As robots are deployed in spaces with multiple humans with limited viewpoints, they will need to incorporate balancing these audiences into their motions. This proposal seeks to explicitly model several aspects of the communication to improve performance: understanding which members of their audience are physically able to observe their actions, balancing the constraints of multiple observers, and weighing efficiency and clarity against communication goals. To investigate these, we propose a series of experiments using the scenario of a restaurant, a multi-human environment where each audience member has a different view of the scene.