Autonomous driving: levels, technology and future

Alpha Guide » Games » Autonomous driving: levels, technology and future

Autonomous driving has gone from being a science fiction concept to one of the automotive industry's major goals in a very short time. Every year, new systems, pilot projects, and legal changes are announced, bringing us closer to the idea of ​​getting in a car, setting a destination, and forgetting about the rest of the journey.

Even so, we're still far from seeing steering wheel-less cars circulating en masse in our cities. What we do already have on the market are vehicles with varying levels of automation and a good selection of Advanced Driver Assistance Systems (ADAS) that are paving the way toward full autonomy. Let's break down clearly how all this works, what technologies these cars carry, what the SAE levels are, what real progress has been made, and what challenges remain to be overcome.

What exactly is autonomous driving?

Before discussing levels and brands, it's important to clarify that not every "smart" car system implies autonomous driving . The reference used by almost the entire industry is the J3016 standard from SAE International (Society of Automotive Engineers) , which defines quite precisely what constitutes automated driving and how it is classified.

Simply put, autonomous driving is the ability of a vehicle to sustainably take over some or all of the dynamic tasks of driving (accelerating, braking, turning, monitoring the surroundings, reacting to unexpected events, etc.), with varying degrees of human supervision. At the highest levels, the car can drive without any human intervention.

To achieve this, an autonomous vehicle combines sensors, processors, actuators, and software capable of perceiving its surroundings, interpreting what is happening, planning a maneuver, and executing it using the car's controls. All of this occurs within the conditions for which it has been designed, known as the operational design domain (ODD) : road type, weather, traffic, lighting, etc.

The main motivation behind this technological race isn't just convenience. It's estimated that over 90% of traffic accidents are related to human error . Reducing this factor, improving traffic flow in cities, and facilitating mobility for people with driving difficulties are some of the major goals behind the autonomous car.

How does a self-driving car work on the inside?

An automated driving car is not just a vehicle that "drives itself," but a complex technological platform. For it to function, it needs to perceive its surroundings, accurately locate itself, decide what to do, and execute that decision safely. All of this relies on several technological components.

First, there are the environmental perception sensors , which in a way replace the driver's eyes and ears. These include cameras, radar, lidar, ultrasound, infrared sensors, and night vision systems, among others. Each has its strengths and limitations, which is why they are usually combined.

Next comes the positioning and navigation aspect . An autonomous vehicle must know its location with much greater accuracy than a conventional mobile phone GPS provides. To achieve this, it uses satellite navigation systems (GPS, GLONASS, Galileo, etc.), inertial navigation, and, at the most advanced levels, high-definition maps that are continuously compared with the data recorded by the sensors.

Everything these devices capture is sent to control units with very powerful processors . That's where artificial intelligence comes in: computer vision, machine learning, and planning algorithms that transform raw data into concrete decisions about the steering wheel, accelerator, or brakes. This is what's known in the industry as automated driving systems (ADS).

Finally, the actuators (electric power steering, electronically controlled brakes, engine management, etc.) are responsible for executing the commands. If we add to this the communication systems with other vehicles or with the infrastructure (V2V, V2I), the concept of the connected car emerges , key to total autonomy in complex urban environments.

Levels of autonomous driving according to the SAE

To avoid confusion, the SAE establishes six levels of automation, from 0 to 5 , which have become the global standard. What changes from one level to another is, essentially, who is responsible for the driving tasks: the human, the automated system, or a combination of both.

Level 0: no automation

At level 0, all decisions and actions rest with the human driver . There may be driver assistance systems such as blind spot monitoring or forward collision warning, but these systems do not continuously control steering or speed. They are merely alerts; they do not take control of the vehicle.

Level 1: Driver assistance

Level 1 includes the first systems that control either longitudinal or lateral movement , but never both at the same time. For example, adaptive cruise control manages the accelerator and brakes to maintain a safe distance from the car in front, or a parking assistant steers the vehicle while the driver operates the pedals.

The main objective here is to make driving more comfortable and somewhat safer , but the responsibility still lies entirely with the driver, who must supervise at all times.

Level 2: Partial automation

At level 2, which is already quite widespread in the market, the vehicle can simultaneously manage steering and speed within a specific scenario (usually a highway or traffic jams). This allows the car to maintain its lane, adjust its distance from the vehicle ahead, and move in heavy traffic almost "on its own."

