In-Vehicle Voice Assistant — 2020 Research Deck
A 2020 research deck on voice assistant UX for in-vehicle infotainment. Covers the problems voice assistants solve in the car, the current market landscape, and recommendations for improving the Telenav voice assistant experience.
Alan Tan · February 2020
Overview
- 01
What problems do voice assistants solve in the car?
- 02
The current voice assistants landscape.
- 03
How can Telenav improve upon current voice assistant solutions?
What problems do voice assistants solve?
Anytime your eyes are not on the road you increase the likelihood of getting into an accident.
Voice assistants improve driver safety by helping drivers maintain attention on the road — using their voice to control functions of the infotainment system.
The latest voice assistants in the market
Android Auto
CarPlay
Mercedes
BMW
See Mercedes and BMW go head-to-head →
Past voice assistant experiences have been mediocre
Compare the original Sync 3 VR system to the Polestar running Google Assistant for controlling temperature. You can see how far along voice assistants have matured over the years — today's systems sound more like you're having a conversation than in the past.
Sync 3 — older one-shot VR
Polestar — Google Assistant conversational
Limitations of CarPlay / Android Auto in-market
One-shot commands are not conversations.
Conversations require cooperation between the speaker and listener. In Android Auto, some tasks cannot be completed using voice only — for example, in a find a destination task the user must complete it by tapping on the screen.
Android Auto destination task — voice → touch handoff
Even conversations with real people can be problematic
Voice systems cannot be 100% effective in maintaining road and car safety because of diversion of attention.
Even a real conversation diverts attention from the road
How to improve upon the voice assistant experience
Talk more naturally to the assistant
- · The assistant should try to make sure the user is satisfied with its actions, and continue to listen to allow override of its decision.
- · The user should be able to barge in while the system is talking — and respond to the command or response without waiting for the system to finish.
Make tasks easier to complete
- · Users should be able to easily multi-task using their voice — continuing to view their original content while engaged in a secondary task.
- · Make it multi-modal: whatever is asked using voice should also be accessible via touch.
Be aware of its surroundings
- · The assistant is aware of the vehicle status and current states of apps — and can adapt appropriately to a driver's current context.
- · The assistant is aware of driver distractions, and can adjust based on those distractions.
Allow users to talk more naturally to the assistant
The assistant should try to make sure the user is satisfied with its actions, and continue to listen to allow override of its decision.
The user should be able to barge in while the system is talking and respond to the command or response — without requiring the user to wait for the system to finish talking first.
Natural-speech assistant — barge-in and override
The assistant should make tasks easier to complete
- Users should be able to easily multi-task using their voice, meaning they can continue to view their original content while engaged in a new task.
- Make it multi-modal: whatever is asked using voice should also be accessible via touch.
Multi-tasking through voice while content remains on screen
The assistant should be aware of its surroundings
- The assistant is aware of driver distractions, and can adjust based on those distractions.
- The assistant is aware of the vehicle status and current states of apps, and can adapt appropriately to a driver's current context.
Context-aware assistant — adapts to driver state and vehicle status
- voice UX
- automotive
- research