Distributed automated reasoning: the role of knowledge in distributed problem solving
Douglas John Macintosh · 1990
Research in distributed artificial intelligence (DAI) has largely focused on distributed problem solving issues that are domain specific. It is not common to find domain independent research. It is also atypical in DAI to find research whose domain did not require a significant amount of effort in order to implement. There are also many recurring themes, or focal points in DAI research which suggest that there are characteristics of distributed problem solving systems that are domain independent. In this thesis, we address each of these issues and make three distinct contributions to DAI with respect to them. One such contribution addresses the need within the DAI community for a generic distributed emulation facility and development environment for distributed intelligent systems. In this research, we have developed such a system called SIMULACT. This system is domain independent and is itself a distributed system. SIMULACT can perform simulation that is time driven, event drive, or a combination of both. We demonstrate in this research the value of having a testbed such as SIMULACT, by building on top of it a powerful domain independent distributed reasoning system called DARES. DARES is the second contribution that this research makes to DAI, since it is the first distributed reasoning system that has the ability to reason about distributed domains. Furthermore, DARES is considered to be domain independent because it employs the first-order predicate calculus, which can be used to represent a wide variety of domains. By being a domain independent distributed system, DARES is a vehicle in which domain independent data can be experimentally obtained with respect to distributed problem solving. This ability to access distributed problem solving in a domain independent fashion is not typically done in DAI research. The third contribution that this research offers, is that DARES was used to obtain experimental data indicating the nature of the role that knowledge plays in distributed problem solving. We consider this significant because the behavior revealed by this data is domain independent and characteristic to distributed problem solving. This was done in one research effort, where the recurring themes in DAI mentioned above are the result of numerous efforts.