This paper presents a computational architecture that addresses the Grand Challenge problem of system integration for intelligent control software. Robotics research has made many advances in sensory processing, control, and planning, but few efforts focus on the problem of dynamically integrating the "best available" approaches into a single architecture to support development of robot control systems. The role of architecture will be discussed, and typical approaches will be reviewed. Then, a novel approach to intelligent system architecture will be presented. This new approach takes advantage of the maturation of object-based software engineering to help manage the complexity of intelligent machine development and attempts to combine some of the desirable properties from other modern robot control architectures.

Full Paper (postscript)

Last updated: 3 March, 1997 by Anthony Alford alford@vuse.vanderbilt.edu