BACK TO BASICS - AUTONOMOUS DRIVING Everything about self-driving cars and their technology
Fully autonomous vehicles are not really here yet, at least not in the ways we’ve envisioned them in the past. They are coming, but in the meantime, autonomous capabilities allow for some pretty powerful and convenient functionality.
For people who are just becoming acquainted with new mobility technology, the term “autonomous vehicle” may be somewhat confusing. What is an autonomous vehicle? Is a self-driving car autonomous all the time, or can people take control of driving functions where and when they desire? How does autonomous vehicles work exactly? This article provides insights into the world of autonomous vehicles.
What is an autonomous vehicle?
Primarily, an autonomous vehicle has an internal computing platform that can take over some—or all—of the duties of driving and free up the vehicle’s driver to perform other tasks, such as work-related activities, entertaining themselves, or doing nothing at all. Of course, the phrase “take over the duties of driving” does not imply simplistic functionality; even if those duties just translate to parking, there are myriad inputs, variables, and decision-making processes that have to be taken into account.
How does a self-driving vehicle work?
First of all, sensors—including radar, sonar, and/or LiDAR (which stands for light detection and ranging)—are utilized along with cameras for self-driving cars to get an idea of their surroundings and what lies in the vicinity ahead. One of the chief functions of autonomous vehicle platforms is to detect obstacles and/or moving objects—be they pedestrians, other vehicles, motorcycles, or bicycles—and avoid these so that collisions don’t occur. A radar transceiver sends a radio wave out that reflects off objects and obstacles and bounces back to the transceiver’s sensor so it can be processed. The length of time the wave takes to come back determines the distance to any particular object or obstacle. Sonar and LiDAR work the same way, except that sonar uses sound instead of radio waves, and LiDAR uses lasers.
In addition to these sensors, driverless cars utilize cameras as well as global navigation satellite systems (GNSS)—of which the U.S.’s government-funded global positioning system (GPS) is one type. Self-driving vehicles take all of these inputs—sensors, cameras, and GNSS data—and synthesize them to come up with a very accurate three-dimensional model of what surrounds them. By constantly updating this data dozens or even hundreds of times per second, autonomous vehicle platforms can see which objects are in motion (such as cars that make up oncoming traffic) and be able to predict their trajectories seconds (or even minutes) into the future.
Let's look at the streets of Paris through the eyes of a Tesla Autopilot:
The other main function of autonomous vehicle platforms is to actuate braking, acceleration, and steering, so vehicles can drive themselves (supplanting human drivers) when and where necessary. In order for this to happen, autonomous vehicle platforms need to be physically wired into the electronic control units (ECUs) that control these functions of the cars they’re installed in.
SAE Autonomy Levels of self-driving vehicles
The circumstances under which a car takes control of driving functions and what a driver needs to do when that happens are described by levels of autonomy as defined by the Society of Automotive Engineers (SAE). These levels are as follows:
- Level 0: At this level (“No Automation”), there’s no autonomous functioning of a vehicle whatsoever; cars that are not autonomous vehicles operate at SAE Autonomy Level 0.
- Level 1: This level (“Driver Assistance”) allows a vehicle to assist a driver with specific functions, such as adaptive cruise control (ACC); the vehicle can take over either acceleration/braking or steering, but not both at the same time.
- Level 2: At this level (“Partial Automation”), advanced driver-assistance systems (ADAS) permit a vehicle to take over both acceleration/braking and steering simultaneously. Drivers still need to keep their hands on the vehicle’s steering wheel and their feet on the brake and accelerator pedals.
- Level 3: This level (“Conditional Automation”) allows a vehicle to fully take over driving under a limited set of circumstances. Drivers may take their hands off the steering wheel, but they must still keep their attention on the road and on their immediate surroundings. Drivers must be ready to take back control of their vehicle at a few seconds’ notice (in cases where drivers are unwilling or unable to take back this control, their vehicle will gradually decelerate to a full stop).
- Level 4: At this level (“High Automation”), drivers can take their attention completely away from the road and the task of driving, so they can engage in other activities, including socializing with other passengers and being entertained. It’s envisioned that in a number of SAE Level 4-capable vehicles of the future, drivers’ seats will be able to rotate 180 degrees so that drivers can face the rear of vehicles and interact directly with rear-seat passengers.
- Level 5: This level (“Full Automation”) of autonomy is what most people envision when they think of driverless cars; the car does absolutely everything itself; there is no human “driver” any longer. In fact, purpose-built SAE Level 5 vehicles usually lack all driving controls, such as a steering wheel, manual brakes, and an accelerator pedal; the person who was previously the vehicle’s driver is now simply considered another passenger.
- Generally speaking, the higher one goes on the SAE autonomy scale, the more expensive the technology is that’s necessary to achieve such a level.
Self-driving car benefits
Why do we need self-driving cars? It’s a great question—after all, just because we may be capable of making this technology work, the query remains—why is it of any benefit to us?
The immediate answer should be obvious: safety. Annual fatalities from traffic-related accidents number at least 1.35 million worldwide. A common objective of self-driving car developers is to one day achieve “Vision Zero”—an idealized goal of lowering the number of traffic-related fatalities to zero through the use of various strategies, including (in many cases) the universal adoption of autonomous vehicles (and in a large number of cases, the prohibition of non-autonomous vehicles) on public roads.
Beyond safety, there are the factors of comfort and convenience to consider. Imagine if drivers no longer had to operate their vehicles themselves. What could they do with that newfound freedom and time? Rapid advances in car cockpit technology have made it possible for people to relax, to connect with coworkers, family, and friends, and to communicate and share messages, data, and even video streams with them. Infotainment systems now let passengers and drivers alike watch movies, listen to music, and enjoy the latest video games and interactive entertainment inside their driverless cars.
As technology matures, this infotainment will become more immersive, and augmented-reality (AR) and virtual-reality (VR) experiences will let vehicle occupants leverage their real-world location and surroundings to achieve enhanced productivity and leisure. Vehicles will become intelligent, and technologies such as driver monitoring will allow autonomous vehicle platforms to know when (and if) vehicle occupants can and should take over driving or perform other duties.
Eventually, when more cars are self-driving vehicles, intelligent transportation systems (ITS) will be able to use artificial intelligence (AI) and Big Data to optimize traffic flows on roads and make journeys for millions of travelers more efficient.
The current state of self-driving cars’ technology
Currently, many new vehicles come with a fair amount of SAE-defined autonomy, typically incorporating at least some capabilities that would be considered SAE Levels 1 and 2—features such as adaptive cruise control (ACC), lane-keeping assist (LKA), and active parking assist (APA). Some cars have gone further and incorporated advanced features, such as traffic sign recognition (TSR) and remote control valet parking.
But true SAE Level 3 operation—where drivers can take their hands off the steering wheel (while their eyes remain on the road) and let the car self-drive—has only been included with just a handful of cars so far, including vehicles from Audi, Nissan, and NIO. And a major caveat of this SAE Level 3 functionality is that it’s only operational on highways, and possibly only at low speeds or in single lanes, depending on the car manufacturer.
For SAE Level 4 functionality, one has to turn to “robotaxi” services that are operated by companies like Waymo, Aptiv, and DeNA. In limited, precisely defined areas (where weather and traffic are generally mild), the robotaxi vehicles from these companies can take over most driving functions. However, in many cases, a backup driver is still present in the vehicle for safety. For truly driverless SAE Level 5 operations, one has to turn to low-capacity, minibus-style autonomous vehicles that are currently in use in a highly select number of locations in the world, including some airports and office parks. These vehicles literally lack any type of manual driving controls at all as well as anything that could be called a “driver’s seat.” At the same time, many of these vehicles do not operate in mixed traffic—instead, they drive in special lanes or on dedicated roads, and only at low (under 40 kph) speeds.
To see more advanced progress in terms of SAE-defined autonomous functionality, one may have to look at other autonomous vehicle types—self-driving trucks, buses, and industrial/commercial vehicles. In some of these cases, clear task-based functions make autonomous driving easier and a better fit based on particular operating constraints and/or environments.
Driverless cars in the future
Going forward, more automakers will start to add highway-only SAE Level 3 capability to their cars. Slowly over time, the caveats that are present with this functionality in today’s cars will disappear. At some point in time, this Level 3 operation will transition to SAE Level 4—but only on highways.
It may be a decade or more before we see SAE Level 3 or 4 operation on normal streets due to liability reasons. Until the robotaxi-style SAE Level 4 operation that’s been successfully demonstrated by Waymo and others can be replicated universally on all roads, in all towns and cities, under any weather or traffic conditions, the risks are too great for carmakers to incorporate this functionality in their vehicles. (In 2018, an Uber self-driving car accidentally killed a pedestrian in Arizona—an incident that sent shock waves through the self-driving vehicle industry and led to a re-examination of many testing practices.)
However, it’s possible that in the medium-term future, some automakers may allow SAE Level 4 or 5 functionality to be “switched on” in certain geographic areas at certain hours of the day or in specific weather conditions (clear, sunny skies, for instance). While the “Vision Zero” goal of zero fatalities from traffic accidents may never be achieved in reality, it’s possible—even likely—that we will one day get very close to it.
In summary, it’s been a long journey research- and development-wise to get to the level we’re at currently regarding self-driving vehicle technology. Many millions of kilometers have been driven in test vehicles, and billions of kilometers have been run in simulations to get autonomous vehicle platforms to the state they’re in today. While some industry observers were predicting we’d all be driving SAE Level 5-capable autonomous cars by now, progress has been slower on several fronts than many researchers expected. Component costs, technology constraints, and simply the sheer number of potential scenarios that a vehicle can encounter when traveling on public roads have all been limiting factors that have impeded the rapid achievement of universal SAE Level 5 functionality. But if the recent past is anything to go by, it’s quite possible that by 2025, we’ll be well on our way to the long-envisioned dream of fully self-driving cars where the word “drive” will only apply to machinery, and the occupants in a vehicle will all be passengers.