Autonomous Driving , How Autonomous and When?




Autonomous driving technology will fundamentally transform the automotive industry, changing the way vehicles are built, operated, sold, used and serviced. Dozens of companies are already working on the sensors and other technologies required to enable these vehicles to know where they are and what is around them. More than 50 companies are building artificial intelligence (AI)-based software to make autonomous vehicles smart enough to communicate with their passengers, other vehicles and the cloud, while planning the best path for a safe, fast and entertaining journey.

Press announcements from players in the autonomous driving market often create the impression that it will all be available soon — that is, all the benefits of driving safety, improving the quality of lives, delivering economic benefits, revolutionizing transportation logistics and enabling smart city integration. However, while much of the autonomous vehicle technology appears to be developing quickly, virtually no commercially available vehicles have yet passed the required level 4 advanced driver assistance systems (ADAS) for autonomous driving.

Based on a six-level definition of driving automation from SAE International (see graph below), Level 2 (Partial Automation) is the highest practical level that is currently in production in any significantly large volume. Level 2 ADAS was delivered on 5.1 million vehicles in 2016, and was forecast to rise to 8.6 million units in 2017 (final results pending more data inputs).


Currently, as to the future of autonomous driving, leading players tend to fall into one of two camps:

  • CAMP #1: believes in a gradual deployment of increasingly sophisticated automation that will incrementally take more and more vehicle control. This first stage of autonomy is slated for high-end luxury vehicles, and will work in only select situations. Level 3 (Conditional Automation) technology will allow drivers to engage an autopilot that can drive the vehicle in certain situations, such as specifically designated highways. However, a human driver must be ready to retake control of the vehicle after the vehicle exits the highway. Proponents of this camp feel that lessons learned from Level 3 vehicles will eventually lead to a fully autonomous car. Today, companies such as GM, Audi and (even) Tesla, among others, are in this first group. Tesla’s Autopilot is one of the prime examples in this camp. These companies also intend to make Level 4-capable vehicles eventually.
  • CAMP #2: suggests that it will require vehicles to attain SAE Level 4, where a robot controls the vehicle entirely, before a viable and sustainable autonomous market can gain real traction. This group argues for an entirely new transportation system, contending that there is no safe way to deploy a Level 3 system because it will still require a human — who may not be paying attention — to retake control in a dangerous situation. Ford, Volvo and Waymo are in this second group, with Waymo leading by a wide margin.
Two Paths to Autonomous Driving

Key Forces Behind Scenarios for Autonomous Driving in 2025

The path of the autonomous driving market is destined to be fraught with sudden and often majorly disruptive twists and turns, driven by a complex relationship among four interrelated areas: technology, regulations, costs and consumer acceptance.

Autonomous driving in some form will happen at some point, and we should expect to see multiple launches of autonomous vehicles somewhere around 2020. However, the full impact of autonomous vehicle technology on society and the economy will not begin to emerge until approximately 2025.

Here is the analysis on these four key forces:


Regulations affect vehicle automation in multiple ways. In addition to road and driving regulations, as well as functional safety requirements (for example, ISO 26262), regulations regarding environmental considerations and liability issues will come into play. For example: the legal or moral culpability when it comes to accidents involving autonomous vehicles.

In terms of road regulations, there is a fundamental legal distinction between automated vehicles (SAE Level 3, with a human driver) and autonomous vehicles (SAE Level 4 and above, without a driver). Regulations regarding autonomous vehicles are among the biggest barriers preventing autonomous vehicle on all streets. Right now, some of the latest incidents from Tesla and Uber have put even automated vehicles in an awkward position. While autonomous vehicles may be less likely to cause an accident than human drivers, there are standing evidences indicating that accidents would still occur (at least before all cars become autonomous and other vehicles, including bicycles and pedestrians, would only behave logically).

In Europe, a 2016 amendment to the Vienna Convention on Road Traffic stipulates the extent to which drivers must be in control of their vehicles. According to this latest amendment, systems are now deemed to be controllable at all time, virtually banning autonomous driving. To achieve Level 4 autonomy, the United Nations Economic Commission for Europe (UNECE) is already elaborating on a further update to the Convention on Road Traffic to enable the use of driver-less systems in the future.

