Autonomous vehicles around the world: How close are we to regular driverless operation?

A few years ago, reports of cars capable of handling part of the drive on their own were more of a novelty from technology fairs. Today, however, it's no longer about isolated prototypes. Autonomous vehicles are gradually penetrating regular traffic and are starting to appear in situations previously reserved solely for human drivers. Expectations have also shifted. Instead of questioning whether this technology will take hold, there's increasing discussion about where it makes the most practical sense and what demands real conditions place on it.

In this article, we will examine what underpins autonomous driving, the principles that keep an autonomous vehicle on the road, and the situations in which the technology is already being applied. We will focus on specific technical foundations and operational experiences to demonstrate what current systems can do, where they have their limits, and what factors determine their further development. Thanks to this, it will be clear why autonomy is moving ever closer to regular traffic, but at the same time remains cautious where the environment is too complex or unpredictable.

The technology autonomous vehicles rely on

For an autonomous vehicle to drive safely, it needs to continuously gather accurate information about the space around it. It uses a combination of sensors that capture various types of data. Cameras recognize lanes, traffic signs, object shapes, and their movement. Radar measures the distance and speed of cars in front and beside the vehicle, reliably even in adverse weather or darkness. Another technology works with light flashes that reflect off surrounding obstacles. Based on the time it takes for them to return, it creates accurate spatial information about the surroundings. Thanks to the combination of these sources, a detailed image of the road situation emerges.

Sensors, however, are not the only source of data. The car also needs to know exactly where it is located. It uses highly detailed maps that are much more accurate than standard navigation. These contain information about the exact road profile, lane placements, and junction shapes. The system continually compares these details with real-time sensor data. If both layers match, it is confident in accurately interpreting the environment.

All gathered information flows into the vehicle's central computing software. It analyzes this data in real-time and decides how the vehicle should react. The software evaluates the trajectories of other cars, recognizes obstacles, assesses potential risks, and determines suitable speeds and maneuvers. These decision-making algorithms are based on a vast amount of driving data that helps predict common situations and less frequent scenarios.

Part of the whole system includes safety backups. If a sensor temporarily fails to provide sufficient data, the other technologies can compensate. Likewise, backup computing units are present in the car, ready to take over critical functions if the main system fails.

Autonomous driving thus relies on the integration of multiple technologies, together creating a robust and reliable system. Only through this synergy can the vehicle's response be fast, predictable, and stable, provided the conditions and data are sufficiently clear.

Automation levels according to SAE and their true meaning

The international organization SAE International has created a scale defining six levels of autonomy from 0 to 5. This framework is used by automakers and regulatory authorities worldwide, serving as a unified description of how much work a car can handle independently and when human intervention is still necessary.

Level 0

Driving is entirely in human hands. Systems can warn of risks but do not interfere with driving.

Level 1

The vehicle assists with one specific task, such as adaptive cruise control or light lane corrections. Overall control remains with the driver.

Level 2

The system combines multiple functions simultaneously. The car maintains its lane, adjusts speed, and responds to traffic ahead. Responsibility, however, stays with the human driver.

Level 3

In specifically defined situations, the car can take over driving and monitor traffic itself. The driver must be ready to take control if the system requests it. This level is primarily applied on highways.

Level 4

The vehicle is fully capable of driving in designated areas or under specific conditions. Typical examples include robotaxis in restricted urban zones. Outside these areas, autonomous driving would not operate.

Level 5

Full autonomy without restrictions. The car should handle any traffic without human intervention. This level is not yet available in regular traffic.

Where autonomous cars are already operating today

In regular streets, autonomous operation is most advanced in the USA. The most visible service is Waymo One, which transports passengers without a driver in several parts of Phoenix, such as Tempe or Chandler, and in selected areas of San Francisco and Los Angeles. The rides take place in pre-mapped zones where the system has been extensively tested over time. A similar model is expanding to Austin and Atlanta.

In California, Cruise also operates driverless cars. Its vehicles primarily run in San Francisco and several smaller cities. Despite stricter regulatory oversight, operations continue.

