Like most things, it started small. We began using cruise control to monitor the driving speed. Then backup cameras appeared on our consoles. Soon, we were trusting sensors to help us assist in parking. And now? Currently, the Advanced Driver Assistance System (ADAS) sector is at Level 2+ autonomy and moving ever-nearer to Level 3, and the demand for it is at an all-time high. The ultimate goal, of course, is to eventually move towards complete autonomous driving, when eyes off the road offer drivers the opportunity to become just another passenger.
It was once believed that just like with most technologies, we would be able to create these futuristic self-driving vehicles through progressive improvements, evolving from level 1 to eventually full autonomy. But some of the technology giants have adopted a different approach, and have directly jumped to the development of level 4+ autonomy.
Which approach will deliver full autonomy earlier and more sustainably? OEMs and traditional car manufacturers have certainly paved the way for the technology, as they have almost perfected the implementation of ADAS (level 2+). This has also helped in making the technology accepted and popular among the masses. But interestingly, they appear to be in no rush to take the leap of faith towards the next levels, which can be pinned to multiple factors, discussed later in the blog post.
Meanwhile, technology conglomerates have skipped all the ‘hassle’ and have been investing billions of dollars into developing a functioning level 4 autonomous technology. This leads us to a very interesting crossroads when it comes to the development of autonomous driving technology.
Will ADAS evolve into fully autonomous driving?
Theoretically, it makes sense that we progress from ADAS (levels 1 and 2) and build towards higher autonomy. But practically, there are considerable roadblocks to overcome if ADAS has to serve as a bridge from non-autonomous to fully autonomous vehicles, which include but are not limited to the business models of the stakeholders involved as well as the technological gap.
Market Dynamics of ADAS and Autonomous Driving Development
Consumer technology companies continue to enter the ADAS space to offer their expertise and partnership. Meanwhile, automakers are joining forces as well. For example, many high-profile car manufacturers are collaborating to produce a purpose-built autonomous vehicle to enable their global mobility service business. Others invest in autonomous vehicle companies to harness the self-driving system offerings and integrate them into their vehicles.
Traditionally, automakers would develop technology for their own use. Having a differentiated product in the market was advantageous—and it still is. As ADAS developments continue, though, OEMs are realizing that competing in the technology field has not—and will not— be their new modus operandi. It takes billions of dollars to develop these technologies; a partnership with a technology developer makes much more sense.
Suppliers of ADAS have relied on a bottom-up approach to building autonomous technology and have been making massive strides. They already have a great business case, have been generating profits for a long time as they sell their technology to carmakers, and constantly upgrade their systems and save lives along the way.
On the other hand, technology companies are heading full steam towards building level 4+ autonomous driving solutions, but interestingly they have neither any previous ADAS technology framework nor any experience when it comes to automobile manufacturing. There is a huge technology jump when going from level 3 to level 4 that requires a completely different business model, which creates a very interesting division in the market between OEMs and tech giants.
While the approach might seem ambitious, it has yielded considerable results. For instance, Alphabet’s self-driving car division, Waymo, has been running an impressive self-driving ride-hailing service in a small suburb outside of Phoenix, Arizona for almost a year now. Recently they have announced the expansion of the commercial service to San Francisco, starting with a “trusted tester” rollout, which is exciting news.
Still, the fact remains that currently scaling the level 4 autonomous technologies on the same level as ADAS poses exorbitant costs and prolonged testing. This is one of the major hurdles for the tech companies keeping them matching the OEM ADAS’ scale of operations and actual miles on the road.
The Technology Gap Between ADAS and Autonomous Vehicles
The difference in business model and operational approach emanates from the huge differences in the technology requirements between ADAS and Autonomous Vehicles. A few notable ones include:
Sensors – For now, ADAS mainly relies on cheaper sensors like ultrasonic for parking and high-resolution radars for lane-keeping, although LiDAR is also used by some OEMs, for example, to complement cameras and high-resolution radars to take complete control over functions like managing the speed and steering control, thereby providing a remarkably safe driving experience.
LiDAR, which is more expensive than other sensors, for now, will definitely be necessary for level 3 autonomy and onwards. It will also become a norm even in ADAS, as LiDAR integration adds to the comfort of the drivers by enabling features like an automatic door opening or pothole detection and to the safety of vehicles by offering high-resolution data and sensors redundancy for example in turning situations in urban environments. LiDAR’s ability to offer better resolution than Radar makes it a key part of the sensor fusion concept that paves the road to fully autonomous driving.
