This year has thus far proven to be a rough ride for the “self-driving” industry. Tesla and Elon Musk have lost interest in developing a network of robo-taxis, instead shifting focus to the “Tesla Bot” while Waymo’s robo-taxi was bamboozled by a single static traffic cone.
By contrast, there is heightened awareness of the potential for advanced driver monitoring to reduce fatalities. One indication is a provision for driver monitoring systems (DMS) and other safety technology such as automatic emergency braking and crash avoidance systems included in pending U.S. infrastructure legislation.
Having studied the automotive DMS market since 2017, I have observed a recent and radical shift in sentiment towards safety technology. This responds primarily to growing recognition that passenger vehicles won’t offer any form of “self-driving” functionality any time soon.
Advances will therefore come from safer drivers, rather than from using technology to replace human drivers.
That means automotive DMS is finally heading for mass-market adoption.
More than 30 DMS companies are targeting the automotive sector, but just three—Cipia, Seeing Machines and Smart Eye—appear able to thrive
Let’s explore what I call “Key Competency Indicators” for suppliers, concluding with an assessment of the market landscape through 2026.
Optical path refers to IR illuminators and CMOS image sensors. Advanced systems operate at 940 nm, with a frame rate of 60 frames per second. IR light at 940 nm is invisible to the human eye, but is also harmful at high power. That requires understanding of, and conformity with, IEC 62471 for eye safety is vital for DMS operation.
Optimum measurement of driver eye gaze, face and head pose is obtained using an alternating, strobing illumination pattern operating at 60 Hz. That requires use of dedicated driver ICs. Future performance gains will come by replacing IR LEDs with 940-nm VCSELs (Vertical Cavity Surface Emitting Lasers).
VCSELs emit energy across a much narrower range of frequencies than LEDs, increasing optical power output at 940 nm, thereby raising the signal-to-noise ratio at the central wavelength. VCSELs also have a much faster rise-and-fall time than LEDs, which is critical for strobing illumination at 60 Hz.
Seeing Machines is the technology leader for the optical path, ahead of Smart Eye.
In an era of AI and machine learning, the company with the largest dataset wins. Facebook, Amazon, Netflix and Google prove the point. For automotive DMS, Seeing Machines possesses the largest dataset, with approximately 6 billion kilometers of naturalistic driving data captured by its “Guardian” system in just the last three years, and about 7.5 billion kilometers total.
Guardian is capturing training data at a rate of about 55 million kilometers a week, across a fleet of over 30,000 commercial vehicles. No other DMS supplier has a system to capture a comparable volume of training data.
Functional safety compliance
Functional safety is described by ISO Standard 26262 and the Automotive Safety Integrity Level (ASIL) rating. Also relevant is ASPICE (Automotive Software Process Improvement and Capability dEtermination), a standard among German automakers.
DMS performance requirements have long since surpassed simple beeps and chimes when the software detects distracted or fatigued drivers. Automakers are now investigating how to use DMS signals to provide real-time analysis of the driver’s attention state and engagement level, with the goal of varying sensitivity of the automatic emergency braking and lane-keeping systems.
Fusing DMS with the braking and steering systems to modify the vehicle position on the road substantially raises the functional safety requirements as well as ASIL specifications for DMS software.
Companies developing commercial-grade software rarely possess the expertise to achieve ISO 26262 compliance. Functional safety is thus a substantial barrier to entry even for established companies such as Apple, Google and Microsoft seeking to compete in the automotive DMS market over the next five years.
Seeing Machines is the clear leader for functional safety compliance.
Human factors research
Early DMS technology gathered raw data such as head-pose estimation (X,Y,Z) eyelid opening (a measure known as PERCLOS, or Percentage Closure), and eye blinks. More recently, precision eye-gaze vector measurement (X,Y,Z) is possible, but remains much more challenging and requires additional computational resources and advanced optical path expertise.
Advanced DMS systems have moved far beyond raw data gathering to actionable high-level signals, which provide a real-time assessment of the driver’s state. Parameters measured include visual and cognitive distraction, drowsiness, attention state and impairment.
Again, Seeing Machines is the clear leader when transition from raw data gathering to high-level signals, a function of human factors and behavioral research.
To achieve robust and precise head, face, and eye tracking at 60fps requires a system-wide design approach, including optical components, image processing and related algorithms. I call this the “imaging signal chain,” which requires both that both DMS hardware and software elements be designed in parallel to optimize performance, power consumption and price.
