Definition: Self-driving cars use cameras, radar, LiDAR, and AI to drive without a human.
Current status: Level 2 cars are on sale. Level 4 robotaxis are being used in America. Level 5 is not yet available.
Core tech: Sensor fusion, neural networks, teleoperation, LiDAR, radar.
Pros: Fewer crashes and better access.
Cons: Concerns about cost, safety, and ethics.
There are six levels of driving automation: Level 0 (no automation) – Level 5 (full automation). Most cars today are level 2. Level 3 and 4 cars are being tested in some cities in America.
Levels Of Automation
|
Level |
Who Drives |
Where It Works |
Driver Fall-back? |
Examples |
|
Level 0 — No Automation |
The human driver controls all aspects of driving. | Everywhere. | Always. | Traditional manual vehicles. |
|
Level 1 — Driver Assistance |
The driver handles most driving but may get single-function support (e.g., steering or speed). | Highways or simple roads. | Always — must supervise. | Adaptive Cruise Control, Lane Keeping Assist. |
|
Level 2 — Partial Automation |
The vehicle can control steering and acceleration/braking simultaneously, but the driver must stay alert. | Highways or controlled conditions. | Yes — must monitor the road and take over instantly. | Tesla Autopilot, GM Super Cruise, Ford BlueCruise. |
|
Level 3 — Conditional Automation |
The vehicle drives itself in certain conditions but may request driver intervention. | Limited scenarios (e.g., traffic jams, mapped highways). | Yes — driver must be available to resume control. | Mercedes-Benz Drive Pilot, BMW Personal Pilot L3 (in select markets). |
|
Level 4 — High Automation |
The vehicle drives itself in defined, geofenced areas without human input. | Specific cities or test zones. | No — within the operational area. | Waymo One, Zoox, Cruise robotaxi fleets. |
|
Level 5 — Full Automation |
The vehicle operates entirely on its own in all environments and weather conditions. | Everywhere. | None — no human driver or controls. | Not yet available commercially (long-term development goal). |
(Source: Imagination)
LiDAR provides 3D cloud data. This helps a car find its surroundings, spotting objects nearby and planning where to go.
LiDAR emits laser pulses and measures the time it takes these pulses to bounce off an object and return. The time it takes to come back shows how far away it is. LiDAR is the primary system for telling self-driving cars where to go, whilst identifying road signs, safe distances and traffic signals.
AI uses the data provided by the sensors to make human-like decisions. This involves following road signs and looking out for hazards on the road.
Within the AI system, neural networks help the car take information from the cameras. In certain test scenarios, teleoperation allows a human to remotely control the car.
The Sensors Involved In Autonomous Driving
|
Camera |
Radar |
LiDAR |
Ultrasonic |
|
|
Sensing Principle |
Passive light (image / video) | Active radio waves | Active laser light (time-of-flight) | High frequency sound-waves |
|
Output Data |
2D Image/Video, Object Classification, Colour | Range, Velocity | 3D Point Cloud, High-resolution depth | Short-range distance detection |
|
Poor Weather Performance |
Poor (affected by fog, heavy rain and low light) | Excellent (unaffected by most weather) | Moderate (Struggles with rain, fog and snow) | Excellent (unaffected by most weather) |
|
Resolution / Accuracy |
High resolution for classification, poor 3D depth | Low spatial resolution, excellent velocity accuracy | Extremely high spatial resolution and depth accuracy | Low resolution, designed for very close objects |
|
Cost |
Low | Medium | High (this is falling quickly though) | Very low |
|
Primary ADAS Use |
Lane Keep Assist, Traffic Sign Recognition | Adaptive Cruise Control, Blind Spot Monitoring | High-accuracy Mapping, Obstacle Avoidance | Parking assist, obstacle detection at low speed |
Benefits
Safer Roads: Most crashes happen because people make mistakes. Self-driving cars can help stop that.
Less Traffic: These cars can share information to keep traffic moving.
