“I think we’re very close to having fully self-driving cars,” Tesla CEO Elon Musk declared in February, only to surprise investors two months later when he posted on X that Tesla would unveil robot taxis in August.
If you’re having deja vu, you’re not alone: Musk has been talking about self-driving cars since 2016, when Tesla unveiled a feature it called “full self-driving.” In 2019, he predicted that full self-driving was just around the corner, saying, “Maybe in two years we’ll have cars that don’t have steering wheels or pedals.”
That’s not yet the case, but Tesla isn’t alone in its rocky road to full self-driving: Other companies, including Cruise and Waymo, have also faced setbacks as they try to transition the world to a future where self-driving vehicles (AVs) are the norm.
Consider Cruise, the General Motors subsidiary whose California license was suspended last October after one of its self-driving cars struck a pedestrian and dragged him 20 feet before coming to a stop. Or Alphabet Inc.’s Waymo, formerly Google’s self-driving car project, which said in February it was voluntarily recalling software for its self-driving robotaxi vehicles after two of its vehicles crashed into the same pickup truck within minutes of each other.
Despite these setbacks, startups and established companies alike are working to solve the autonomous driving challenge.
Waymo began testing its vehicles on highways this year, and Cruise resumed driverless operations in Dallas in June, but those vehicles will be supervised by a safety driver behind the wheel.
On the software side, self-driving car startup Wave Technologies raised more than $1 billion in a Series C round that attracted investors including SoftBank, Nvidia, and Microsoft. One thing is clear: the race to self-driving cars isn’t slowing down.
Big vision with big challenges
To understand the challenges of developing a fully autonomous vehicle, let’s start with how it works.
Self-driving cars generally follow a “sense-plan-act” architecture. Take Waymo for example: their vehicles rely on a range of sensors (LiDAR, radar, cameras) and advanced software using artificial intelligence to process all sensory inputs in real-time, make split-second decisions on what to do, and send signals back to the car’s steering, braking and acceleration control systems.
Their design philosophies differ: Tesla, for example, eschews much more expensive LiDAR in favor of cameras, ultrasonic sensors, and radar, while startups like Wayve are developing AV2.0 self-driving tech that aims to learn and adapt to new scenarios without the need for pre-programmed rules or detailed maps, upending traditional “sense-plan-act” architectures.
The industry has made great strides in bringing autonomy to market using “sense-plan-act” architectures, with many cars on the market in the U.S. now equipped with autonomous features such as driver assistance and enhanced cruise control.
Waymo and Cruise are piloting technology that could be considered Level 4 autonomy, which would enable fully autonomous driving in limited service areas and scenarios.
The ultimate goal is a Level 5* autonomous vehicle that can drive itself on any road and in any situation without any input from the passenger. Developing such a vehicle comes with several challenges, including technical, regulatory and ethical issues.
First, let’s consider the technical challenges. AVs rely on sensors to assess their operating environment in real time. When many AVs are driving on the same road, there’s a risk that their LiDARs will interfere with each other. Sensors are less reliable in adverse conditions — think of a lane divider covered in snow — it’s difficult for AVs to make informed driving decisions in real time.
Additionally, the machine learning algorithms used to train the artificial intelligence at the heart of AV systems require vast amounts of data about every possible driving scenario that may be encountered on the road. While current AVs can handle many standard scenarios, unexpected “edge” cases are particularly challenging for AV systems to handle.
Second, the lack of federal regulation and variations in regulation by state create challenges for companies looking to design and deploy AVs at scale.
In December 2020, the National Highway Traffic Safety Administration (NHTSA) requested comments on the development of a framework for the safety of automated driving systems. In the document, NHTSA stated that the four key functions the agency should focus on are sensing, perception, planning, and control.
Progress has been slow. In April, NHTSA finalized standards to require pedestrian automatic emergency braking (AEB) for all passenger cars and light trucks by September 2029. While a comprehensive federal framework continues to evolve, states are filling the gaps with their own regulations. California, Michigan, and Arizona have been the most aggressive players, implementing a variety of regulations spanning AV testing, reporting, and deployment.
