Use this URL to cite or link to this record in EThOS: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.336332
Title: Contributions to time-bounded problem solving using knowledge-based techniques
Author: Chatterjee, Niladri
ISNI:       0000 0001 3530 3018
Awarding Body: University of London
Current Institution: University College London (University of London)
Date of Award: 1995
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Abstract:
Time-bounded computations represent major challenge for knowledge-based techniques. Being primarily non-algorithmic in nature, such techniques suffer from obvious open-endedness in the sense that demands on time and other resources for a particular task cannot be predicted in advance. Consequently, efficiency of traditional knowledge-based techniques in solving time-bounded problems is not at all guaranteed. Artificial Intelligence researchers working in real-time problem solving have generally tried to avoid this difficulty by improving the speed of computation (through code optimisation or dedicated hardware) or using heuristics. However, most of these shortcuts are likely to be inappropriate or unsuitable in complicated real-time applications. Consequently, there is a need of more systematic and/or general measures. We propose a two-fold improvement over traditional knowledge-based techniques for tackling this problem. Firstly, that a cache-based architecture should be used in choosing the best alternative approach (when there are two or more) compatible to the time constraints. This cache differs from traditional caches, used in other branches of computer science, in the sense that it can hold not just "ready to use" values but also knowledge suggesting which AI technique will be most suitable to meet a temporal demand in a given context. The second improvement is in processing the cached knowledge itself. We propose a technique which can be called "knowledge interpolation" and which can be applied to different forms of knowledge (such as symbolic values, rules, cases) when the keys used for cache access do not make exact matches with the labels for any cell of the cache. The research reported in this thesis comprises development of cache-based architecture and interpolation techniques, studies of their requisites and representational issues and their complementary roles in achieving time-bounded performance. Ground operations control of an airport and allocating resources for short-wave radio communications are two domains in which our proposed methods are studied.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.336332  DOI: Not available
Keywords: Computer software & programming
Share: