Use this URL to cite or link to this record in EThOS:
Title: Distributed abductive reasoning : theory, implementation and application
Author: Ma, Jiefei
ISNI:       0000 0004 2711 6310
Awarding Body: Imperial College London
Current Institution: Imperial College London
Date of Award: 2012
Availability of Full Text:
Access from EThOS:
Access from Institution:
Abductive reasoning is a powerful logic inference mechanism that allows assumptions to be made during answer computation for a query, and thus is suitable for reasoning over incomplete knowledge. Multi-agent hypothetical reasoning is the application of abduction in a distributed setting, where each computational agent has its local knowledge representing partial world and the union of all agents' knowledge is still incomplete. It is different from simple distributed query processing because the assumptions made by the agents must also be consistent with global constraints. Multi-agent hypothetical reasoning has many potential applications, such as collaborative planning and scheduling, distributed diagnosis and cognitive perception. Many of these applications require the representation of arithmetic constraints in their problem specifications as well as constraint satisfaction support during the computation. In addition, some applications may have confidentiality concerns as restrictions on the information that can be exchanged between the agents during their collaboration. Although a limited number of distributed abductive systems have been developed, none of them is generic enough to support the above requirements. In this thesis we develop, in the spirit of Logic Programming, a generic and extensible distributed abductive system that has the potential to target a wide range of distributed problem solving applications. The underlying distributed inference algorithm incorporates constraint satisfaction and allows non-ground conditional answers to be computed. Its soundness and completeness have been proved. The algorithm is customisable in that different inference and coordination strategies (such as goal selection and agent selection strategies) can be adopted while maintaining correctness. A customisation that supports confidentiality during problem solving has been developed, and is used in application domains such as distributed security policy analysis. Finally, for evaluation purposes, a flexible experimental environment has been built for automatically generating different classes of distributed abductive constraint logic programs. This environment has been used to conduct empirical investigation of the performance of the customised system.
Supervisor: Broda, Krysia ; Russo, Alessandro Sponsor: Deputy Rector's Award ; International Technology Alliance
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral