The motivation for the Grand Challenge effort stems from three main hypotheses.

The first hypothesis is that, given complete enough observation data on a large enough population of subjects, it is possible to design learning algorithms which are able to generate sensible interpretations of sensor data and correct predictions of user's behavior and motion. In particular, the Grand Challenge effort will provide the data necessary to support the development of the learning and data mining approaches used in the perception and mobility thrusts. Our hypothesis is that models of behaviors and actions can be extracted from the large corpus of data collected over time from subjects, and that the models can be used in conjunction with on-line data to predict user motion and intent, and to detect anomalies in physical behavior.

The second hypothesis is that there are novel sensors, or novel sensing strategies, which can drastically improve the applicability of the data to specific applications. They include new types of accelerometers, and body worn video and audio recording devices.

Finally, a third hypothesis is that a greater selection of sensors will lead to more effective analysis and prediction algorithms through sensor fusion.

The type of data collected under the Grand Challenge effort and the type of processing planned for this data are guided by these three hypotheses.