When humans and robots work together on cooperative tasks, each partner needs appropriate space to act. The new DFG project SEMIAC explores perceptual and behavioral strategies of a mobile robot which considers implicit non-verbal signals of a human collaborator for its own navigation and acting. SEMIAC stands for Semantic Environmental Models for Implicit Action Coordination with humans. The joint DFG project with University of Magdeburg started in December 2022 and is funded for three years. While Madgeburg focusses on the perception and interpretation of interaction-relevant features (head and body poses, facial expressions, scene context), in Bielefeld this is combined with adaptive robot behaviors considering human intentions. The overall goal of the project is to reduce interruptions and interaction breakdowns in the close collaboration between humans and robots.