Planning for a Class of Sequential Control Problems
1989 (English)Report (Other academic)
Sequential control is probably the most common control problem in industry. Despite its importance, fairly little theoretical research has been devoted to this problem, and practical problems are typically solved by ad hoc methods, sometimes quite time-consuming.
Sequential control is really a planning problem. In this thesis, we describe and analyse sequential problems in a systematic way based on a formalism from artificial intelligence (AI). We focus on solving the planning prblem, i.e., the problem of finding a sequence of actions which transforms a given initial state into a desired final state.
The complexity of planning using a method based on exhaustive search increases exponentially with the number of state variables, ans so there is a need for alternative methods.
We suggest a method for solving the planning problem for a class of sequential problems. The complexity of this method increases polynomially with the number of state variables. The method is proven correct and complete, and it always returns a minimal plan if there is a plan at all. We also discuss how planning can be handled outside this subclass.
Place, publisher, year, edition, pages
Linköping: Linköping University , 1989. , 46 p.
LiTH-ISY-I, ISSN 8765-4321 ; 1007
Sequential control problems
IdentifiersURN: urn:nbn:se:liu:diva-104094OAI: oai:DiVA.org:liu-104094DiVA: diva2:694543