Acoustic Management of Robotic Ego-Noise

You are reading version 1 of this direction, frozen 2026-10-04. The latest version is version 2.

Version 1 of 271 papers104 conceptsfrozen 2026-10-04

When we design robots, we often focus on what they see and how they move. However, as robots transition from isolated factory floors to shared domestic spaces, a neglected sensory dimension has emerged as a primary barrier to adoption: sound. Specifically, ego-noise—the whirrs, clicks, and hums generated by a robot’s own motors, cooling fans, and joints—creates a persistent acoustic interference that degrades both machine perception and human trust.

The Problem of Proximity

In traditional audio processing, "noise" is usually external, like a television in the background or wind. But for a robot, the most disruptive noise source is itself. Because microphones are typically mounted directly onto the robot's chassis, they are physically proximate to the actuators. As Yihui Fu and colleagues note in the Alpha-Mini Speech Challenge, this proximity creates a low signal-to-noise ratio that standard speech datasets simply don't account for.

This internal noise floor is not just a nuisance; it is a perceptual veil. It masks the subtle acoustic signatures a robot needs to understand its world. For instance, Yefei Wang and Huaping Liu have shown that internal noise can overwhelm the sounds of objects being manipulated, such as the sloshing of liquid in a bottle, which is vital for audio-visual grounding.

The Psychological Barrier

Beyond technical performance, ego-noise is a social liability. Aimee Allen and Dana Kulic found that these "consequential sounds" significantly increase human anxiety and decrease the willingness to share a workspace (Robots Have Been Seen and Not Heard). A robot that sounds like a grinding machine feels like a piece of industrial equipment, regardless of how friendly its visual design might be.

For assistive technology, the stakes are even higher. In the development of robotic guide dogs, Hochul Hwang and Sunghoon Ivan Lee argue that mechanical noise is a functional safety hazard. For blind or low-vision users, the robot’s noise can mask critical environmental cues—like the sound of an approaching car or a pedestrian's footsteps—effectively "blinding" the user's auditory awareness (Lessons Learned from Developing a Human-Centered Guide Dog Robot).

Mitigating Noise Through Awareness

Researchers are tackling ego-noise through three primary lenses: signal processing, control theory, and path planning.

1. Proprioceptive Filtering If a robot knows how it is moving, it can predict the noise it is about to make. Joon-Hyuk Chang and Moa Lee demonstrated that by feeding a robot's motor "on/off" status into a deep neural network, the system can create noise-adaptive representations that significantly improve speech recognition. Essentially, the robot uses its own internal state to "subtract" its voice from the environment.

2. Quiet Control Rather than filtering noise after it happens, some researchers aim to prevent it at the source. Shangqun Yu and the UMass Amherst team developed quiet locomotion controllers for quadruped robots. By optimizing foot-swing trajectories to minimize impact sounds, they achieved a 10 dB reduction in perceived loudness—a 50% decrease—making the robot viable for high-stakes assistive tasks.

3. Socially Aware Navigation Finally, robots can learn to be "polite" by choosing paths that minimize their acoustic footprint. The ANAVI framework by Vidhi Jain and Yonatan Bisk allows robots to predict how their movements will sound based on room geometry and materials. This enables a robot to decide when it needs to be silent and when its noise will be safely masked by ambient sounds like a running dishwasher or a television.

The Path Forward

The shift toward "quiet" robotics represents a move from viewing robots as isolated agents to viewing them as members of an acoustic ecosystem. As highlighted by the RoboVox Challenge led by Mohammad Mohammadamini, the future of the field lies in moving beyond "smart home" simulations toward rugged, hardware-aware models that treat a robot's own mechanical signature not as an error, but as a known variable to be managed.

Papers in this version

Showing 60 of 71 papers.

History & citationVersion 2 · 2026-10-04

Current · Permanent link

  1. Version 2 (latest) · 2026-10-04 · 56 papers
  2. Version 1 · 2026-10-04 · 71 papers
Cite this direction

Latest version (a living page):

@misc{lacuna_robotic_ego_noise_interference_10ed2017,
  title = {robotic ego-noise interference},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/robotic-ego-noise-interference/rdr_10ed20177a044483a7e813815c6b0e0c},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-21},
}

This exact version (version 1, frozen text):

@misc{lacuna_acoustic_management_of_robotic_ego_noise_10ed2017_v1,
  title = {Acoustic Management of Robotic Ego-Noise},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/acoustic-management-of-robotic-ego-noise/rdr_10ed20177a044483a7e813815c6b0e0c/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-21},
}