All participants involved in the manufacture and operation of autonomous vehicles agree that a set of universal standards must be developed to ensure the safety of passengers and other traffic participants. However, those discussions are still in the early stages and will likely take a rather long period of time to fully develop, which could directly delay the availability of SAE Level 4 or above autonomous vehicles.

Consumer and Social Acceptance

Consumer trust and buy-in will be essential for the broad adoption of autonomous vehicle technology. Currently, overall consumer acceptance is not keeping up with autonomous vehicle technology.

Percentage of Consumers Who Feel Full Self-Driving Vehicles Will Not Be Safe, Source: Deloitte

According to Deloitte’s Global Automotive Consumer Study, which polled over 20,000 people in the United States, Germany, Japan, South Korea, China and India, found that the majority of those surveyed would “not” consider riding in fully autonomous vehicles. Gen Y/Z consumers are generally more interested in fully autonomous vehicles, but enthusiasm for fully autonomous vehicles has flattened and even declined in several markets.

Change in Consumer Preference for Full Self-Driving Vehicles from 2014 to 2017, Source: Deloitte

Also, from a societal point of view, the morality of actions taken by a computer has not been addressed. A common issue raised is the moral dilemma of how a computer decides between two unavoidable, lethal options (as emphasized in Will Smith’s acting in I, Robot the movie). Another still-unaddressed moral issue has to do with culpability in case one of these vehicles is involved in some criminal activity. For example, who should be responsible for an unregistered autonomous vehicle without a driver trafficking drugs in the neighborhood?

Initially, the primary market for fully autonomous vehicles may be as a replacement for taxis or for-hire transportation (think Uber). Later, it may be as a replacement for personal vehicle ownership with mobility services.

Some automaker executives have said that they don’t expect consumers to be able to purchase a fully autonomous car for personal use for some time after they become commercially available (see later section on Cost of Technology). As a result, CAMP #2 has suggested that to achieve large-volume numbers and a highly disruptive scenario, people must give up ownership of their own vehicles and migrate to a mobility-as-a-service (or pay-as-you-go as some preferred) model. Already, more and more consumers are opting to use transportation services such as Uber and Lyft. The continued growth is expected to gnaw away at new car purchases as people are able to travel more efficiently and less expensively through ride hailing. With Uber, Lyft and other shared riding serving as education tool, younger people are already less likely to get driver’s licenses than before. A study by the University of Michigan Transportation Research Institute (UMTRI) found that the percentage of 20 to 24-year-olds with driver’s licenses had fallen from 92% in 1983 to 77% in 2014. Studies in Germany and other regions have shown similar results too.

The total market for autonomous vehicles would be relatively small, unless a large number of people decided to abandon their personal vehicles for a mobility service. Cost reductions based on high-volume production may not occur quickly in a scenario where mobility services are the primary market for autonomous vehicles.

Cost of Technology

The sensors on the vehicles today are expensive and require a new internal computing platform and infrastructure. Initial versions of autonomous vehicles are likely to cost more than $100,000, based on the numerous new technologies that will be required in the earliest vehicles and their low initial volume. Some even said it would be above $250,000.

Classification of Driving Assistance Systems and Features by Autonomous Level

Costs are certain to fall as companies commercialize the sensors, but how fast and how much they will fall is uncertain. Delphi has indicated that it expects to be able to offer an entire suite of sensors to enable autonomy to automakers for $5,000 when sales reach high volumes (without the actual timeline).

