Progress and Prospects of Multimodal Fusion Methods in Physical Human-Robot Interaction: A Review

Recent advances in physical Human-Robot Interaction (pHRI) have shown the potential and feasibility of robot systems for active and safe collaboration with humans. This induces high demand for real-time perception about the dynamic unity formed by robot, human, environment and operating objects. Single sensory modality is inadequate to give a robust estimate of the state of the unity. This paper provides a comprehensive review of state-of-the-art multi-modal fusion methods exploring their classifications, feasibilities and challenges based on the fact that pHRI is spatio-temporal sharing, multi-modal and multi-task. Finally, several directions of research are recommended based on the discussion of multi-modal fusion methods.