To successfully operate in human environments, a robot needs to be able to properly interpret its surroundings and manipulate human objects. In human environments such as a household, we can bring to bear significant information about the expected spatial layout and the function of common household objects. However, for most reasoning methods for robotic perception and manipulation, humans play a secondary role. In this proposal, we propose to make humans central to our reasoning algorithms. We argue that such explicit modeling and consideration of humans, even when they are not present, will enable robots to better perceive, manipulate, and plan. Through such reasoning, we will make advances in the following areas: (a) Modeling 3D Scenes with Objects and Humans, (b) Human Activity Detection and affordances, and (c) Robot Manipulation for assistive tasks.
This research has the potential to significantly impact the quality of life of a large number of elderly people. In fact, one of the most promising applications of this research is in assistive care for the elderly and those with difficulty in performing daily tasks. Assistive care for the elderly is an increasingly important problem: a growing fraction of the US population is elderly; by 2030 about 1 in 5 Americans will be over 65. Many elderly people (including about 50% of those 85 years or older) require some personal assistance. Just to mention one basic example: a key concern in elderly care is that the subject does not drink enough water throughout the day and becomes dehydrated. Our robot (or a monitoring system) can observe the water consumption, remind the subject to drink water when needed, and even bring some if desired. Other outreach activities include incorporating young students and minorities into robotics research, curriculum development, and open-source robotics software.
