This dataset represents ambient data collected longitudinally in 49 community homes. This is a resource for analyzing naturalistic behavior in a home and building activity recognition models that operate in the wild. Data are collected continuously from ambient sensors while residents perform their normal routines. The data fields are date, time, sensor identifier, and message. The sensors consist of PIR (motion) sensors and magnetic door (open/close) sensors. Sensors are attached to ceilings and identified by their location in the home (e.g., Bathroom, Bedroom, DiningRoom, Bed, Bath, OfficeChair). If a home contains more than one room of a given type, the corresponding sensors are distinguished by a trailing letter to differentiate the rooms (e.g., BedroomA, BedroomB). The lens of most motion sensors are constrained to cover a 1 meter diameter area. To detect movement in a larger area, an unconstrained sensor is angled to cover an entire room or region and is indicated by Area (e.g., BedroomArea). There is one file per home. These datasets represent entries with raw sensor identifiers - a version of the data with mapped sensor names can be found in zenodo dataset 10.5281/zenodo.15708568. The resource also includes text files that denote the region of the home (or intersection of regions) where each sensor is located. Note: Other CASAS smart home and smartwatch datasets are also available, look for more at https://zenodo.org/communities/casas.

Citation: Please cite the following paper when using this dataset:Cook, D., Crandall, A., Thomas, B., & Krishnan, N. (2013). CASAS: A smart home in a box. IEEE Computer, 46(7):62-69, 2013. https://doi.org/10.1109/MC.2012.328