Future Work

~ ~ The future of this project will be determined by its usefulness to the Competitive Product Development Institute (CPDI) at Digital Equipment Corporation. However, even if SNEAKERS meets the needs of CPDI, it still has room for improvement.

All of the unimplemented options in SNEAKERS can be implemented. There is the potential for forty agents in the system as it stands, but only twenty-six are included. Also, the agent types and aspects are not all inclusive, so even more types of agents could be added. The rule sets for each agent could be expanded. Some agents have only one rule, while others have ten.

There is no geometric reasoning in the system. For example two supports are allowed to pass through each other. Bracing is handled as an arbitrary length rod of undecided material. Bracing could be changed to be a particular rod length and material type, just as supports are, and the tower designed to accommodate the allowable bracings as well. All of this would make a better system, but would not greatly increase its usefulness in educating managers and engineers.

If this interface were to be used as a realistic general CE support system, then there would have to be extensive changes. Areas that would have to be added include: negotiation, database management, agent communication, and support for multiple users.

Negotiation between agents with conflicting suggestions is a difficult problem, but it would be necessary in a general system to weed out some of the comments that the user sees, through negotiation. Otherwise, the number of comments would be unmanageable. Precedence for this work is also given in [Sriram et al 1989].

More work would have to be done to expand the COOL representation of the database. There is no support for version control and time dependence in the databases. Research into these areas of database management can be found in [Katz & Chang 1987], [Katz 1990], and [Snodgrass 1990].

A more realistic system would need to take into account the research reported in [Daley & Fotta 1991] and [Fotta & Daley 1991] which are concerned with communication among agents with varied backgrounds. The domain in SNEAKERS is simple enough that the agents can communicate with an easily managed vocabulary.

A full-blown system would have to use multiple processes on multiple machines and be able to support multiple users. It should have a means of keeping all users informed of the agents' comments and allow communication between the users.

There are many ways in which this project could be expanded. SNEAKERS itself is a fairly complete system, and needs no major improvements to fulfill its task. But it is not a general CE support system, and therefore, would have to be radically altered in order to fill that role.