Using multimodal displays to signify critical handovers of control to distracted autonomous car drivers

Until full autonomy is achieved in cars, drivers will still be expected to take over control of driving, and critical warnings will be essential. This paper presents a comparison of abstract versus language-based multimodal warnings signifying handovers of control in autonomous cars. While using an autonomous car simulator, participants were distracted from the road by playing a game on a tablet. An automation failure together with a car in front braking was then simulated; a rare but very critical situation for a non-attentive driver to be in. Multimodal abstract or language-based warnings signifying this situation were then delivered, either from the simulator or from the tablet, in order to discover the most effective location. Results showed that abstract cues, including audio, and cues delivered from the tablet improved handovers. This indicates the potential of moving simple but salient autonomous car warnings to where a gaming side task takes place