About the LCI
The Laboratory for Computational Intelligence supports research in computational reasoning, perception, and robotics. The objectives are to identify the constraints and to define the computations that make intelligent reasoning, action and perception possible. LCI provides the intellectual environment, shared resources, and technical support for a wide range of research projects directed by the participating faculty.
The research strategy is to investigate the theory, design, and implementation of intelligent systems that link perception to action. This requires mathematical analysis of the problems to be solved, development of a theory of knowledge representation for those problems, design of algorithms and architectures to solve the problems, and implementation and testing of these systems in realistic settings.
The long-term objective is to identify the constraints and to define the computations that make intelligent reasoning, action and perception possible, in man or in machine.
One focus for the research is visual perception for robotics and for remote sensing. Research includes: knowledge representations for vision; model-based systems for recognition and motion tracking; analytic methods for early vision, especially shape from shading and photometric stereo; and parallel algorithms and architectures for effective implementation and integration of vision modules. Applications are to robot vision systems that recognize known 3-D objects, to navigation and control of mobile robots, and to natural resources management, using satellite and aerial imagery of the earth's surface.
A second focus of research is on logical reasoning methodologies with applications to decision making and planning, diagnosis, recognition, theory formation, and belief revision. One of the main avenues of research is the development of representational and computational mechanisms for reasoning and decision making in the presence of uncertain and incomplete information, and the application of these ideas to such domains as planning and diagnosis. This research area also forms the underpinning of work into natural language processing and discourse interpretation.
A third focus for our research is the design of frameworks for specifying, modeling, verifying and synthesizing robotic control systems using constraint-based techniques and logical specifications.
One on-going research project, named Dynamo, makes use of multiple radio-controlled vehicles to investigate problems in multi-agent robotics. This project integrates research on planning, reasoning, perception and control. The goal is to generate the appropriate cooperative and competitive behaviors for complex tasks such as playing soccer.
Another project is developing a new class of symmetric, high degree of freedom robots called Platonic Beasts with modular, multi-purpose limbs for locomotion and manipulation.
One IRIS project is to build vision systems for object recognition, motion tracking, navigation, and diagram understanding. These vision systems will be used for robotics and telerobotics applications.
Another IRIS project is to develop theories of reasoning, using mathematical logic as a basis for hypothetical reasoning, default reasoning, belief revision, and model-based diagnosis.
Many other projects are described in the Web pages of individual lab members.