Rethinking Legibility in Social Robot Navigation: Impact of Intent Representation and Human Attention

Legible robot motion lets nearby people anticipate what a robot will do next, but most prior work on legibility assumes open spaces, static interactions, and fully attentive pedestrians. Through two hallway user studies with a Hello Robot Stretch 2, we show that legibility works best when a robot expresses its immediate interaction choice, e.g. which side it will pass on, rather than its final destination, and that this benefit holds up even when pedestrians are cognitively distracted.