Communication between embodied AI agents has received increasing attention in\nrecent years. Despite its use, it is still unclear whether the learned\ncommunication is interpretable and grounded in perception. To study the\ngrounding of emergent forms of communication, we first introduce the\ncollaborative multi-object navigation task CoMON. In this task, an oracle agent\nhas detailed environment information in the form of a map. It communicates with\na navigator agent that perceives the environment visually and is tasked to find\na sequence of goals. To succeed at the task, effective communication is\nessential. CoMON hence serves as a basis to study different communication\nmechanisms between heterogeneous agents, that is, agents with different\ncapabilities and roles. We study two common communication mechanisms and\nanalyze their communication patterns through an egocentric and spatial lens. We\nshow that the emergent communication can be grounded to the agent observations\nand the spatial structure of the 3D environment. Video summary:\nhttps://youtu.be/kLv2rxO9t0g\n