Designing AI navigation for cars
I worked with Cloudmade, who make in-car systems, to lead the prototype design for an AI navigation app.
Design lead • User research • UX/UI/Interaction design • HMI design
The brief
Design an AI enabled, in-car navigation prototype.
Approach
Machine learning systems have to learn over time, so the challenge was to design a UX that could adapt as it learned more about who was driving the car.
This adaptive behaviour is something I think is going to become more prevalent in UX design for AI-capable systems. I wrote a blog post about designing for non-human users, which you can read here.
-
Design discovery included research into driver cognition, in-car testing of the test unit and UI concept work.
-
I refined the prototype UI and designed all of the different screens we required for the MVP.
Design discovery
Design discovery included research into driver cognition, in-car testing of the test unit and UI concept work.
As part of this process, I went out and recorded my own drives, to get a sense for things that would affect UX design, for instance, what causes driver distraction and when is the right time to make proactive suggestions for route changes and alerts.
Creating a visual identity for the AI navigation assistant was important too, as well as a tone of voice, as we’d be utilising conversation as well as manual interface interactions.
I designed multiple driver scenarios to demonstrate how multiple screens would work together. These included a HUD, a central console and a dashboard display.
UI design
After design discovery I moved on to refining the prototype UI and designing all of the different screens we required for the MVP.
We were going to utilise OpenMaps for core navigation, so we could create custom map components. Key considerations were things like driver awareness, driver feedback and effective use of motion design.
We tested various different approaches to using voice and motion sequences for things like notifications and suggestions. As the system had to be proactive, rather than just reactive to driver actions, we needed to make sure that any driver communication was timely and notifications were succinct and noticeable, but didn’t distract the driver from driving.