Several companies, including Ford, BMW, Waymo and Tesla, have announced plans to manufacture fully autonomous cars within the next five years. As these cars are still under development, safe deployment of these vehicles is likely many years away. Semi-autonomous cars will be available to buyers much sooner. Volvo, for example, is actively testing these vehicles in Sweden (SciPol brief available).
Fully autonomous vehicles require several different technologies to function. First, they require sensors to collect visual and location data. Visual sensors can include light imaging detection and ranging (LIDAR), radio detection and ranging (RADAR), or cameras, or a combination of all three. Each technology has its own advantages and disadvantages, so multi-sensor systems are often used so the strengths of one sensor offsets the weakness of another. LIDAR can produce a 3D map 100 meters around a vehicle, but they produce large amounts of data and do not perform well in rain or snow. RADAR is effective at detecting object distance and motion. It does not have the detail of LIDAR, but it is effective in a variety of weather conditions. Cameras can see in color, and therefore are potentially the best at object recognition, but they require the most data of the three types of sensors.
Computer vision entails collecting the visual sensor data, processing it using a computer, and interpreting it in a way that is useful to the vehicle, such as ignoring background scenery or recognizing a road hazard before applying the brakes.
For location data, fully autonomous vehicles use powerful on-board computers to produce local maps using sensor data. Local maps give the location of the vehicle relative to the objects in the environment. Global Positioning Sensor (GPS) equipment can provide a vehicle’s absolute location, but local maps are required when GPS data are not available, such as inside a parking garage.
Performing under a minimal risk condition refers to an autonomous vehicle’s ability to return to a safe location in the event of a system malfunction, or to operate safely in a degraded state. Since autonomous vehicles will be expected to operate in a variety of roadway types, speeds, geographic areas and environmental conditions, the circumstances of the minimal risk condition may vary.