Typical examples include lane-keeping assist working in conjunction with adaptive cruise control , traffic jam assist, and advanced parking systems that control both the steering wheel and pedals. Even so, a Level 2 car is not designed to react autonomously to every unexpected situation , and the driver must remain constantly vigilant and ready to intervene.

Level 3: Conditional Automation

Level 3 represents a significant leap forward, as the vehicle now features an automated driving system capable of handling all dynamic tasks under certain conditions. The driver can "cede" control to the car, which then drives itself without continuous supervision, but with important nuances.

The system has a well-defined operating environment (for example, congested highways at low speeds) and is capable of managing the driving until it detects that it is about to leave that environment or encounters a situation it cannot resolve. At that point, it requires the human driver to take back control, allowing a limited time for the handover.

If the driver does not respond in time or there is a failure in some key component, the system must be able to bring the vehicle to a situation of minimum risk , such as stopping it in a controlled manner in a safe place.

Level 4: High automation

At level 4, the car can operate without human intervention within a specific operational design domain . The driver is no longer expected to react to a request for assistance from the system; in fact, there might not be anyone in the driver's seat while the vehicle operates within its intended environment.

The automated driving system is capable of handling unexpected situations and managing emergencies to reach a safe location on its own, even without driver intervention. However, it still has geographical and road-type limitations: it may be designed, for example, only for specific urban areas, closed campuses, or particular highway routes.

The user can still request to take control or deactivate autonomous mode, but unlike Level 3, the vehicle can delay deactivation if it deems it safer . And if no one responds, it is obligated to come to a safe stop on its own.

Level 5: Full automation

Level 5 is the "holy grail": a car capable of moving in any environment and under any condition in which a human could drive . There are no restrictions on road type, weather, or specific area, at least in theory.

In this scenario, the vehicle can do without elements like a steering wheel or pedals , because there is no driver. A person simply indicates the destination and the car takes care of everything, whether it's a complicated urban journey or a secondary mountain road in the middle of fog.

Although some companies are working on prototypes close to this concept —especially in the form of autonomous taxis—, the reality is that there are still no level 5 vehicles available to the general public , nor are they expected to be in the short term.

Autonomous cars and ADAS: where are we now?

It's important to clarify that currently, there are no Level 5 cars , or even Level 4 cars, available to individual consumers at dealerships. What we do have are cars with multiple Advanced Driver Assistance Systems (ADAS) and some specific systems that reach advanced Level 2 and even Level 3 in highly controlled scenarios.

Advanced Driver Assistance Systems (ADAS) encompass all those driver assistance features that are already familiar: automatic emergency braking, lane keeping assist, adaptive cruise control, traffic sign recognition, 360-degree cameras, automatic parking, driver fatigue detection, etc. Their function is to reduce the risk of accidents and alleviate the driver's workload , but in most cases they do not completely replace the driver.

In Spain, current regulations are very clear: the driver is not allowed to let go of the steering wheel while driving , even if the car's technology could handle it. This effectively limits the use of systems that, in other countries, are considered Level 3 automation, so here they continue to be used as advanced driver assistance systems that require the driver's constant attention.

In other places, such as some US states or certain European and Asian countries, autonomous driving tests have been authorized in open traffic , especially with fleets of robotaxis or shared mobility services. These are highly monitored environments with fairly controlled conditions, but they demonstrate that the technology is maturing.

Current examples of automated driving systems

To put all of the above into perspective, it is worth reviewing some of the most talked-about systems that are already being deployed in different markets and that clearly illustrate the variety of approaches.

BYD God's Eye

God's Eye is BYD's advanced driver assistance system , initially launched outside of Europe, which is now being incorporated even into its most affordable electric models. It uses a combination of LiDAR sensors, cameras, and radar to constantly monitor the car's surroundings.

With this information, it can control speed, maintain lane position, change lanes on the highway, and avoid collisions in certain situations. Its interaction with connected traffic lights has also been tested, receiving real-time data on the traffic light phase to anticipate stopping or resuming movement.

Although its behavior in the city and in complex maneuvers is very advanced, its actual classification falls within a level 2 of automation : the driver has to continuously monitor and be ready to take control at any time.

Tesla FSD (Full Self-Driving)

Tesla's well-known FSD is a software package that offers highly advanced driver assistance features . Unlike other systems, it relies almost exclusively on a set of cameras around the vehicle, foregoing lidar and instead using artificial intelligence models trained on millions of kilometers of real-world user data.

Among its capabilities are door-to-door navigation, automatic lane changes on the highway, traffic light and sign recognition, automatic parking, and the famous "Summon" feature , which allows you to move the car at low speed from your mobile phone or key.

