As written in the MIT Technology Review,"understanding where we are in the pursuit of self-driving cars can be as confusing as understanding where we are in the pursuit of AI." Autonomous vehicles are among the most demonstrative innovations enabled by AI.
More complicated than sectors like Finance (highly digitalized with structured data) and Retail/Manufacturing (operating AI in restricted areas/ closed environment), autonomous driving requires complex AI capabilities from the fundamental infrastructure to deep integration with the auto industry and related industries.
Unlike the basic recognition of 2-D texts and individual 3-D objects for other AI applications, the technological requirements for autonomous cars are more comprehensive. For example, face recognition works highly effectively because it is viable for machines to capture the whole face's features. No matter how different each face may be, human body parts and facial parts are fixed. But things are different for autonomous driving because the environment is changing and unpredictable; a minor failure could cause destruction.
The deeper the research goes, the harder it seems to be to achieve fully autonomous safe vehicles. In 2017, the auto industry predicted great progress to be achieved in five years. In 2018, people tended to believe substantial technological breakthroughs are hard to realize, so big advancement would be likely to happen in 10 years. This year, the attitude has changed to “at some point in the future.” It takes time to gather feedback and remodel, and it also requires time for other technologies, those of sensors and electrical systems, to fuel AI development.
Ethics and policy challenges follow along with technical issues. For example, an ethical quandary about how AI should act when getting involved in an inevitable accident overhangs development. Similarly, it can be hard to determine who to blame when an AI system does not manage to avoid some collisions.
The transformative technological development of autonomous vehicles can easily interrupt other industries as well, such as insurance. The liability policies for the insurance industry are likely to change, involving a re-think of what body injury, property damage and collision mean in the autonomous era. The potential buyers and insurees are also subject to change from car owners or car drivers to system providers or car manufactures.
Moreover, due to the reduction of accidents, revenues to the insurance businesses will drop significantly. According to Autonomous Research, 42% of revenue from global insurance premiums come from car insurance. This disruptive technology is likely to bring an enormous economic shock.
There will be a lot more obstacles than previously expected.
However difficult it may seem, autonomous driving in some form is the inevitable future, as the auto industry is getting increasingly intelligent. Currently, there are several standards for the level of vehicle automation. According to SAE, there are six levels from Level 0, no automation, to Level 5, full automation, based on execution of steering and acceleration/deceleration, monitoring of the driving environment, the fallback performance of dynamic driving tasks and system capability.
Several social and economic benefits are empowered by autonomous driving.
1) AV reduces traffic accidents and the resulting casualties and/or costs due to errors caused by human operation. According to the World Health Organization, 1.35 million people were killed on roads in 2016, and a 2013 study demonstrated that human error is the biggest cause of accidents. Reducing or removing human error from driving seems to be the best way to reduce road accidents.
2) AV provides a convenient way for vulnerable groups (elderly, disabled) to travel. The adoption of autonomous cars will change public transit. Currently fixed-route, fixed-schedule public transportation can be upgraded to point-to-point, on-demand services that will largely improve the convenience for the elderly, vulnerable and disabled individuals who need assistance in moving.
3)AV saves labor and time costs, reducing traffic congestion, improving travel efficiency, and saving parking spaces in the urban area. The average commute time in 2018 reached a record high, according to data from US Census Bureau, with 4.3 million workers with commutes of 90 minutes or more in the US. The metropolitan areas are facing severe inefficiency caused by congestion and insufficient parking space. Relieving people from driving, autonomous cars provide extra time and space for people to engage in production and thereby contribute to economic gains. Besides, instead of parking cars near apartments and offices, autonomous vehicles are able to send themselves to low-density areas.
Despite the discussion on the timeframe to achieve it, autonomous driving research institutes and companies never cease to attract investment in R&D. Between 11/01/2018 and 11/01/2019, 394 AI companies were established after 2011 attracted USD 28 billion from venture investors. Among them, 30 autonomous driving-related startups raised one fourth of the total amount.
In the meantime, governments release and draft policies and rules, mainly to resolve the current challenges of road testing. European countries are especially supportive of public transit. Besides, countries are also preparing infrastructure to welcome this new technology.
The Netherlands, Singapore, Norway, the US and Sweden are the top five among the 25 countries that are best prepared for autonomous driving in terms of technology, policy, infrastructure and consumer acceptance, as measured by KPMG 2019 Autonomous Vehicles Readiness Index.
ADAS (Advanced Driver Assistance Systems) and V2X (Vehicle to Everything) define the internal and external requirements for autonomous driving. Currently, ADAS (~L2) is the main revenue driver for passenger car-focused autonomous driving startups in the short run.
The global ADAS market is expected to see a CAGR of 18.1% from 2018 to 2024. The penetration rate is about 8% to 12% for developed vehicle markets in the EU, the US and Japan. China’s ADAS penetration is around only 2% to 5%, according to Gasgoo Auto Research Institute, the top three applications among which are Blind Spot Detection (BSD), Driver Monitoring Systems (DMS) and Automatic Emergency Braking (AEB).
Currently, revenue flows for autonomous driving companies are mostly from installing ADAS systems. Demand for intelligent vehicles is driven by consumer needs. For most consumers who tested functions in a smart driving system, they crave other functions and rarely switch back to the manual modes.
Although fully reliable autonomous driving passenger vehicles are to be seen in the long run, companies that develop commercial driverless vehicles have already generated revenue by deploying low-speed driverless cars in a closed environment, such as parking lots, scenic spots, high-speed trains, harbors, airports, neighborhood blocks, exhibition halls and so on.
As a result, trucks are recognized to be the first frontier in the embrace of autonomous driving technology, because highways are a simple, predictable environment compared with city roads. As predicted by the American Trucking Association in 2017, there will be a deficit of 75,000 drivers by 2024. Autonomous buses in public transit systems are set to be another early service that is enabled by AI, which has already been provided in countries including Norway, Sweden and France.
Such a comprehensive project has to be backed by central and local governments. Chinese regulators increasingly emphasize autonomous driving-related businesses.
In 2018, China’s National Development and Reform Committee released Intelligent Vehicle Innovation Development Strategy (智能汽车创新发展战略) and the Ministry of Industry and Information Technology published V2X Industry Development Action Plan (车联网（智能网联汽车）产业发展行动计划). The former policy indicates that, by 2020, intelligent vehicles achieving DA(L1), PA(L2), CA(L3) will account for over 50% of new vehicles; the latter category indicates the penetration rate of V2X users will be over 30%. The latter emphasizes the integration of telecommunication, computing and the auto industry.
With strong national support, traditional vehicle manufacturers have stepped into the field of self-driving cars and began to assemble ADAS products. Domestic traditional car companies such as Great Wall Motor, BAIC Group, Changan Automobile and GAC Group have successively released the development plans of autonomous vehicles. Overall, the traditional car companies in China are stably making progress in automatic driving. Besides, tech companies and startups, led by Baidu, are actively exploring autonomous systems.
Based on the formula ‘market scale = unit price * penetration rate * production,’ Guosen Securities predicts China’s ADAS market will rise above CNY 100 billion (USD 14.15 billion) in 2020, with decreasing system prices, increasing penetration and relatively stable production.
However, compared with the US and Europe, China has three disadvantages in autonomous driving realization: 1) relatively low labor costs of Chinese commercial vehicle drivers; 2) stricter regulation and laws for road tests; 3) the relatively low product quality by Chinese auto OEMs. To the realization of autonomous driving, the top-down methods of resource allocation are as critical as bottom-up technology development around autonomous operation.