Can AVs master the ‘language’ of driving?
Thursday, 22 February 2018
Driving is a game of written, but also unwritten rules. Driverless cars can be programmed to obey traffic rules, interpret signs, and respond to lights and road markings. Although still in its infancy, technology is now being developed that will allow autonomous vehicles (AVs) to communicate with each other enabling smoother navigation and reducing accidents. However, the unwritten rules of the road are where things get tricky.
Self-driving cars also need to communicate with human-driven vehicles, cyclists and pedestrians, to understand the social signals and gestures used in traffic - which also vary according to country or region - and be able to adapt to unusual situations such as accidents, lane closures and technical failures. This is the complex language of driving that the autonomous car yet has to master. It includes the ‘language of headlights’, such as flashing to give way to someone, or as a warning or a gesture of thanks. It also includes the language of bumper stickers to convey information warnings, and ‘sign language’, hand gestures that tell someone to slow down or sometimes convey hostility.
A less than friendly 'road' gesture
Sounds are another form of communication, with horns used to convey approval, disapproval, warnings or emergencies.
Different countries have their own ‘code’ of honking to convey complex instructions, like in Cairo, where four short honks followed by one long one means ‘pay attention!’
Try this in the UK, though, and you might end getting a less than friendly hand gesture in return.
Humans are able to consider the situational context and react accordingly. But an autonomous car? It could perhaps be ‘taught’ to read road signs or recognise stickers, but unexpected situations require human interpretation. A driverless car could misinterpret a hand gesture or sign, possibly resulting in a catastrophe. So how can they be programmed to recognise these often non-standard visual and auditory signals? As driverless cars become increasingly popular, it remains to be seen whether engineers will one day find a way of programming driverless cars to master the subtle yet essential language of driving.
Adapted from an article by Abdesalam Soudi, Sociolinguist, University of Pittsburgh