Traditional binaural models, such as those based on cross-correlation and equalization/cancellation algorithms, use a bottom-up approach. They simulate the functions of the ascending auditory pathway from signals at the ears up through higher-level structures, typically the midbrain. In contrast, top-down methods are frequently used in practical robotics applications, for instance, when applying simultaneous-localization-and-mapping (SLAM) techniques. Such top-down approaches, however, are generally absent from models that foster understanding of our auditory system. This talk focuses on binaural-localization models that integrate traditional signal-driven models with psychoacoustically motivated top-down structures, for example, by means of feedback loops. The aim is to design advanced models that operate in architectural environments in order to understand both the environment and activities within the environment. These models are presented in the context of the authors' current work on a contributed book entitled, “The Technology of Binaural Understanding,” which is in press at Springer. The discussed approaches include: models to estimate room-impulse responses and binaural-activity maps from running signals; algorithms to utilize head movements for strategically reading the environment; and algorithms for the automatic labeling of sonic events using deep-learning techniques. [Work supported by the EU-FET project “TwoEars,” Contract No. 618075, NSF BCS-1539276, and CISL (RPI).]Traditional binaural models, such as those based on cross-correlation and equalization/cancellation algorithms, use a bottom-up approach. They simulate the functions of the ascending auditory pathway from signals at the ears up through higher-level structures, typically the midbrain. In contrast, top-down methods are frequently used in practical robotics applications, for instance, when applying simultaneous-localization-and-mapping (SLAM) techniques. Such top-down approaches, however, are generally absent from models that foster understanding of our auditory system. This talk focuses on binaural-localization models that integrate traditional signal-driven models with psychoacoustically motivated top-down structures, for example, by means of feedback loops. The aim is to design advanced models that operate in architectural environments in order to understand both the environment and activities within the environment. These models are presented in the context of the authors' current work on a contributed b...