This is the first year of a three-year continuing award. The research addresses the continuation of exploration of the use of deictic models and pointing strategies for constructing goal- directed sensory-motor behaviors that are far more succinct than classical methods, and exploration of trial-and-error learning of these behaviors. The investigation involves deictic modeling of visual routines, in which realistic fixation routines are developed that can work with grayscale images, leading to visual behaviors that have the flexibility to take advantage of task context; deictic modeling of short-term memory that allows routines to use local memory; modeling of modular task scheduling that supports the production of complex visual behaviors involving multiple goals from simpler behavioral routines; and a deictic model of hand-eye control via teleassistance, in which humans are able to assist a robot in performing some of the more difficult control functions in a complex task.