Finally, there are many ethical questions that continue to spark public debate. One big one is the classic “trolley problem”: in a crisis situation, how should an autonomous vehicle decide whose safety to prioritize? This dilemma goes beyond programming complexities and involves significant moral judgments traditionally made by humans. Who should define these ethical frameworks? Engineers, regulators, or someone else?
Another concern is the risk that hacking could lead to cars being used for criminal or other illicit purposes. As we navigate these challenges, one thing is clear: there are no easy answers. A continued dialogue between governments, the private sector and the public is essential.
Potential massive disruption
Despite the challenges, AVs have the potential to disrupt the entire passenger vehicle value chain once they eventually reach maturity.
McKinsey research predicts that Level 4 autonomous driving systems for passenger cars could generate $170 billion to $230 billion in revenue by 2035. Original equipment manufacturers (OEMs), traditionally strong in hardware, will need to match this prominence in software development and deployment. A key question for many OEMs is whether to develop these capabilities in-house (like GM and Cruise) or form strategic partnerships (like Daimler Truck with Waymo and Torc Robotics).
The widespread use of AVs will also have a significant impact on adjacent industries such as insurance. Some argue that with AVs, responsibility shifts from the “driver” to the OEMs and technology providers who design and implement the autonomous driving solution. However, manufacturers such as Tesla with their “Autopilot” features argue that the ultimate responsibility for the vehicle lies with the driver, not the company. Clarification of responsibility will be crucial for the development of AVs.
AVs may disrupt the entire automotive business model, challenging traditional car ownership and leading to a ride-sharing future. New pricing models based on service fees and subscription models may proliferate. Increased efficiency and utilization of ride-sharing and car-sharing services may significantly reduce the total number of vehicles on the road. Competitive advantage may shift from companies that manufacture high-quality vehicles at scale to companies that deploy artificial intelligence and software to deliver a superior driving experience.
The (near) future of AV
The big question is, when will true Level 5 self-driving cars be widely available?
S&P’s latest forecast predicts that Level 5 autonomy won’t be on the roads until 2035 or later. But we’ll likely start to see targeted use of Level 4 AVs before then. We’re already seeing niche applications, such as in John Deere farm tractors, small delivery vehicles on university campuses, and ride-hailing services offered by companies like Waymo in cities like San Francisco and Phoenix.
Many experts predict that long-haul trucks will be the first to see self-driving vehicles on the road: Daimler Trucks recently unveiled its first self-driving truck and said it will have driverless semi-trucks on the roads by 2027.
For passenger vehicles, we are likely to see a more localized rollout before widespread use nationwide.
Factors such as population density, road infrastructure and local government support will determine which cities will be attractive for AV expansion.
Consumer acceptance is essential for a fully autonomous future. A recent S&P Global Mobility consumer survey found that 65% of buyers want Level 2+ hands-off autonomous driving on highways. Yet a 2021 McKinsey survey found that only 25% of buyers are “very interested” in advanced self-driving capabilities (defined as L2+, L3, L4) when buying their next car. Interest in AVs is not yet universal, but this may change over time.
Potential benefits of AVs include reducing accidents caused by human error, saving driving time, and reducing carbon emissions through increased efficiency. Startups and established companies alike are innovating around AV’s key technologies, business models, and use cases.
The competition is fierce, with the US, China, Germany and Japan investing heavily in this technology race, and the boundaries of what is possible with autonomous vehicles will continue to be pushed on a global scale.
The development and widespread adoption of self-driving cars will be a winding road, so fasten your seat belt and get ready for an exciting journey.
*Level 4 or “High Automation” is defined as “The system is fully responsible for the driving tasks within a limited service area, with the occupant acting solely as a passenger and without any need for involvement.” Level 5 or “Full Automation” is defined as “The system is fully responsible for the driving tasks, with the occupant acting solely as a passenger and without any need for involvement.” For more information, visit https://www.nhtsa.gov/vehicle-safety/automated-vehicles-safety.