LIDAR, the Key to L3 Autonomous Driving (as Redundancy to Camera and Radar)

The most expensive sensor today is LIDAR, a laser range-finding system that most autonomous vehicle developers are using. Price reductions are pinned to producing a low-cost LIDAR sensor based on solid-state electronics, rather than a rotating series of lasers, like the units developed by market leader Velodyne (which is used in Google’s autonomous driving program, now Waymo). Several companies are working on such solid-state systems, including Quanergy, Continental, TriLumina and LeddarTech. Google’s autonomous driving division, Waymo, is developing its own customized LIDAR too. Some other companies, including Tesla and AutoX, are using cameras and radar rather than LIDAR. If a camera-based system could be operational for SAE Level 4 or above by 2020, it would greatly reduce costs. Nonetheless, judging by Tesla’s recent slippage on technology promise, I probably wouldn’t bank on camera system to deliver Level 4 in the future.

In addition, if Level 3 systems gain in popularity quickly, it could help to drive down the cost of the equipment by establishing a market for the sensor suppliers earlier than a Level 4/Level 5 autonomous vehicle could. So far, consumers have shown not just a great interest but also willing to spend more on Level 3 systems.

Automotive Semiconductors for Level 3 Autonomous Driving

Another aspect of the cost of technology is, while autonomous vehicles may be less likely to cause an accident than human drivers, there is a possibility that insurance rates for such cars may be more costly because of the high replacement cost of these vehicles.

Readiness of Technology

The implementation of autonomous vehicles will require mastery in a series of technology areas, including:

  • Sensing: Sensor and software technology that enables vehicles to understand their location not just geographically, but also situationally — that is, with context-awareness of such things as road signs, pedestrian traffic, local traffic rules, construction activity and traffic jams
Some Notable Semiconductor Suppliers for Level 3 Autonomous Driving
  • Decision Making: Algorithms for predictions and decision making for planned pathways and trajectories. Deep learning and AI will be required to detect, predict and react to the behavior of other road users, such as vehicles, pedestrians, cyclists, animals and other dynamic roadway events
  • High-Definition Maps: These are readable by computers, which act as the reference point for where objects are in the real world, in order for the machine to be able to pinpoint its exact location in the environment. Some of the most innovative ideas have been using each vehicle with advanced sensors to record road map in very small-sized files and uploading these files to a centralized data center to build an nearly-real-time map that can downloaded by all vehicles to serve autonomous driving purposes. However, the realization of this concept probably requires some degree of maturity of 5G technology and deployment. For now, the finalization of 3GPP Release 14 in 2017 was crucial to establishing C-V2X capabilities for LTE to compete with the long-established capabilities of dedicated short-range communications (DSRC)/802.11p.
REM (Road Experience Management), Real-Time Mapping via V2I, Source: Intel/Mobileye
  • User interfaces (UIs): These have a smooth transition to looping in or looping out the driver (in case of Level 3 driver participation), or provide in-vehicle experiences for autonomous operating vehicles. These UIs provide specific working or entertainment environment alerts. User experience may also be used to convey trust to riders that the ride is safe and secure via transparent and forward-looking information on driving maneuvers. By using sensors and AI to monitor the driver’s behavior and the content being consumed, the vehicle will be able to draw the driver’s attention back to the warning messages and the core responsibility of piloting the vehicle
  • In-vehicle versus cloud deployment, and connectivity of information to the vehicle and back: Companies must be able to learn from the information coming in from the vehicle, and then deploy the lessons learned back into the fleet
Delay of Connected Car Could Be a Hiccup to Development of Autonomous Driving

A range of technology companies, large suppliers and automakers are exploring these technologies and developing hardware and software implementations for use in-vehicle and/or the cloud. The viability of the concept has been proved in multiple showcases of vehicles traveling on public roads, or in test vehicles that are currently being operated. The next frontier for autonomous driving is to ensure that the vehicle can understand and make sense of the data in order to devise safe, efficient and smooth driving maneuvers.

Four Scenarios

Based on an evaluation of the above drivers, foresees the following four scenarios evolving by 2025.

Scenario #1: Experimental Transportation Scenario

In this scenario, the world of autonomous driving has not progressed very much beyond what we have today. Level 3 technology, which allows for partial automated driving, is maturing. Vehicles increasingly come equipped with more advanced ADAS (mostly Level 3) systems, but customer adoption remains low and limited to certain regional markets, due to limited trust in the technology or consumers’ ability to afford the technology. All vehicles still require a driver. Level 4 autonomous vehicle technology remain limited to pilots and trials, although the trials will have become larger in scale by 2025.

