Using Sound to Get Around

When it comes to environmental sensing in robotic systems, one size most definitely does not fit all. Conventional cameras are perhaps the most frequently used option in high-end systems, but when light levels are low, they tend to lose accuracy or even fail outright. LiDAR overcomes this particular issue, however, environmental factors like fog or smoke render it ineffective. These types of trade-offs continue to pop up as one goes down the list of visual sensing instruments, from thermal cameras all the way to more exotic equipment.

So what are we to do if none of the available options meet the needs of our application? Aside from twiddling one’s thumbs and waiting for a technological breakthrough, the best path forward is to choose an entirely different sensing modality. And of the available alternatives, acoustic sensing is one of the most capable. I mean, if it is good enough for dolphins, it should be good enough for us too, right?

Very often it is. However, acoustic sensing setups have problems of their own. In particular, they require a lot of complex, specialized equipment, and as you might expect, they perform poorly in noisy environments. But that may change in the near future, thanks to the work of a team of researchers at Seoul National University in Korea. They have developed a simple auditory sensor called 3DAR (3D Acoustic Ranging) that uses just a single microphone to locate humans — and it even works in noisy environments, like factories.

By mimicking the ability of dolphins and bats to selectively listen to sounds from specific directions, the team engineered a meta-structure phase cancellation mechanism. This mechanism artificially adjusts the phase of sound waves coming from different paths, amplifying sounds from desired directions while canceling out unwanted noise. After integrating this meta-structure with a single microphone mounted on a rotational device, the team managed to replicate 3D beamforming previously achievable only with multi-sensor arrays. This enables accurate 3D positioning of humans or other objects with just one sensor.

Moreover, taking cues from dolphins’ use of dual-frequency communication, the 3DAR system separates communication channels into audible frequencies for human-robot interaction and inaudible frequencies for robot-to-robot communication. This clever design minimizes interference and allows multiple robots to coordinate without relying on complex network infrastructures.

The researchers see many potential use cases for their technology. In industrial settings, for instance, 3DAR can track workers’ positions in real time ro prevent accidents by making robots aware of their human colleagues’ locations. It could also enable voice-based commands to free workers from using gestures or buttons, increasing operational efficiency.

Field tests have already demonstrated the system’s effectiveness. A quadruped robot equipped with the 3DAR sensor successfully interacted with humans through sound commands and even detected the source of gas leaks by locating sounds accurately. In noisy factory environments, the technology achieved over 90% accuracy in locating human sound sources and an impressive 99% accuracy in robot-to-robot sound localization, even under challenging conditions.

Looking ahead, the researchers plan to further enhance their system by integrating it with artificial intelligence, including large language model-based cognitive systems. This would allow robots not only to locate sound sources, but also to understand and interpret the meaning behind sounds much like humans do.