The costs of sensors rise exponentially for autonomous vehicles, as the sensor set required is far more complex and requires a very high-performance output. Level 4 cars usually require a lot of these sensors since they don’t have ADAS’ scale of operations, meaning costs that are rather on the high side.
For example, a robotaxi usually has a sensor suite consisting of 25 – 30 cameras plus a host of radars and LiDARs. By contrast, a level 2 ADAS vehicle has 8 or fewer cameras, 2 radars, and, in very few cases so far, a single LiDAR. Having said that, price point concern will take a back seat as the volume manufacturing and adoption of the technology increases, and advances like MEMs based LiDARs become more commonplace, therefore decreasing the costs.
While the lower number of sensors makes ADAS data less dense, the huge volume of ADAS-equipped vehicles on the road makes up for it, accumulating a large number of real-world hours and eventually closing the gap of data collection with Level 4 technology. In conclusion, the number of sensors required for ADAS makes it cheaper to implement and scale, which is one of the reasons for its wide-scale applicability.
Processors – With the increasing requirements for processing speed in ADAS applications, processors are used for everything from building a real-time 3D spatial model of a car’s surroundings to calculating proximity and threat levels based on the environment. However, due to the length of processes of approval and qualification of any new technology in the automotive industry, processing technologies for Autonomous Vehicles are sometimes even less powerful than smartphone processors.
On the other hand, since Autonomous Vehicles have so many sensors, they need a particularly powerful processor, definitely a lot more than what is being employed by ADAS manufacturers, and an adequate power supply to run that processor. This has been one of the major concerns for a fully Autonomous Vehicle, as discussed in one of our previous blogs.
Mapping – Almost all cars today require accurate mapping functionalities to employ GPS capabilities and navigate through cities efficiently. ADAS data is also used in mapping function that stores and updates geographical and infrastructure information gathered via vehicle sensors to determine the exact location of the vehicle and other information about the traffic and other variables on the road. This function maintains the information and communicates it to system control even if GPS coverage fails.
Other companies have outsourced real-world testing and are using their customers’ driving data. Every company that’s working on ADAS is now trying to leverage and create machine-learning databases. In fact, millions of vehicles are already equipped with high-definition maps accurate to a few centimeters that get near real-time updates from crowdsourcing.
In comparison, Autonomous Vehicles, laden with sensors and powerful processors, do have more refined data. Since complete autonomous driving already relies on LiDAR data, they create highly accurate 3D maps with every drive. The point cloud on display in the video below using the Blickfeld LiDAR sensor gives one a sneak peek into the detailed nature of the data collected. But since OEMs have so many more sensors on the road already collecting data, and some of them also employ at least one LiDAR sensor in their sensor suite, no one would bet against them for catching up with the Level 4 feature set.
OEMs or Tech Giants – Who will take the lead?
For the driverless groups, it is either complete autonomous driving or bust – there is no plan B. This is in contrast to the more linear and step-by-step approach by ADAS players, in collaboration with successful car manufacturers.
While some still remain skeptical that the evolutionary approach adopted by ADAS companies can ever deliver level 4 autonomy, it cannot simply be dismissed or put completely out of question. And if OEMs do manage to achieve this, it would be ground shaking, to say the least, because that would mean that the world’s biggest technology conglomerates are left in the wake of the more stable and progressive technological advancement strategy.
The interesting thing to note is while the autonomous driving players are in the testing stage, the ADAS OEMs are already in series production, still control all the volume, all the production, and are currently the only ones that can provide a real business case. But the technological jump required from level 3 to level 4, along with the fact that OEMs are now very comfortable with what they are currently providing their customers, can mean that it will take a while until we see them moving to the next level.
The main premise is that no one can really succeed in this space alone. It takes a large amount of money to develop and bring-to-market technology, including the systems that harness it. Accomplishing this at scale takes expertise from multiple sides. While Level 4 autonomous vehicles are still a relatively distant reality, developments in ADAS are progressively shortening the distance. In the meantime, we must focus on our capabilities surrounding infrastructure and connectivity to enable a world where eventually, self-driving cars become a reality!