Signals must remain discriminable in noisy environments
Aliases: signal-to-noise ratio · alarm masking · field discriminability
What it is
Signal discriminability under noise is the ability to detect a robot cue and distinguish its state amid background sound, bright light, occlusion, vibration, and competing alarms. Noise includes any competing information, not sound alone. A clear laboratory cue can disappear in deployment.
Why it happens
Detection depends on contrast with the background in intensity, spectrum, colour, location, and temporal pattern. Simply increasing brightness or volume captures attention but can mask other alarms, disturb occupants, and habituate. Distinct rhythm, spatial direction, and cross-modal redundancy improve effective signal-to-noise ratio, although similar patterns across robots confuse source identity.
Studying it
Target sites or reproducible worst-case backgrounds should be used to measure detection, confusion matrices, response time, localisation, misses, and false alarms. Distance, viewpoint, spectrum, illumination, concurrent devices, and sensory limitation can be varied. Performance during a realistic primary task is more informative than deliberate cue search.
Where it stops holding
Every modality has failure conditions: hearing protection attenuates sound, sunlight reduces projection contrast, and occlusion blocks lights. Salience cannot remove attentional limits or make every state highest priority. Hospitals, homes, and night settings also constrain noise and light disturbance.
Applying it
- Measure site spectra, illumination, sightline occlusion, and existing alarms before selecting contrasting channels and rhythms.
- Give safety cues redundancy without a common failure source and make multi-robot signals localisable.
- Test during the primary task with customary protective or assistive equipment, judging misses, false alarms, source confusion, and latency rather than visibility alone.