View More


Elements of Automated Driving


Driving automation operates thanks to the interaction of several systems that enable the vehicle to understand the driving environment, make intelligent decisions, and navigate safely to a destination:

  • Localization and Mapping determines where the vehicle is located within its environment. This requires building a specialized map of the surrounding environment, either from scratch or by drawing from a baseline of prior knowledge that is well-understood and trusted to be mostly correct, and then localizing the vehicle within that map. This system helps a vehicle correctly interpret the data its sensors gather.

  • Perception combines information from the Mapping and Positioning system with data from vehicle sensors – including cameras, LIDAR, RADAR, global positioning systems (GPS), and inertial navigation units (INU), among other inputs – to collect and interpret information about the vehicle’s current situation and its relationship to its environment. This includes the location and movement of the full range of obstacles, both static and dynamic, including infrastructure, vehicles, pedestrians, bicycles, and more. The amount and complexity of data for analysis makes this one of the most challenging steps in automated driving.

  • Prediction helps the vehicle imagine where other vehicles, pedestrians, bicycles, etc. are likely to be in the future. Often there are multiple possible predictions (known as hypotheses).

  • Planning determines one or more safe courses of travel for the vehicle, including decisions such as which lane to travel, where to position the vehicle relative to other dynamic objects, and how much space to afford obstacles. Critically, the Planning system must make decisions about how to safely guide the vehicle under conditions of uncertainty, such as when other vehicles on the road may be blocked from view, or if they behave in unexpected ways. Multiple hypotheses may lead to multiple possible plans, with the ultimate choice depending on the actions of other vehicles, pedestrians, and more.

  • Control executes the planned driving trajectories set by the planning system, which are updated constantly based on new information. This is accomplished using actuators that direct vehicle drive functions.

  • Coordination communicates with other vehicles, the road infrastructure, and cloud databases.

  • External Human Machine Interaction (e-HMI) manages the communication of information between the vehicle and humans in the traffic environment. Importantly, while communication between driver and vehicle is obviously important, particularly in managing the handoff of control, so too is communication between the vehicle and humans outside the vehicle, such as drivers and pedestrians.

Driving automation operates thanks to the interaction of several systems that enable the vehicle to understand the driving environment, make intelligent decisions, and navigate safely to a destination.


An overlapping set of core technologies and tools make the fundamentals of automated driving possible. These include:

  • Artificial Intelligence (AI) is a broad term for technology that processes information and makes decisions to achieve a certain goal. This may be accomplished via a rule-based system, such as if a vehicle perceives a stop sign and follows a programmed command to stop, or via machine learning, in which a system might process large volumes of information to differentiate a car from a bicycle.  

  • Computer Vision is the process of gathering information from sensors and using it to perceive the surrounding environment. This process leverages artificial intelligence to draw knowledge from the data, identifying and differentiating individual elements, such as cars, pedestrians, trees, and roads.

  • Predictive Algorithms are used to anticipate the likely behavior of other objects in the road environment, such as the expected future position of another vehicle on the road based on its current trajectory and proximity to other vehicles.

  • Decision Algorithms choose the vehicle’s proper path based on the predicted behavior of others on the road. Importantly, decision algorithms must operate despite uncertainty, which varies based on conditions including visibility and traffic congestion. 

  • Maps are baseline representations of the core elements of the physical world the vehicle occupies. These include both high-definition maps, which can be generated ahead of time and used by a vehicle when it enters an environment, or generated in real time using algorithms such as Simultaneous Localization and Mapping (SLAM).

  • Sensors gather data from the driving environment or from the vehicle itself. These include systems that gather data about the world, such as video cameras, LIDAR, and RADAR; those that track location, such as GPS; and those that monitor the movement of the vehicle itself, such as inertial measurement units, or wheel speed and angle monitors.

  • Actuators are used to control the physical operation of the vehicle, opening or closing the throttle, turning the wheels, or engaging the brakes. Importantly, while much of automated driving is performed by computers at very high speed, actuators are limited by physical constraints, including vehicle dynamics and the speed of the actuator itself. Thus, automated drive systems must account for the lag between issuing commands and the vehicle’s physical response.

  • Simulation is used to test the performance of automated driving software in a virtual environment. Data gathered from real world testing is used to recreate a variety of traffic scenarios in simulation to test and measure system response and ensure proper operation. 

  • V2X Communication consists of a direct information exchange between vehicles, with roadside traffic management systems, and with pedestrians via digital short-range communication and via cellular networks. These vehicle-to-vehicle (V2V), and vehicle-to-infrastructure (V2I), and vehicle-to-pedestrian (V2P) communications share information about road signals, signs, road conditions, and other vehicles or pedestrians that may be difficult to see. They can also alert drivers of approaching vehicles, pedestrians, red lights, and slow or stationary vehicles. Vehicle-to-network (V2N) supports map data generation and map data updates, as well as various kinds of information delivery and remote control. Together, V2X communications provide an additional means for automated vehicles to gain knowledge about surrounding traffic. The information obtained is combined with data from on-board sensors to enable the vehicle to make better decisions for vehicle control, traffic safety, efficiency, and driver interaction.