Regulation is driving demand for driver monitoring system (DMS) across Europe. In pushing back the technical boundaries, Cipia (formerly Eyesight Technologies) has formed several partnerships (Mobileye, OmniVision and Ambarella) to bring DMS to mass market in a way that fits the needs of the OEM based on a specific vehicle model stage in its lifecycle: planning or production. Yet in-cabin sensing has a lot more in store beyond driver monitoring, and will evolve with autonomy and mobility services. Tal Krzypow, vice president of product management at Cipia gave us his vision of DMS.
Could you tell us a little about Cipia?
Cipia is a computer vision AI leader specialising in in-cabin sensing, which refers to Driver Monitoring System (DMS) and Occupancy Monitoring System (OMS). Our DMS employs advanced computer vision machine learning algorithms and AI in order to analyse the visual cues of the driver. A second layer of algorithms translates the visuals using human factors science to establish the physiological state of the driver in terms of drowsiness, distraction, identity, driver actions and other elements.
To date, we are already working with four automotive manufacturers to implement our solution in production lines, and are proud to have vehicles featuring our technology on the road. Cipia also caters to commercial fleets through telematics service providers. We offer both a software solution for integration and a standalone hardware device which make it easy for telematics service providers to offer their fleet customers a driver monitoring solution.
The idea of a car watching the driver is not new. We have seen driver alert systems being offered by a number of OEMs using different guises. What is your USP? And can your solution detect when drivers wearing face masks become distracted or drowsy?
Indeed, there are multiple technologies out there that monitor drivers. For example, there are indirect measures such as tracking the steering wheel movement or the lateral position in the lane through the ADAS camera and detecting patterns which are indicative of drowsiness. The challenge with such technologies is that they are indirect. The statistics are very clear with respect of the causes for accidents, with distraction and drowsiness in the lead. This means the real reason is not vehicle behaviour, but driver behaviour and the best way to prevent accidents is to monitor the cause – the driver. By the way, while drowsiness is known to be monitored in an indirect manner, we’re not familiar with indirect monitoring for distraction. It seems there is a consensus that the best way to monitor driver distraction effectively is by tracking the driver’s gaze.
Cipia is not a new computer vision company and previously operated in the consumer electronics industry. However, we’ve always been an embedded computer vision company, meaning we run locally on the device in real-time, and heavily optimise for lean hardware requirements. This know how and expertise serve us well today when we can offer to our automotive customers lower total system cost, thanks to the lean nature of our software.
Face mask is something we’ve accounted for even before Covid-19. They were already widely used in Asian markets, and cover a significant part of the face. From the get-go, we trained the algorithms to cope with faces occluded by a face mask. Naturally, with the outbreak, we gave this aspect an extra emphasis by training the system to cope with more face mask types. Regardless, since the eyes are the most important aspect while assessing drowsiness and distraction, a face mask does not impede the ability to assess this important accident-causes.
We understand that Cipia has struck several strategic partnerships to bring DMS to mass market. Could you tell us a little more about these and how each fits the needs of the OEM based on a specific vehicle model stage in its lifecycle: planning or production?
During the planning stage of a new production programme, the OEM has the flexibility to spec any hardware they desire. The partnership we have with Mobileye is ideal for that lifecycle stage. First and foremost, the integration of the DMS software on and ADAS chip means that there is no need for a dedicated DMS ECU. Moreover, Mobileye’s EyeQ is not just any ECU, but an ASIL B rated ECU, ensuring the hardwares fit for functional safety requirements. It also means that the DMS is already integrated on the SoC, tested and validated, further saving the integration time and cost. However, beyond these commercial and technical benefits, we are discussing two systems that belong together.
At L1, with the availability of Adaptive Cruise Control and Lane Keeping Assist, detection of distraction or drowsiness can lead to automatically triggering the ACC and LKA, keeping the driver and those around them safe, until the driver refocuses on the road.
At L2 with systems such as Traffic Jam Assist or Highway Driving Assist, the automated functions spare the need to use the pedals and steering wheel, but they do not spare the driver of their responsibility. TJA and HDA do not deal well with a vehicle that abruptly cuts into your lane, nor do they cope with the loss of lane markings. The driver has to be attentive at all times in order to safely enjoy these systems, and a DMS is a natural answer.
At L3 the vehicle can drive on its own an entire segment of the road, let’s say the highway. However, the highway is bound to end eventually, and the vehicle has to hand off control back to the driver. The question is whether the driver is cognitively available and ready to assume control, which can only be verified with a full-fledged DMS. The tight link between the ADAS and DMS is clear throughout these levels of automation and also highlights the benefit of having both systems reside on the same safety grade chip, with minimal latency and highest efficiency. The future holds additional benefits with the systems feeding each other, making the DMS aware of the external conditions, and the ADAS aware of the driver’s state, for the safest and smartest driving experience.
