hungry than existing systems.

Advanced driver assistance  systems (ADAS) in production cars today use devices like radars, ultrasonic sensors, and cameras. Most of these sensor modules contain microprocessors that process and analyse raw data from the sensor, before passing some of it on to separate modules for cruise control, lane keeping, collision warning, and so on. A few, like Mobileye’s cameras, do almost everything, from gathering raw data to outputting results.

Either way, says Glenn Perry, vice-president and general manager of Mentor Graphics’ Embedded Systems Division, the scheme is inefficient and does not scale up well to fully driverless Level 5 automated driving: “The little microcontrollers in the sensors introduce latency, increase power consumption and cost, not only in the original car but in replacements if you have a collision. And at the end of the day, only a subset of all the data that the sensor saw gets transmitted to the ADAS module.”

Mentor’s solution, called DRS360, is for the car’s sensors to pump their raw data straight to a central platform that then fuses them into a single view of the world. The automated driving algorithms can then all work on the same, complete data set. As well as reducing the number of electronic control units (ECUs) in the car—which can number 150 or more in luxury vehicles—Mentor says its system uses an order of magnitude less power.

Mentor is already a big player in the automotive world, as a supplier of infotainment and display systems intellectual property and design tools. Auto makers used to make their own proprietary operating systems for in-car entertainment but Linux-powered devices are now found in about half of all models. Mentor believes that automated driving is due for similar simplification and consolidation. “I don’t think anyone believes you can get two hundred ECUs in a car today down to one,” says Perry, “But there’s a notion that you might be able to get it down to five or six, each of those focused on a domain.”

Mentor’s DRS360 system, which it hopes to have driving a development ‘mule’ vehicle later this year, comes with only basic proof-of-concept automated driving algorithms at the moment. Mentor expects carmakers to either integrate their own algorithms—or license them from companies such as Waymo.

But despite a growing enthusiasm for centralized automated driving systems from companies like Mentor, Nvidia, and Intel, the market for them is still unproven, according to Ian Riches, director for Automotive Electronics at research firm Strategy Analytics. “Adding a Mobileye-based camera system is a comparatively easy task for an OEM,” he says. “Developing a DRS360 platform is a much larger undertaking. While large parts of the ADAS product line-up remain optional features, you can see the sense in maintaining a distributed approach. However, there comes a tipping point when fitment is so ubiquitous that centralization could bring advantages in terms of cost, performance, innovation, and differentiation.”

Start-ups have known this for years. Almost all the small companies developing their own autonomous driving systems use centralized fusion of raw data, says Eric Gonzalez of Udacity, an online education provider that is cosponsoring an open-source self-driving car competition.

Karl Iagnemma, CEO of Nutonomy, which is currently testing its vehicles in Boston and Singapore, agrees: “What they are proposing is not new at all. That’s how data fusion should be done. In practice, some groups get lazy, or can’t get access to raw data, but it’s well known that this is suboptimal.”

The question now is who can make the first must-have technology platform for automated vehicles. Mentor is betting that its pedigree in designing rugged, practical systems for production cars will give it an edge. It also helps, says Perry, that Mentor is relatively low profile. “The big carmakers are extraordinarily paranoid about exposing their data to companies like Google,” he says. “They think that if Google gets its hand on it, it’s gone.”

The Conversation (0)

Why Functional Programming Should Be the Future of Software Development

It’s hard to learn, but your code will produce fewer nasty surprises

11 min read
A plate of spaghetti made from code
Shira Inbar

You’d expectthe longest and most costly phase in the lifecycle of a software product to be the initial development of the system, when all those great features are first imagined and then created. In fact, the hardest part comes later, during the maintenance phase. That’s when programmers pay the price for the shortcuts they took during development.

So why did they take shortcuts? Maybe they didn’t realize that they were cutting any corners. Only when their code was deployed and exercised by a lot of users did its hidden flaws come to light. And maybe the developers were rushed. Time-to-market pressures would almost guarantee that their software will contain more bugs than it would otherwise.

Keep Reading ↓Show less