In China, the scope of autonomy is even greater. Companies like Baidu Apollo or AutoX operate robotaxis in Beijing, Shenzhen, Wuhan, or Guangzhou. Some city routes allow full autonomous rides without a driver across dozens of kilometers. Additionally, China boasts autonomous buses running on regular routes in places like Shanghai and Shenzhen.

Europe progresses more cautiously, but autonomous driving is not an exception here either. In Germany, Level 3 systems can be utilized on selected highway sections. In Finland and France, autonomous minibuses operate in smaller urban areas, often within campuses or residential districts. Some European cities are also testing robotaxis with an operator on board, such as in Stockholm or Paris.

Where does autonomy have its limits?

The biggest challenge for autonomous systems is situations requiring more than just precise data. These include scenarios where traffic rules overlap with human interaction. Typical are intersections without traffic lights, where drivers give way to each other using simple gestures or eye contact. Autonomous vehicles do not use such signals and rely solely on measurable data. This leads to more cautious responses in some situations, slowing down traffic.

Problems also arise in scenarios that do not match pre-mapped data or test-driving information. These include temporary changes like provisional signage, detours, or roadworks. While autonomous software can handle them, it often requires greater distance or slower responses to ensure accurate interpretation.

Another limitation is the system's inability to understand the wider context. A human driver often foresees a traffic jam approaching, notices a child on the sidewalk might run into the road, or predicts a vehicle ahead slowing down for an unseen pothole. Such situational reading is still challenging for autonomous driving, as it relies only on measurable and unequivocal data that may not capture the full intent of surrounding participants.

A significant limitation is also the lack of foresight into human behavior. Drivers often predict actions based on subtle details, such as how one holds the steering wheel or the vehicle's positioning in the lane. For the system, evaluating such behavior is difficult, opting for a more cautious strategy.

These limitations are not obstacles that halt progress. They demonstrate, however, that autonomy currently operates best in environments where rules and participant behavior are as straightforward as possible, and where data is reliably interpretable. Anything deviating from expected patterns is challenging for current systems and requires ongoing research and refinement.

Myths surrounding autonomous driving

"An autonomous car drives completely by itself."

In reality, no commonly available system today can manage all situations without supervision. Autonomous driving is always limited to specific conditions, selected routes, or precisely defined areas. Outside of these, the system shuts down and requests the driver to take over. The idea of the car "driving anywhere on its own" does not yet match reality.

"A machine always reacts faster than a human."

An autonomous system has a quick response time, but only when the data is clear. In scenarios lacking context or with a confusing environment, the system might hesitate or choose an overly cautious strategy. The issue isn't speed but rather scenario comprehension.

"Autonomous cars make mistakes because the software isn't reliable enough."

In practice, issues are primarily caused by the environment. Provisional signage, unexpected obstacles, or the behavior of other road users can lead the system into scenarios not already captured in data. While the technology is robust, the world around it is highly dynamic.

"Once autonomy is engaged, the car assumes all responsibility."

The legal framework varies by country. In many instances, the driver remains obliged to be ready to intervene. Only in some regions is the liability partially transferred to the manufacturer or service provider. Thus, there's no universal rule applicable everywhere.

What will the next step towards full autonomy look like?

The immediate future won't be about cars driving completely without a human driver, but rather about gradually expanding the range of situations they can handle independently. The technology will move from selected districts and highways to larger areas where the system will operate with greater certainty and fewer limitations.

A pivotal role will be played by connecting vehicles with infrastructure. Traffic lights, road signs, and navigation data will be able to provide more accurate information directly to the car, reducing uncertainty in situations that are currently complex for autonomy.

Legislation will also be crucial. Once clear conditions for responsibility and data handling are established, services like robotaxis can expand to additional cities and states. Thus, development will likely follow a path of stable, clearly defined rules rather than rapid leaps.

The coming years will mainly bring broader and more reliable application of current systems. Full autonomy remains a goal for the more distant future, but the technology will become a more accessible and natural part of regular traffic operations.