Although Domestic Service Robots (DSRs) that support people in everyday environments have been widely investigated, the DSR’s ability to predict and describe future risks resulting from their own actions is still insufficient. In this study, we therefore focus on the linguistic explainability for the DSRs. Most existing methods do not explicitly model the region of possible collisions; thus, they do not properly generate descriptions for regions of possible collisions. In this paper, we propose Nearest Neighbor Future Captioning Model that introduces Nearest Neighbor Language Model to future captioning regarding possible collisions, which enhances the model output with a nearest neighbors retrieval mechanism. Moreover, we introduce Collision Attention Module, which extracts attention regions of possible collisions, which enables our model to generate descriptions that adequately reflect the objects associated with possible collisions. Experimental results demonstrated that our method outperformed baseline methods on the standard metrics.

会議名 : AICカンファレンス2023
開催地 : 慶應義塾大学日吉キャンパス
日時 : 2023年3月4日
第2章ポスター発表要旨
ポスター要旨-6