The Sound of Safety: How Zoox Designs Audio With Intent
The sound of safety: How Zoox designs audio with intent
Zoox keeps innovating its sound design to improve safety
At Zoox, we're building a new form of urban transportation. A bidirectional, fully symmetrical robotaxi designed around riders, not drivers. That distinction shapes everything. When you remove the constraints of a traditional vehicle, you can build the experience from the ground up, from how you think about seating, doors, controls, and even sounds.
Reimagining sound poses unique opportunities. Because there is no human driver in a Zoox robotaxi, there is no one to make eye contact with a pedestrian or signal intent with a wave. Automated driving systems need new ways to fill that gap. How does a robotaxi communicate in the absence of a driver? How should it communicate with pedestrians, cyclists, and other road users while being a good citizen of the road?
Designing sounds with intent
Inside the robotaxi, sound plays a key role in guiding riders through their whole journey, from start to finish. A mix of chimes and voice cues provides feedback, delivers important information, and prompts action when needed.
Outside the robotaxi, we’ve rethought how sound can communicate with the world. The robotaxi carries 32 speakers at each end, capable of beamforming - the ability to pinpoint sound in a specific direction and intensity, allowing us to reach the right person with the right sound, at the right moment. A traditional car horn emits sound in all directions, potentially contributing to unnecessary noise for everyone nearby. Directional audio lets us be more precise, alerting a pedestrian who is crossing ahead without startling the cyclist to the right. This precision keeps journeys flowing smoothly without adding unnecessary noise to the environment.
That precision extends to the horn itself. Because our robotaxi is built from the ground up, we had the opportunity to rethink how we can further communicate intent with the horn. The result is a horn with multiple expressions: a louder horn for communicating with other vehicles, a friendlier horn for pedestrians, and a rear horn intended to help prevent rear-end collisions. Each one is calibrated for its context.
Rethinking sound as a communication tool raises a larger question: could intentionally designed sounds do even more to advance safety?
Partnering with VTTI to evolve the language of sound
Researching better methods inspired Zoox Safety, Experience Design, Research, and Sound Design teams to collaborate with the Virginia Tech Transportation Institute (VTTI) and explore novel ways to use non-verbal audio cues to signal the intent of a robotaxi to the people around it.
The goal was specific: could these cues communicate a vehicle's intent as effectively as conventional signals, such as a honk or road noise? And how would that differ for people with low vision or visual impairments?
The study was conducted at the Virginia Smart Road, a closed test track at VTTI in Blacksburg, Virginia. Forty participants, balanced by gender and vision status (low-vision or visually impaired), participated in the study. Using a human-operated Zoox test vehicle, participants experienced two real-world scenarios: a crossing outside the crosswalk and a crossing in the crosswalk.
In the mid-block crossing scenario, the purpose-designed audio cues incorporated intentionally defined sound principles related to urgency, semantic meaning, assertiveness, and awareness. The research suggests these were as effective as a horn at influencing pedestrians' decisions not to cross, and were almost unanimous in increasing awareness of the vehicle's presence.
Listen to the new research sounds on our Soundcloud.
Participants also perceived the sounds as less assertive and less aggressive than those of a traditional car horn. The research suggests that in certain scenarios, thoughtfully designed sounds can replace honking with a more intentional approach. In the crosswalk scenario, the findings pointed toward a longer-term opportunity, with continued research needed into how people learn to interpret new audio signals over time.
More research is needed, but these initial results are promising. The research reflects exactly the kind of iterative progress Zoox is committed to by learning from real-world behavior and building toward a clearer, more effective language between robots and the people around them, while also potentially reducing urban noise pollution.
For additional details, read the VTTI news article.
