Non-technical Abstract: Robots have the potential for wide-ranging positive impacts on society in complex or dangerous applications such as disaster relief, elder care and the reshoring of manufacturing jobs. However, existing robots have had limited success is these domains, mainly because the planning and control algorithms are not robust to misconceptions in the robot's "understanding" of its environment nor to small imperfections in the robot's ability to execute the required actions. The overall goal of this project is to develop the sensing, planning, and control algorithms necessary to overcome these problems, and hence necessary to allow robots to work productively in complex domains shared with humans.
The key activities of this project are the development of new ways of representing uncertainty in the state of the world that support efficient planning for robots. These new representations and algorithms provide principled and practical methods of integrating perception and action in complex domains. The resulting algorithms are tested in the context of a real robot performing household tasks in a kitchen environment.
The project also involves a thorough integration of research and education. Graduate and undergraduate students are involved in all aspects of the research. Furthermore, the research in this project forms the basis of an undergraduate subject on robot planning algorithms under development at MIT.
The overall goal of this project is to develop the estimation, planning, and control techniques necessary to enable robots to perform robustly and intelligently in complex uncertain domains. Robots operating in complex, unknown environments have to deal explicitly with uncertainty. Sensing is increasingly reliable, but still inescapably local: robots cannot see, immediately, inside cupboards, under collapsed walls, or into nuclear containment vessels. Planning, whether in household and disaster-relief domains, requires explicit consideration of uncertainty and the selection of actions at both the task and motion levels to support gathering information.
In order to explicitly consider the effects of uncertainty and to generate actions that gain information, it is necessary to plan in belief space: that is, the space of the robot's beliefs about the state of its environment, which we will represent as probability distributions over states of the environment. For planning purposes, the initial state is a belief state and the goal is a set of belief states: for example, a goal might be for the robot to believe with probability greater than 0.99 that all of the groceries are put away in acceptable locations, or that there are no survivors remaining in the rubble.
This project is developing a systematic, integrated approach to finding plans efficiently in high-dimensional uncertain domains. By factoring the belief space and exploiting a decoupling between geometric and probabilistic reasoning, this approach can employ constraint satisfaction methods to generate good solutions relatively efficiently. This program of basic research provides conceptual, formal, algorithmic, and software results that are of use in mobile manipulation robotics, as well as artificial intelligence more generally, including applications from medical diagnosis and treatment to electronic commerce to managing energy production and distribution systems.
