Self-driving cars explained: how autonomous is my car?

Self-driving cars will be on our roads by 2025. Here's everything you need to know about driverless cars and autonomous driving...

Self-driving cars explained: how autonomous is my car?

Driverless cars were once the preserve of science fiction, but they are rapidly becoming a reality. They promise a number of benefits, including reducing road traffic accidents and providing new mobility solutions for disabled and older people. 

The UK Government believes that autonomous cars are “the future of road travel” because of their potential to revolutionise the way people and goods travel. It expects self-driving vehicles to be in use on UK roads from 2025, and that, by 2035, 40 percent of new cars in the UK could have self-driving capabilities.

So it is investing heavily in the new technology that is needed to support their introduction. In February 2023 its Connected & Automated Mobility programme announced an £81 million combined investment with industry to promote seven projects that aim to accelerate self-driving innovation. The innovations include trials of self-driving buses, HGVs and taxis. 

Are autonomous cars safe?

However, before driverless vehicles can be used on public roads a lot of laws and regulations need to be put in place, though, along with improvements in infrastructure. Issues such as accident liability - who will be held responsible for any accidents that happen when cars are being driven autonomously - need to be resolved. And road networks may need to be upgraded so the software systems that feed information to self-driving cars can interpret signs and road markings accurately. 

Here we’re going to examine the different levels of autonomous cars and how they work, as well as looking at the history of driverless cars and how safe they are, and the factors that can influence how well they can operate.  

What is a self-driving car?

A self-driving car is one that is able to take complete control of the acceleration, braking and steering. Semi-autonomous cars require a driver to be onboard and ready to take back control when necessary, but they don’t need them to keep an eye on what the car is doing when it is driving itself. 


Fully autonomous cars require no assistance or monitoring at all from the driver. They are able to read and interpret road signs, understand road markings and furniture, such as traffic lights, and navigate their way to pre-set destinations, keeping up with other traffic but avoiding collisions. 

How do self-driving cars work?

Self-driving cars use a combination of radar, cameras and lidar (light detection and ranging sensors) to create and maintain a continuous map of their surroundings. 

Video cameras are used to read traffic lights and road signs, to track other vehicles and keep a lookout for pedestrians and other vulnerable road users. 

Lidar sensors bounce pulses of light off the car’s surroundings to measure distances, detect road edges, and identify lane markings. There may also be ultrasonic sensors in the wheels to detect curbs and other vehicles when parking.

All this information is constantly pulled together and processed to work out where and at what speed the car should be travelling, and instructions are sent to the car’s acceleration, braking, and steering systems.

What are the levels of autonomy in self-driving cars?

Autonomous cars are divided into six levels according to SAE International, the US-based society of automotive engineers, depending on how advanced they are and how much of the driving they can do. Below, we explain what those levels mean.

Level 0 autonomous cars

In Level 0 driving, the driver is completely in control of the car, for both speed and direction, with only warnings and alerts assisting them in extreme circumstances. Today most cars have some driver assistance systems, although budget trim levels and models such as the Dacia Sandero don’t have any and are considered to be Level 0. However, Level 0 cars do usually have active safety features, such as anti-lock braking, autonomous emergency braking, blind spot warning, electronic stability control and lane departure warning systems. 

Dacia Sandero front cornering

There are also some low-volume performance cars at this level, and the lack of driver assistance is said to create a 'purer' performance car.

Level 1 autonomous cars

This is the first of the ‘assisted driving’ levels, where the car can ‘support’ driving functions. Most mainstream new cars, such as the Ford Puma, Peugeot 3008 and Volkswagen Tiguan come with Level 1 technology that enables the car to control certain elements of travel.

Ford Puma front right driving

Examples of Level 1 technology include the most basic adaptive cruise control and lane keeping or centring assistance systems. Often the systems work in isolation, simply controlling either speed or sideways movement. Although vehicles with these systems offer greater support, the driver must still always play an active role in driving. Examples of Level 1 technology include the most basic adaptive cruise control and lane-keeping or centering assistance systems.

Level 2 autonomous cars

The second level of ‘assisted driving’ involves the car being able to ‘control lateral or longitudinal movement’. Vehicles at this level can combine a couple of systems together, such as lane-keeping assist and adaptive cruise control. Advanced driving assistance systems (ADAS) are rated Level 2, and they provide continual assistance with acceleration, braking and steering. These systems generally use a combination of sources, such as cameras, lidar and radar, to gather information that is used to alter the car’s speed and direction of travel. 

