Lingodroids are language learning robots that play games to construct shared lexicons. These fascinating mobile robots have learnt lexicons for places, distances, directions, and time. Our latest work demonstrates how to handle referential uncertainty, that is, how to resolve a word’s meaning between multiple candidates.
Further, we have demonstrated language learning between robots that have different embodiment and cognitive capabilities. They differed in their sensors (camera images versus laser ranges); algorithmic approach (biologically-inspired network versus probabilistic particle filter); and spatial representation (semi-metric topological map versus occupancy grid).