Modeling the Formation of Language in Embodied Agents: Conclusions and Future Research