More Freedom: Older people and people with disabilities can travel more easily.
Better for the Planet: Smoother driving uses less fuel and makes less pollution.
Challenges:
Safety and Trust: Past test crashes have worried people. This has gone some way to reducing trust in self-driving cars.
Rules and Responsibility: If a self-driving car crashes, who’s at fault?
High Cost: The technology is still very expensive to build.
Bad Weather: Rain and snow can block the car's sensors.
1,353 self-driving car crashes in the US (Techpilot).
By June 2024, about 200 self-driving car accidents had been reported in the US (G&M Direct Hire).
In early 2025, Tesla said its self-driving cars crashed once every 69 million miles. Normal cars crash once every 963,000 miles (Tesla).
Waymo found 96% fewer crashes at junctions. Also, 91% fewer crashes where airbags went off (Waymo).
Fairness and Bias - Self-driving cars need to be fair. They can’t treat people differently because of their age, gender, or identity. The car’s computer makes decisions using code, not feelings. It follows rules but doesn’t have emotions or care like humans do. People can think about what’s fair in a moment — a computer can’t.
Accountability – If a self-driving car crashes, who is responsible? Is it the occupant, the software programmer, the manufacturer, or the AI itself? This creates a murky line of responsibility in the event of an accident.
The Trolley Problem – There’s an old question called the trolley problem. It asks what someone should do if they have to choose between hurting one person or several people. The same idea can be used for self-driving cars. Imagine a self-driving car stuck in the middle lane with something coming toward it. A person might try to swerve or brake, even if it means being hit from behind, to avoid hurting others. But what would a computer do? A self-driving car doesn’t have feelings — it only follows facts and rules. So, would it choose to hit another car to keep its passenger safe? That’s one of the big questions about how these cars should be programmed.
Different Approaches:
A report by McKinsey & Company suggests that 2040 could be when we see fully autonomous cars on our roads. Emily Shao, leader within McKinsey’s Advanced Industries and Travel, Logistics & Infrastructure Practices, has the following to say on the matter:
“By 2040, my kids will be in their late teens, early 20s. They probably won’t have driver’s licenses, because they just want to use AVs to get back and forth.”
Mingyu Guan, leader of McKinsey’s Automotive & Assembly Practice in Greater China, added to this, saying “Ten years, 15 years down the road, I imagine there’s no more need for driver's licenses. All the vehicles on the road will be autonomous driving equipped.”
These quotes raise the question, what’s the holdup?
Self-driving cars will change how cities work and look. Because cars won’t need to park for long, parking spaces could be turned into parks, walking areas, or fun places for people to meet. This would make cities greener and nicer to live in.
Self-driving taxis, called robotaxis, could also help by letting people share rides. That means fewer cars on the road and less pollution.
Traffic could also move more smoothly. Self-driving cars can travel at the same speed and stay close together safely. They could even “talk” to traffic lights. This would mean drivers don’t wait too long at red lights. All this could make your daily trip to school or work quicker and easier.
Autonomous technology is here. However, its everyday use is still limited by regulation, safety validation, and infrastructure readiness.
The UK doesn’t allow self-driving cars right now. They are looking to introduce them around 2027.
All cars will likely be self-driving by 2040.
Tesla is not the only self-driving car. Waymo and Alphabet produce fully autonomous taxis, and brands like Ford want to get into the self-driving space.
The biggest problems of self-driving cars:
Driverless cars are vehicles that perform activities like steering and accelerating without human input.
Driverless cars use advanced sensors that fuse information together about the area around the car. This information is passed to the car, which will apply the brakes if needed.
Self-driving cars led to 1353 crashes in the US in 2023. This was lower than human-driven cars. Tesla announced that in Q2 2025 their driverless cars had an accident every 6.69 million miles vs every 960,000 miles for human-driven cars.
Robotaxi services are fully autonomous vehicles designed to carry passengers without a human driver. They operate within geofenced areas, meaning they can only drive in specific, well-mapped city zones.