Despite its name, which might make it seem like magic in some videos, FSD is currently considered advanced Level 2 automation . Tesla emphasizes that the driver must remain attentive and prepared to intervene, because the system does not yet have full decision-making capabilities in all possible situations.

Ford Blue Cruise

Ford has introduced its BlueCruise system in models like the Mustang Mach-E , which in Spain is sold by subscription after a trial period. This system allows what the brand itself calls "hands-free, eyes-on-the-road" driving on certain highways and motorways designated as Blue Zones.

While the car is driving through one of these zones, BlueCruise monitors steering, speed, braking, and lane position , while also maintaining a safe distance from other vehicles. An interior camera monitors the driver's gaze and head position to ensure they remain focused on the road.

The driver can take their hands off the wheel when the system authorizes it , but must remain ready to act. The system adapts to slow traffic, stopping and starting in traffic jams, and can relinquish control if the user requests it or if they leave one of the Blue Zones.

Mercedes-Benz DRIVE PILOT

Mercedes has gone a step further with its DRIVE PILOT , which has been certified as a Level 3 SAE system in some US states, becoming one of the first of its kind to be officially recognized.

Under specific conditions—approved sections of highway and speeds up to approximately 65 km/h, especially in heavy traffic—DRIVE PILOT can take over driving completely . In this mode, the driver is legally authorized to perform other tasks permitted by law, as the system is considered responsible for driving while active.

To achieve this, the vehicle incorporates LiDAR, radar, redundant cameras, and a high-precision positioning system , as well as an architecture designed to safely manage failures. The system is offered via subscription, and Mercedes is working to expand its availability to other markets and road networks.

Robotaxis and level 4 "geofenced" projects

Companies like Waymo (a subsidiary of Alphabet, Google's parent company) and some Chinese operators have been testing autonomous taxis in designated urban areas for years . These vehicles operate in Level 4 mode within a specific geographic area, without a human driver, although always under remote supervision and with strict limitations.

In Waymo's case, they use various types of sensors, including LiDAR, radar, cameras, positioning systems, and sound detectors, to recognize even emergency sirens. Their software has been trained in simulators, on closed circuits, and later in real traffic, accumulating millions of kilometers of experience.

Key technologies in autonomous driving

For all of the above to work reliably, automated vehicles rely on a set of interconnected technologies. Each manufacturer chooses its own "recipe," but the basic ingredients are usually the same.

Environmental perception sensors

Self-driving cars are equipped with various types of sensors, each with its specific role. Their combined operation allows for the construction of a dynamic, three-dimensional image of the vehicle's surroundings.

Ultrasonic sensors are perfect for detecting very close objects, such as when parking. They work by emitting inaudible sound waves and measuring the time it takes for them to bounce back. Their low cost has made them common in parking assistance systems and perimeter alarms for years.

Satellite positioning systems —GPS, GLONASS, Galileo—allow you to know the exact location of a car on the planet thanks to trilateration with multiple satellites. On their own, they have a margin of error of several meters, but when combined with other correction techniques and high-definition map data, their accuracy improves to levels suitable for advanced automation.

Inertial navigation relies on accelerometers and gyroscopes to estimate the vehicle's position, orientation, and speed without external references. It is often used to complement satellite positioning to mitigate temporary errors or signal loss, although on its own it does not achieve the accuracy required for Level 4 or 5 navigation.

Infrared sensors and night vision systems provide valuable information in low-light conditions, allowing the detection of pedestrians, animals, or obstacles that conventional cameras would struggle to see. Some high-end models have offered this technology as an option for years.

Video cameras are now ubiquitous. They are inexpensive and have greatly improved in quality, but the challenge lies in the software capable of interpreting what they show: lane lines, signs, pedestrians, bicycles, vehicles, roadworks, etc. Their performance can be affected by heavy rain, fog, lens dirt, or low light.

Radar uses electromagnetic waves to detect objects and measure their distance, speed , and trajectory. It works very well up to several hundred meters and is less sensitive to adverse weather conditions. This is why it has become widespread in functions such as adaptive cruise control, blind spot monitoring, and automatic emergency braking.

LiDAR (Laser Imaging Detection and Ranging) emits laser pulses and measures the time it takes for them to return, creating 3D maps made up of point clouds with a very high level of detail. This information allows for highly accurate reconstruction of the environment, including road edges, obstacles, street furniture, and other vehicles.

