Discovering innovations that matter since 2002

Autonomous vehicles | Photo source Pixabay

Tech Explained: Autonomous driving


This Tech Explained focuses on autonomous vehicles - explore the workings behind the growing number of driverless cars

It seems like science fiction – get into a car, tell it where to go, then relax while the car drives itself to the destination. In fact, this day may be a lot closer than we think. There are already a large number of driverless vehicles in testing around the world. At Springwise, we have also seen a big increase in innovations related to autonomous driving, including AI software that understands how people react and a kit that can retrofit cars for autonomous driving. So how does driverless technology work?

Driverless, or autonomous, vehicles, use sensors and software for control and navigation. Most self-driving systems work by using sensors to create and maintain a map of their surroundings. These Lidar (light detection and ranging) devices measure distances using pulses of light. Software then analyses this data and uses it to build a map of the surroundings and to plot a path.

In addition to Lidar, autonomous vehicles also use sensors such as radar and global positioning system antennas. These help the vehicle to determine the location of other objects, such as other vehicles, pedestrians and bicyclists. Cameras recognise traffic lights, street signs, road markings and other types of signals. Algorithms and predictive modelling tell the vehicle how to avoid obstacles, how to follow traffic rules and how to discriminate between similar objects, such as a bicycle and motorcycle.

In order to sift through all the data and make decisions on how to respond, autonomous vehicles also feature an on-board computer that can analyse the data. It is impossible to write a program that can take into account every potential traffic situation. Therefore, the programmes used by autonomous vehicles use machine learning or artificial intelligence to learn how to recognise new situations. For example, one system learns how to identify pedestrians by first analysing thousands of photos of people walking or running across or near roads.

Once the information is analysed, the onboard computer sends ‘instructions’ to the vehicles’ ‘actuators’. These are the devices which control acceleration, braking, and steering.

Some self-driving vehicles are being developed which will be ‘connected’. These vehicles will be able to communicate with other vehicles and with infrastructure, such as traffic lights. In the future, these connected vehicles could help to reduce traffic congestion and accidents.

One obstacle is that self-driving vehicles have difficulty duplicating subtle, nonverbal communication, such as making eye contact, that goes on between pedestrians and drivers. Sensors also struggle with some weather conditions, such as heavy rain. They can have trouble in tunnels and in heavy traffic. For these reasons, vehicles that are completely capable of self-driving in every situation are probably still much further off. Presently the reality is vehicles that can often self-drive, but need a driver to take over on occasion. Will autonomous technology one day allow vehicles to drive without any human control?