Autonomous driving refers to the use of sensing, computing, software and vehicle-control systems to perform some or all of the tasks involved in driving without continuous human control, depending on the system's defined automation level. The concept encompasses a wide range of technologies rather than one single capability. Modern vehicles may provide driver-assistance functions such as adaptive cruise control, lane-centering assistance and automated emergency braking, while more advanced systems can perform broader driving tasks under specified conditions.
The internationally recognised SAE driving-automation framework describes levels from Level 0 through Level 5, distinguishing systems by who performs the driving task and under what conditions. The distinction between driver assistance and genuine automated driving is important because many systems marketed with terms such as autopilot or assisted driving still require a human driver to supervise the environment and remain responsible for the driving task.
Autonomous driving systems generally combine several categories of technology. Cameras provide visual information about road markings, vehicles, pedestrians, traffic lights and other objects. Radar can measure distance and relative velocity and can operate under conditions where optical sensing may be challenged. LiDAR uses laser pulses to construct detailed measurements of surrounding objects and environments and is used in some automated-driving systems. Ultrasonic sensors can assist with short-range perception and parking. Positioning systems, inertial sensors and detailed maps can provide additional information. No single sensor is universally sufficient for every driving environment, so automated systems may combine multiple sensing sources through sensor fusion and use redundancy to improve reliability.
Perception is the process by which an automated-driving system interprets sensor data. Machine-learning models and other algorithms can detect and classify objects, estimate their locations and movements, identify road structures and interpret traffic conditions. The system then performs prediction, planning and control. Prediction estimates how relevant road users may behave, planning determines a suitable path and speed, and control converts the planned trajectory into steering, acceleration and braking commands. These functions operate repeatedly and in real time. The complexity arises because road environments contain uncertain human behaviour, incomplete information, changing weather, unusual objects and situations that may not match previously observed patterns.
Autonomous-driving capability is commonly described using SAE levels. Level 0 involves no sustained driving automation, although warning or momentary intervention systems may exist. Level 1 provides assistance with either lateral or longitudinal control under defined conditions. Level 2 can provide simultaneous assistance with steering and acceleration or braking, but the human driver remains responsible for supervising the driving environment and driving task.
Levels 3 through 5 progressively transfer the dynamic driving task to the automated system under increasingly broad conditions, with Level 3 involving conditional automation, Level 4 high automation within defined operational domains and Level 5 automation intended to operate under all conditions in which a human could drive. The exact capabilities and legal status of a system must therefore be considered separately from marketing terminology.
A major concept in autonomous driving is the operational design domain, or ODD. An ODD defines the conditions under which an automated-driving system is designed to operate, potentially including road type, geographic area, speed range, weather, lighting, traffic conditions and other environmental constraints. A Level 4 system, for example, can perform the complete driving task within its specified operational domain while not necessarily being capable of doing so everywhere.
A system can therefore be highly automated without being a universal driver. When conditions fall outside the system's operational domain, it must follow its defined procedures, which can include requesting a human takeover for systems designed around human fallback or executing a minimal-risk condition for systems capable of doing so.
Safety is a central challenge because an autonomous-driving system must handle both ordinary and unusual situations while maintaining reliable control. Redundancy can be applied to sensors, power supplies, computing systems, communication pathways and braking or steering systems. Systems can monitor their own operation and identify degraded conditions. Verification and validation involve simulation, closed-course testing, controlled public-road testing and analysis of edge cases. However, no finite testing programme can observe every possible road scenario, so safety assurance also depends on system architecture, operational restrictions, software engineering, monitoring and disciplined deployment. Automated driving does not eliminate risk; it changes how driving decisions and failure modes are produced and managed.
Autonomous driving is distinct from connected driving and driver assistance. Vehicle-to-vehicle or vehicle-to-infrastructure communication can provide additional information but is not synonymous with autonomy. Similarly, an advanced driver-assistance system may automate particular functions without taking responsibility for the complete driving task. A vehicle can have sophisticated sensors and artificial intelligence while still requiring continuous human supervision. Terms used by manufacturers can also differ from formal technical classifications, so assessing actual functionality requires examining the system's automation level, operating conditions, supervision requirements and limitations rather than relying solely on product names.
Legal and regulatory frameworks are important because responsibility for driving changes when automation changes. Requirements can address vehicle approval, driver monitoring, cybersecurity, data recording, safety validation, human-machine interaction, insurance and responsibility following incidents. Regulations differ between jurisdictions and continue to develop as automated-driving technology advances. Autonomous-driving systems also raise ethical and societal questions involving accessibility, employment, privacy, infrastructure and allocation of responsibility, but these questions should be distinguished from the technical definition of automated driving. The technology is ultimately a cyber-physical system in which software decisions directly affect steering, braking and propulsion in a physical environment, making reliability, cybersecurity, fail-safe behaviour and human interaction fundamental engineering concerns.