HD mapping, advanced localization and data interpretation

To achieve high levels of autonomy, it's not enough to know "more or less" where the car is. High-definition maps and localization systems that combine satellite data, vehicle sensors, and precise mapping are needed .

Some specialized companies, such as TomTom, HERE, and Mobileye , create these detailed maps, incorporating information on lanes, horizontal and vertical signage, fixed road features, and other details that facilitate automated navigation. Each company uses different sensors (some prioritize radar, others cameras, and others LiDAR) to collect the data.

A typical approach involves comparing what the car's sensors see in real time with a stored map . If the "image" matches, the system can position the vehicle with great accuracy, which is crucial for complex maneuvers or highly structured environments such as multi-lane highways.

An interesting example is Cadillac's Super Cruise , available on some models in the United States. This system allows hands-free driving on pre-mapped sections of highway (where permitted by law), combining HD maps, cameras, adaptive cruise control, and GPS. However, it doesn't reach Level 3: it doesn't change lanes automatically or avoid unexpected obstacles, and the driver must remain attentive at all times.

Vehicle-to-vehicle and infrastructure (V2X) communication

Vehicle communication systems allow cars and the infrastructure itself (traffic lights, information panels, roadside stations) to exchange data in real time. This is called V2V (vehicle-to-vehicle) when they communicate with each other, and V2I (vehicle-to-infrastructure) when they communicate with road elements.

This communication can use 5G or dedicated short-range communication (DSRC) technologies that operate on specific bands, with ranges of up to one kilometer and bidirectional capability. This allows for sending warnings of sudden braking ahead, the presence of a stopped vehicle, slippery road surfaces, or changes in the phase of a traffic light.

Although they are not strictly essential for an autonomous car to function —in fact, companies like Waymo have shown that progress can be made without depending on them—, they greatly improve anticipation and safety , and are considered an important piece of the future mobility ecosystem.

Redundancy and security of systems

In an autonomous vehicle, a software or hardware failure cannot simply result in a total loss of control. Therefore, manufacturers work on architectures with high levels of redundancy in all critical systems: steering, brakes, power supply, communications, sensors, positioning, etc.

The idea is that if a key component fails, another parallel system can take over or at least bring the vehicle to a state of minimal risk. For example, in a braking system, the traditional vacuum pump can be replaced by an electromechanical pump controllable by both the stability control unit and the automated driving system, thus providing multiple ways to activate the brakes even if one part of the system fails.

From level 3 upwards, this redundancy is not a luxury, but a basic safety requirement , especially because at those levels the driver is no longer considered directly responsible for reacting to any technical problem.

Current situation, challenges and future of autonomous driving

Today we have widespread Level 1 and 2 vehicles , some specific Level 3 systems in real-world testing, and very limited Level 4 projects (robot taxis, truck routes, closed campuses, etc.). Level 5, for now, remains in the realm of advanced research and technology demonstrations.

The advances are not only technological. There is also significant work being done on regulatory frameworks and safety standards . In Europe, the new general vehicle safety regulations have opened the door to Level 3 systems, and in countries like Spain, the concept of "automated vehicle" has already been included in the Road Safety Law for the first time.

Even so, major challenges remain: defining legal responsibilities in the event of an accident , ensuring cybersecurity against potential attacks, protecting the privacy of the data generated by these vehicles (routes, usage habits, constant location) and adapting insurance policies to a scenario where the driver may not be the one making the decisions.

There are also significant ethical and social acceptance issues. Many people remain wary of the idea of ​​getting into a car that's driving itself , especially after high-profile incidents involving misused driver assistance systems. Therefore, manufacturers, governments, and technology companies must work together to build trust and demonstrate, with data, that these systems reduce accidents.

In terms of timelines, most realistic forecasts place a significant presence of Level 4 vehicles in specific areas around the 2030s , primarily in the form of autonomous taxi services or freight transport along designated corridors. Widespread Level 5 is generally projected beyond 2040, and some experts doubt it will be fully achieved under all conceivable conditions.

Meanwhile, it is very likely that over the next 10-15 years we will continue to see increasingly autonomous cars, but with a steering wheel and driver in control , supported by increasingly intelligent assistants: almost unattended highway driving, fully automated parking, advanced traffic management thanks to 5G and communication with traffic lights and other infrastructure, etc.

With everything that has already been deployed and what is on the way, autonomous driving is advancing steadily but step by step: today the technology coexists with the driver as a very capable co-pilot, and little by little it will take over more decisions until, in many cases, the human becomes just a passenger who chooses the destination while the vehicle does the rest of the work.