As the population ages, there is an increasing demand for support with activities of daily living (ADLs) in the homes of older adults and in long term care. Socially assistive robots (SARs) have shown success in increasing ADL independence by providing adaptive assistance for a variety of tasks including eating and dressing. The objective of this thesis is to develop new technologies for improving the design of SARs for ADL assistance. Namely, this thesis develops:​ 1) a novel social robot and wearable sensor system for assisting with the ADL of dressing, and 2) a new deep learning ADL recognition architecture for autonomously recognizing and monitoring known and unknown ADLs. Experiments evaluate performance using classification accuracy and real-time functionality using metrics such as usefulness and reliability. Results for classification performance show the developed methods outperform existing work while interaction experiments validate the systems for use with a variety of diverse users.