Certain vehicles are already in production yet will be required to add driver monitoring system in order to comply with European regulation. The partnership with OmniVision is especially suited to address such need. The OAX8000 is an ASIC that fits Cipia’s entry-level DMS for compliance like a glove. Thanks to Cipia’s expertise in neural architectures, we were able to leverage the Neural Processing Unit on the device to achieve the highest level of efficiency. With the competitive position of the chip and its small form factor, OEMs can enjoy a solution that can be integrated into the camera design. Consider it as a “DMS in a box”, with a small footprint, which can be added with minimal trim changes. This is ideal for a vehicle in production where the desire is to affect a minimal change. This is also a great choice for an OEM in the planning phase who is seeking a standalone DMS solution, rather than one integrated with ADAS or the IVI.
Obviously, there are already plenty of vehicles beyond production, on the road – around 1.4 billion of them. While most of these are not required by regulation to be fitted with a DMS, we see very strong interest with commercial fleets. As an example, for carriers of cargo, the adverse effect of each accident is multiplied compared with a private vehicle – it’s not just about, god forbid, injury or death. When a truck or delivery van are involved in an accident, then the cargo may be damaged, the service level is hurt (bearing an associated penalty), the truck has downtime which affects the fleet’s overall schedule, insurance cost rises, etc. The need for DMS in fleets is evident. However, fleets have other video needs as well, for both safety and security. Among our customers who are building DMS devices for fleets, there is a clear preference for Ambarella and especially for the CV family of chips, which are an ideal video platform for safety and security. Cipia’s efficiency on this platform means that there is plenty of room left for additional video capabilities beyond ours, on the same SoC.
As the automotive industry shifts toward higher levels of driver autonomy, what are the opportunities for your business?
The need for DMS increases with the rise in autonomy up to L3. At higher levels, where we can truly call the vehicle autonomous (rather than automated), there will be a clear shift from focusing on the drive to focusing on the ride. The vehicle will transform into a moving lounge on wheels. We already see OEMs interested in Occupancy Monitoring System (OMS) for safety – detecting passengers, their seating position, usage of seatbelts, presence of children, etc. However, as autonomy level rises, OMS will shift from safety to convenience. Whether in a mobility service, or in a privately owned vehicle, understanding who is the audience inside will be a key for a better experience and for tailoring offerings and services for the right people in the right time. The identity, demographics, mood, activities and other aspects of the occupants in the vehicle will serve to personalise the experience in a way that will benefit both the occupants and the service providers. OEMs are very keen today to develop new business models which are based on services and data, to monetise the ongoing usage of the vehicle rather than just the car sale and the annual maintenance. In-cabin sensing is a key enabler for that goal.
Driver monitoring for fatigue and distraction has become a major focus of automotive safety regulators and governments worldwide. This trend looks set to continue in SAE level 2 (partial) and 3 (conditional) semi-autonomous vehicles. To what extent are regulations driving increasing demand for DMS (Euro NCAP, EU GSR, SAFE act of 2020)?
There is no doubt regulation drives demand. In all the RFQs there is a clear reference these days to the regulations. However, OEMs expect more. Once a feature has to go in because of regulation, it is not a differentiating feature, but a tax from the OEMs perspective. Therefore, they are keen on getting more from the system, to create added value that can be used for new experiences and differentiating models – within their own lineup and versus the competition. Facial identification of the driver, using the driver’s gaze for innovative HMI and other aspects.
What can you offer fleets?
Beyond drowsiness and distraction, we added additional capabilities which are sought after among fleets, such as detecting cell phone usage, smoking and wearing a seatbelt. Even driver recognition is a desired feature to ensure only authorised personnel drive the vehicle.
Cipia does not sell to fleets directly, since fleets typically consume fleet management services. They do not integrate DMS software or buy only a device. Our direct customers are the Telematics Service Providers (TSP) who provide a full system, from the in-vehicle installation to the fleet manager’s dashboard. To TSPs we offer our Cipia-FS10 – a DMS device that provides the driver with life-saving alerts, and also provides the fleet manager better tools for managing the fleet, with insights about driver behaviour for training, workforce management, etc. For device manufacturers who wish to build their own device, such as a video platform for fleets, we also offer our DMS software (similar to the one offered to OEMs). The benefit is we have experience working with fleets and TSPs, and adapt our software to cater to the specific needs of fleets.
I guess that in-cabin sensing has more potential than driver monitoring, Consumer technologies such as fitness trackers have been popular for some time, monitoring our heart rates, performance and sleep, but how might the renewed focus on wellness translate to the automotive space?
Wellness is a topic that interests OEMs as they explore potential ways to improve the in-cabin experience. Heart rate, breathing, the level of fatigue, mood, etc. are perceived by OEMs as potential input for systems which will control temperature, lighting, sound, haptics and even scent. However, all these are in early exploration, and the end-use cases are not clear yet. It is also likely that these inputs will come from a variety of sensors, such as the personal smartwatch, the voice assistant, internal radar and, of course, vision.
What’s next for Cipia?
Cipia is currently focused on the DMS. RFQs are changing from hundreds of thousands to millions of vehicles per year, particularly in light of the regulation and safety standards. This is an exciting period, and we are looking forward to seeing our partnerships with Mobileye, Ambarella and OmniVision helping to bring our solution to the mass market. In addition, we expect to see a transition in OEMs interest with respect to OMS as they switch from RFIs to RFQs.