file.pdf (272.08 kB)
Download file

Autonomous critics

Download (272.08 kB)
journal contribution
posted on 01.01.1990, 00:00 by Sarosh Talukdar, Carnegie Mellon University.Engineering Design Research Center.
Abstract: "Decisions made in the early stages of design processes can have profound effects on later stages. Often, information on these effects can be obtained only after great delays, by which time it is too late to use in the design effort. The goal of the work reported here is to develop and demonstrate technologies for shortening information feedback loops to the point where crucial analysis and evaluation information is made available to designers continuously and automatically, as their pieces of the overall design evolve. Our approach is based on the use of expandable libraries of autonomous programs called critics.Each critic keeps track of a developing design, or a piece of a design, evaluates the design from the viewpoint of a downstream stage, and signals the designer when it detects a flaw. The research issues are four-fold. First, each critic must be able to understand the representation schemes in which the designer is working. Second, each critic must be able to decide when to act. Third, each critic must be able to report its critique in terms the designer will understand. Finally, a distributed framework must exist to support the expanding library of critics. A demonstration system with an initial set of three critics has been completed and is described in the paper."


Publisher Statement

All Rights Reserved