Technology readiness in this scenario has significantly advanced over today, but unclear, highly localized regulations hinder commercialization. Mobility as a service is gaining in popularity, but most men still suffer from driving.

Scenario #2: Automated Logistics and Long Commute Scenario

In this scenario, automated vehicle penetration is up because governments have recognized the potential benefits of the autonomous vehicle technology for quality of life, environmental protection and more efficient traffic. Regulations have encouraged the development of joint highway and large city arterial rules and regulations, as well as clear functional safety standards and requirements. This has allowed for the development of vehicles in volume on a global scale.

Level 3 vehicles are in demand, especially for longer commutes on arteries and for long distances, now that properly regulated and retrofitted roads are available. Platooning of commercial trucks to save fuel has proliferated on long-distance routes. Greater volumes of vehicles that operate at higher utilization rates bring down costs for commercial and lightweight vehicle operations. This helps to make autonomous driving an attractive component for mobility as a service on specific roads and in specific areas (than the above scenario). Penetration of autonomous vehicles is still small with these restrictions.

Scenario #3: Emerging Automation Scenario

In this scenario, technology has made a major step forward. The 2021-announced Level 4 vehicles are actually capable of handling driving situations in all instances on all roads in some regions, including major global cities. The functional requirement specifications have been agreed upon and developed.

The cost of the technology is still quite high, however, since the volume of vehicles is small (but more than the above two scenarios). Autonomous functionalities are limited to premium vehicles. The use of a fleet mobility-as-a-service concept and increased utilization help to equalize the higher cost of vehicles. And these fleets could be available in even some developing countries.

Autonomous Car Commercialization + Shared Driving Could Create a New Business Model

Customer acceptance is gradually adjusting as operations of the few autonomous vehicles on the road show that the technology works even in mixed traffic scenarios. Fully autonomous vehicles are still limited to some regions, but the regions and cities that have implemented the technology validate it. Consequently, the number of Level 4 vehicles rises quickly, reaching midsize numbers in those implemented regions.

Scenario #4: Full Mixed Traffic Scenario

This scenario — the most disruptive of the four — describes a 2025 world in which all scenario drivers have developed quickly, supporting autonomous driving in all its required facets in all regions and global cities. Standards and regulations that developed and adapted on a global scale, enabling security for the planning of technology and production. In close cooperation with the automotive industry, regulators agree with the industry on standard in-vehicle and cloud features, as well as requirements for safe operations.

The 2021-announced vehicles spearhead a fundamental change in the automotive industry. The industry is becoming software-driven, technology-driven and data-driven. Platooning is common for commercial fleets on highways. Autonomous vehicles open up transportation options to some individuals who were previously excluded because they were unable to drive. New vehicle concepts also make ride sharing on longer-distance commutes comfortable and further drive the adoption of autonomous vehicles.

In this scenario, the world is leapfrogging to Level 4 for automation concepts. Personally owned vehicles begin to fade. Driving a vehicle increasingly becomes a rare skill.

The Bottom Line

With so many uncertainties, it is difficult to predict which scenario would be the 2025’s. We do know that the societal benefits from autonomous vehicles will outweigh potential disadvantages and concerns. Furthermore, the potential of fundamental business disruptions from autonomous and driver-less vehicles will motivate industry leaders and newcomers to expand research and development activities, and to dramatically advance the progress of the technologies.

Autonomous driving is changing the way technology service providers engage, communicate with and monetize the driver relationship from now until 2025. New opportunities to build customer relationships or to extend existing customer relationships into the car as the third living space are waiting to be seized.

Personally, I would argue Scenario #2 would be the bottom line…but I would be afraid to put any actual money on this bet. But as an everyday driver, driving is tedious despite some occasional fun. Liberating from driving would definitely improve “my” productivity. Go Scenario #4!