A winding road to an autonomous driving future

Why the autonomous driving journey will start with self-driving taxis

Today, as we see automakers and self-driving technology companies forming alliances, we can't help but wonder, what is the significance behind the alliances? What changes will the formation of alliances bring to our future? Will the actual implementation of self-driving technology and the time for self-driving to be truly accepted by society exceed our expectations? Is the investment cost of self-driving technology beyond the scope of what a single organization can bear? Are the formation of these alliances driven by the government and the public's need for regulation of the self-driving industry, or are they driven by the manufacturers' need for unified development standards?

The exploration of answers to the above questions may involve multiple factors. Let us take this opportunity to take a broad view and talk about the positions and views held by Intel and Mobileye on autonomous driving.

The three elements of "Automobile-Technology-Artificial Intelligence"

First, let’s start with the three elements of “cars-technology-artificial intelligence” and discuss:

1. Advanced Driver Assistance Systems (ADAS)

2. Self-driving taxis will be the future of Mobility as a Service (MaaS)

3. Mass production of autonomous passenger cars

When using ADAS technology, the driver still has control of the car, and the role of ADAS is to prompt the driver when necessary to prevent accidents. These prompts are extremely important because driver distraction has been on the rise in recent years.

The Society of Automotive Engineers (SAE) divides autonomous driving into several stages, L0-L5. ADAS is currently at the L0-L2 stage, and its purpose is to reduce the possibility of accidents to an infinitely low level. This key stage of "automobile-technology-artificial intelligence" is currently progressing smoothly. The current penetration rate of ADAS technology is about 22%, and it is expected to rise sharply to 75% by 2025.

At the same time, the "automobile-technology-artificial intelligence" autonomous driving is being differentiated into two stages: the autonomous taxi mobility as a service (MaaS) stage and the mass production stage of autonomous passenger cars. We have noticed that the mentality of many companies in the automotive industry has changed, and they have begun to realize that these two stages cannot be carried out at the same time.

The mass production of autonomous passenger cars (L4-L5) must wait until the deployment of self-driving taxis is complete and mature. This is determined by three factors: cost, regulation, and geographic scale. It turns out that it is difficult for us to meet all three factors at the same time in a short period of time, which is an important reason why industry insiders have begun to think about the best way to achieve mass production of autonomous passenger cars. Many industry leaders realize that if the deployment of fully autonomous vehicles (AVs) is targeted at the field of self-driving taxis first, it may pave the way for the mass production of autonomous passenger cars.

Cost: A self-driving system (SDS) equipped with cameras, radars, lidars, and high-performance computing costs tens of thousands of dollars, and this cost will remain high in the short term. This cost level is a bit expensive for mass-produced passenger cars, but it is acceptable for self-driving taxi service providers. The cost of SDS should not exceed a few thousand dollars, that is, its expected cost should be an order of magnitude lower than the current cost level, only in this way can we achieve the real mass production of self-driving passenger cars.

Regulation: Companies working on SDS development know that regulation is the most difficult issue in the implementation of SDS, and regulation is an area that has received little attention. According to the law, driver's licenses can only be issued to human drivers, and the question of how to balance the safety and practicality of autonomous driving technology in a socially acceptable way has not been resolved.

Compared to private self-driving passenger cars, it is much easier to formulate laws and regulations for self-driving taxi fleets. Self-driving taxi operators are usually issued limited licenses for individual use cases and regions, and operators are required to submit extensive reports to the competent authorities and obtain back-end remote operation constraints. In contrast, if such cars are issued licenses to ordinary citizens, it will require a thorough reform of the current complex laws and regulations for cars and drivers.

The automotive industry has gradually realized that autonomous driving can only be truly implemented after regulations and technology reach a balance.

Scale: The third element is geographic scale. Creating high-precision maps with great detail and accuracy, and keeping them updated, is the main challenge of geographic scale. Geographic scale is critical to the mass production of self-driving cars, because self-driving cars can only fulfill our promise of the autonomous driving revolution if they can work "everywhere". Self-driving taxis can operate in restricted geo-fenced areas, which can postpone the problem of scale until self-driving taxis are mature.

When we take into account factors such as cost, regulation and scale, we can understand why self-driving taxis must mature before mass production of self-driving passenger cars can be achieved.

The automotive industry is increasingly focusing on L2 products, and this phenomenon is becoming increasingly evident. Enhanced ADAS, where the driver is always in full control of the vehicle, will help autonomous vehicles achieve the various advantages of safety and security without the challenges of regulation, cost and scale.

In the meantime, automakers are addressing regulatory, cost and scale challenges by entering a new industry: robotaxi mobility-as-a-service. Once robotaxi mobility-as-a-service has a certain level of acceptance and maturity, automakers can move to the next and most transformative phase of autonomous passenger vehicles.

Intel and Mobileye's autonomous driving strategies

With this in mind, Intel and Mobileye are focused on exploring the most effective path to achieving autonomous passenger vehicles. This path requires long-term planning, but the rewards will be huge for those who can make the significant investments needed to plan for the future. Our exploration will focus on four key areas:

Continue to be at the forefront of ADAS technology research and development. Intel and Mobileye are saving lives with ADAS technology while also having dozens of new production programs each year. In these programs, automakers often put our technology through the most rigorous safety testing, which allows us to verify the technical reliability of our self-driving cars. Currently, Intel and Mobileye have more than 34 million ADAS-equipped cars on the road, which also provides financial "fuel" for our long-term autonomous driving development projects.

Designing a camera-centric autonomous driving system. Building a robust autonomous driving system that is completely camera-driven will allow us to accurately understand what redundancy is necessary in radar and lidar. This will effectively avoid "over-engineering" or "sensor overload", which is also the key to reducing costs.

Road Network Collection Manager™ (REM™) enables crowd-sourced HD map production and construction to solve the scale problem. Based on existing collaborations with automakers, Mobileye expects to have road data sent back by more than 25 million cars by 2022.

The regulatory issue can be effectively addressed with the Responsibility-Sensitive Safety (RSS) model, which balances the feasibility and agility of the machine driver's driving behavior with the social norm of cautious driving.

Intel and Mobileye are working together to seize the global opportunity of self-driving taxis. We are advancing the technology development of the entire self-driving taxi experience, from calling a car, to powering the vehicle, to monitoring the fleet, and we are doing as much as possible in the process. This will allow us to maximize our efforts in the self-driving taxi stage so that we can provide high-quality solutions to automakers when the time comes for mass production of self-driving passenger vehicles.

In the process, we will help our partners realize a safety revolution that saves lives with ADAS technology. We believe that this revolution has great practical significance and will become a powerful historical example in the human journey of exploring autonomous driving.