Mercedes EQE front cornering

An example of a Level 2 system, is the Mercedes Intelligent Drive system that is found in models, such as the Mercedes EQEIt monitors the environment and uses map data to automatically keep the car at the right speed in traffic, including stopping and pulling away, and it can also steer the car to keep the car in lane. The driver must still keep their hands on the steering wheel and remain attentive while it is being used and be ready to take back control. However there are some circumstances in which you can take your hands off the steering wheel. The Ford BlueCruise system on the Ford Mustang Mach-E can take over acceleration, braking and steering, and on certain pre-mapped motorways you can take your hands off the steering wheel if all conditions are met. 

Read our comparison test of cars with advanced driver assistance systems

Level 3 autonomous cars

This is the current highest level of autonomous driving, known as ‘conditional automation’, which allows drivers to take their hands off the steering wheel and for the car to take over all aspects of driving in certain situations without having to remain fully attentive. 

Traffic jam assistants, such as all-lane-keeping systems (ALKS), are an example of Level 3 technology. Although these are ‘hands-off’ systems, the driver must still pay attention and be ready to take back control if necessary, so they’re not ‘eyes-off’ systems. 

Mercedes EQS front cornering

Mercedes is introducing a Level 3 system in the 2024 S-Class and EQS. The Nissan Ariya already features the ProPILOT system and Tesla has upgraded Autopilot to Level 3 autonomy on cars including the Model 3 and Model Y.  

Level 4 autonomous cars

The first of the ‘automated driving’ categories is also the first ‘eyes-off’ and ‘hands-off’ category. There are already car models on sale with some form of Level 4 technology, but they can't be used until they become legal, which is expected to happen in 2025.

The Government describes Level 4 driving as when ‘the driver is only responsible and exercises control when the system is not in use’. The car can be in charge of all driving functions without the driver’s intervention, attention or input, but the systems can be turned off when the driver chooses and then normal driving resumes.

Level 4 cars can deal with all traffic situations on certain types of roads, but their use is likely to be restricted during specific weather conditions, such as rainstorms. 

While the driver needs to be onboard in most instances, in certain situations, such as when a car is being used in valet parking mode, the driver doesn’t need to be in the vehicle. 



We tested a Level 4 autonomous car on a track. Find out what it’s like to drive one

Level 5 autonomous cars

The highest standard of automated driving allows the car to control its speed and direction fully without any need for driver intervention. Level 5 vehicles are often referred to as ‘fully autonomous’. This type of autonomous vehicle is allowed to drive itself without a driver onboard. 

Level 5 cars are in development, but their introduction is still years away.

How safe are self-driving cars? 

On paper, fully-automated self-driving cars should be much safer than those being driven by people. They’ll be able to travel closer together and will be linked through the same data network, making them able to 'see' other road users and anticipate where other cars will go. 

They are also much less prone to human error, which contributes to more than 90% of collisions, according to the UK’s Transport Research Laboratory. It believes that, although autonomous cars won’t prevent all accidents, the reduction in harm will be on the same scale as when seatbelts became mandatory. 

Other research by the University of Michigan’s Transportation Research Institute estimates that pedestrian fatalities could be reduced by 20% when all cars are fully autonomous.

Self-driving cars within a decade

There have been many trials of autonomous cars around the world since 2010, and they have generally proved extremely safe. However, there have been a small number of accidents. There were 37 crashes of Uber vehicles in self-driving mode between September 2016 and March 2018. The first death happened in March 2018 in Phoenix, Arizona, when Elaine Herzberg was hit by an Uber self-driving car as she was walking her bicycle across a poorly lit stretch of a multi-lane road.

The test vehicle had software problems and failed to correctly identify her until it was too late to avoid an impact. Although prosecutors ruled that Uber was not criminally liable for the death, the car's back-up driver pleaded guilty to a criminal charge of endangerment after dash-cam footage released by police appeared to show her eyes off the road moments before the accident.

Although this was the first death involving a fully-autonomous car, there was an earlier fatality involving a Tesla Model S, in which the driver had been using its Level 2 Autopilot system. In May 2016, Joshua Brown was killed when his car hit a trailer being towed across the road. Neither the car’s sensors nor Mr Brown spotted the trailer, so the brakes weren’t applied; some reports suggested Brown was watching a film on a portable DVD player at the time of the crash.

As well as expressing its condolences, Tesla defended Autopilot’s safety record, stating: “This is the first known fatality in just over 130 million miles where Autopilot was activated. Among all vehicles in the US, there is a fatality every 94 million miles.” However, Tesla did acknowledge that the system had failed to detect the trailer, blaming a combination of its high ride height and what it describes as “the extremely rare circumstances” that allowed the car to pass under the middle of the trailer rather than hitting the front or rear of it.

