One important artificial intelligence research direction at the Beckman Institute examines computational aspects of biological intelligence. Theories of cognition, perception, problem solving, and so on must be implementable if they are to be explanatory in nature rather than simply descriptive. This requirement is often neglected, but can place important constraints on the kinds of theories that should be entertained. Some such constraints are abstract, derived from the fact that computation is necessary in realizing the intelligent behavior. Other constraints are more specific, due to the form of the architecture of the computational medium. Regardless, a theory of the mind, whether philosophical, psychological, or neuropsychological, must be computationally tractable to be viable. Observations and theories of human intelligence become guiding principles for artificial intelligence research helping to refine the cognitive theories. Interactions between this group and the Cognitive Science group are ongoing and fruitful.

Another important research thrust examines how computers interact with humans and how that interaction can be made more effective through intelligence on the part of the computer. Computers will be more easily accepted and more useful as tools as they behave more appropriately and predictably when in collaboration with humans. At the Institute, mobile robots, robotic arms, and various sensing devices are used to integrate sensing, planning, navigation, and autonomous plan execution. One project applies the constraint approach to planning finger grasps for dexterous manipulators. In a more logic-oriented approach to constraining inference, Beckman researchers are applying learning to the task of planning in several real-world domains. Robotic grasping and manufacturing process planning are characterized by uncertainty, complexity, and resistance to conventional formalization.

Computer recognition of human faces is also under study. This is particularly challenging when the perceived image is not explicitly present in the database. Knowledge of prior views of a person and the system's general internal 3-D model of the human head must be combined. Face recognition promises many useful applications, including access control, credit card identification, and law enforcement. In a related project, Beckman researchers are investigating ways to infer 3-D shape and layout of scenes from visual cues, including texture changes, motion, and stereo differences. Spatial and temporal understanding allows development of schemes for representation, navigation, and animation. These schemes relate to structures and mechanisms that have evolved biologically, such as eye movements, eye accommodation, peripheral vision, and foveal acuity.