Mobileye is a prime example of hardware/software co-design for automotive, offering both its EyeQ range of processors with custom image accelerators and in-house image algorithms. Qualcomm seeks to replicate this model with its acquisition of Veoneer and the Arriver vision stack software that runs on Snapdragon Ride processors.
Once again, Seeing Machines is the clear leader in hardware/software co-design, with custom designed image accelerators running on its proprietary Fovio processor. It also runs on Qualcomm’s proprietary accelerators and a custom-designed neural processing unit called Occula, now licensable as an IP core.
Cipia has partnered with Mobileye to optimize its DMS software to run on the EyeQ4 processor, while Smart Eye has followed a software-only approach it calls “hardware agnostic.”
Seeing Machines supplies its DMS software to GM for Super Cruise as well as to Mercedes Benz for the new S-Class and EQS. Research suggests it will also achieve a start of production (SOP) this year with BMW, Ford and Jeep. Smart Eye has reached SoP with BMW and Hyundai-Kia; Cipia is set to start production this year with GM. Additional details are here.
Aisin, Jungo and Mitsubishi are among the only other DMS providers reaching SOP, but all three are significantly behind the market leaders in terms of advanced technology development. They will likely fall further behind.
Automakers typically look for technology partners with at least a decade of automotive R&D. Only Seeing Machines and Smart Eye fulfill that requirement for automotive DMS software. While Aisin and Mitsubishi are both established automotive suppliers, their DMS performance is considerably behind the state-of-the-art. For example, neither offers robust eye-gaze tracking.
Established eye-tracking companies such as Tobii lack automotive-specific expertise, creating significant headwind over the next five years.
Seeing Machines recently announced it is working with 16 leading suppliers. They likely include Aptiv, Bosch, Continental, Denso, Garmin, Gentex, Harman, Joyson, LG Electronics, Magna, Mitsubishi, Panasonic, Valeo, Veoneer, Visteon and ZF.
In its latest financial results, Smart Eye provided few details, but I reckon it is working with as many as eight suppliers, including Aptiv and multiple Chinese companies. Similarly, Cipia is likely working with Aptiv and Jabil as well as a number of Chinese companies.
DMS demand is being driven primarily by legislation, including both the European Union General Safety Regulations and the safety technology roadmap published by the European New Car Assessment Program (Euro NCAP) for achieving a five-star safety rating.
In the U.S., proposed legislation would use DMS to monitor for impaired (drunk) driving, while the U.S. National Highway Safety Transportation Administration has made repeated safety recommendations in its exhaustive investigations of fatal Tesla crashes. Robust DMS must be adopted both to monitor driver engagement and to prevent automation complacency.
Automakers are known to demand “military-grade products at consumer prices.” The demands of the automotive sector have ended many a career, exhausted more investors and wrong-footed more suppliers than perhaps any other industry.
When working with suppliers considering automotive, I advise them that “time-to-money” is at least a decade, and they will need investors with deep pockets and nerves of steel. Entering the automotive sector is a corporate-wide commitment that has to start from the board of directors on down. It is not for the faint-hearted.
The chart below includes estimates for automotive DMS supplier market shares in 2026. The focus is five years out since all of the major contenders will have been established by then and market shares should be stable.
Seeing Machines will be the leading supplier of automotive DMS in 2026, with a market share estimate in the range 60-65 percent, ahead of Smart Eye with 20-25 percent. Cipia ranks third with 5-10 percent, primarily as the result of its work with Mobileye.
Jungo and Xperi round out the top five. I expect both to be acquired before the end of 2021, possibly by chipmakers such as Mobileye, Renesas or Texas Instruments. Or perhaps by a software company such as Tobii.
While these names may be unfamiliar, you are extremely likely to drive a car featuring their safety technology before the end of the decade.
While the future market shares for DMS suppliers cannot be known for certain, one forecast I am very confident in making is that DMS technology will have more of an impact on our everyday lives than the Tesla Bot. It’s hard to fathom precisely what Elon Musk is thinking. As noted, Musk is a performer. Tesla Bot may be part of his comedy routine.
>> This article was originally published on our sister site, EE Times.
|Colin Barnden is principal analyst at Semicast Research and has over 25 years of experience as an industry analyst. He is considered a world expert on market trends for automotive vision-based driver monitoring systems (DMS). He holds a B.Eng. (Hons) in Electrical & Electronic Engineering from Aston University in England and has covered the automotive electronics market since 1999.|
- Vision-based driver monitoring systems gain traction
- Breaking down misconceptions about driver monitoring systems
- The fundamentals of facial recognition
- Understanding AV complexity
- The key technologies moving us closer to driverless mobility
For more Embedded, subscribe to Embedded’s weekly email newsletter.