Worryingly, the number of accidents involving cars with Level 2 autonomous systems appears to be increasing. Between July 2021 and May 2022 the American National Highway Traffic Safety Administration (NHTSA) organisation reported that there were 392 accidents involving cars in the USA being driven in semi-autonomous mode, 70% (273) of them involved Tesla models using its Autopilot function. 

Driverless cars – how will insurance work?

These accidents underline the importance of drivers using semi-autonomous systems properly, understanding their limitations and staying alert while using them. They also show that salespeople, marketers and car reviewers need to be responsible in the way these systems are presented to the public. A recent survey found that more than half of drivers think they can buy a self-driving car today, so more work needs to be done to educate drivers about driverless car technology and use. 

What other challenges do driverless cars have to overcome? 

The challenges that need to be tackled range from the technological and legislative to the environmental and philosophical. Here are just some of the unknowns.

Who is in control? 

The main bone of contention here is whether the driver or the car’s autonomous system are to blame for a collision. In order for blame to be correctly apportioned, proposals have been made that a secure online storage space be created where details of accidents can be stored and accessed by insurers and the authorities. 

Another issue is whether or not the driver takes back control of the car quickly enough. At present, there is a lot of research being done into how long a driver needs to retake the wheel. Volvo is one manufacturer investigating the best way to handle this transition, and its solution is a system where the driver pushes two paddles on the steering wheel to activate and deactivate the IntelliSafe Autopilot, Volvo’s self-driving technology. 

Weather Conditions

Cameras and sensors need to be able to read road signs and markings, and they may not be able to do this in extreme weather conditions, such as snow or torrential rain, so the systems could be limited to use in good weather. 

Severe weather warning for weekend

Sharing the road with non-autonomous vehicles 

Many industry insiders believe mixing autonomous and driver-controlled cars will have safety implications, so driverless cars could be relegated to certain types of road or specific motorway lanes to prevent an increase in accidents. 

Artificial versus emotional intelligence

There is also an ethical question to be answered: what if saving a pedestrian or cyclist means sacrificing the people inside the car? If a fatality is inevitable, should an autonomous car be programmed to save its owner or a vulnerable road user?

The Massachusetts Institute of Technology (MIT) put this question to members of the public. It found the majority surveyed thought autonomous cars should sacrifice its occupants, although the respondents also said they would prefer not to ride in such vehicles.

History of self-driving cars

People have dreamed of self-driving vehicles for hundreds of years. In fact, Leonardo da Vinci designed a self-propelled cart back in the 1500s. However, it wasn’t until the 1980s that the first autonomous cars actually appeared: a vision-guided Mercedes-Benz robotic van, designed by the Bundeswehr University in Munich, achieved a speed of 59.6 miles per hour on streets without traffic in 1987. 

The first road-following demonstration that used lidar, computer vision and autonomous robotic control took place in the 1990s in the USA. The Autonomous Land driven Vehicle (ALV) project used these technologies to drive a robotic vehicle at speeds of up to 19mph. 

Various projects were conducted around the world in the following decades, including one that produced the ParkShuttle, billed as the world's first driverless vehicle. This small bus used magnets embedded in the road surface to verify its position. Two pilot projects using the bus were done in the Netherlands, at Schiphol Airport and Rivium business park. 

Google pushes for driverless cars

Google began developing its fleet of driverless cars in 2009, and during the 2010s, many major car makers, including Audi, BMW, Ford, Mercedes Benz, Nissan, Toyota, Volkswagen and Volvo, began testing driverless car systems. 

In 2012, California, Florida and Nevada passed laws allowing driverless cars to be tested on public roads, and that year a Toyota Prius modified with Google's experimental driverless technology became the first self-driving car to be licensed in the United States. 

The first version of Tesla’s Autopilot driving assistance system was launched in 2014, the same year that SAE International launched the six levels of autonomous cars. The following year the UK Government announced it would oversee public trials of the LUTZ Pathfinder driverless pod in Milton Keynes. 

In 2016, Singapore launched the first self-driving taxi service as a pilot, provided by an autonomous vehicle startup company called nuTonomy.

The first production car to have Level 3 autonomous functionality arrived in 2018. The Audi A8 could drive itself at speeds up to 37mph using Audi AI. Drivers would not have to do safety checks, such as touching the steering wheel every 15 seconds, and they could safely turn their attention away from driving tasks. 

In the same year, the world's first fully electric self-driving bus was launched in Switzerland, and Waymo launched the first commercial robotaxi called Waymo One in Phoenix, Arizona. 

Regulations enabling the use of Level 3 autonomous vehicles were introduced in the EU in 2022, and in the UK from 2023. 

For all the latest reviews, advice and new car deals, sign up to the What Car? newsletter here

Next: Best electric cars >>

Also consider