Abstract
Although it is often natural to specify a task to a problemsolver as a conjunction of goals to be attained, the sequential attainment of these goals is often complicated by interactions between them. The interaction problem is discussed and a process is described which allows the use of simple problem-solving techniques which assume interactions do not occur. Such techniques have received a great deal of attention, for example, in the STRIPS system (SRI). We show that when interactions do occur, information which enables problem solving to continue can easily be extracted from the existing data structures. A problem solver, INTERPLAN, has been designed and programmed which incorporates the process.
INTERPLAN tries to find a sequence of individual goals, in the order they will be considered, which will solve a task without interaction. It does this by "debugging" a given initial sequence (usually the given order of the goals in a conjunct). This process is similar to that used in the HACKER system (MIT) for a more procedural representation of a problem.
INTERPLAN tries to find a sequence of individual goals, in the order they will be considered, which will solve a task without interaction. It does this by "debugging" a given initial sequence (usually the given order of the goals in a conjunct). This process is similar to that used in the HACKER system (MIT) for a more procedural representation of a problem.
Original language | English |
---|---|
Publisher | University of Edinburgh Department of Machine Intelligence and Perception |
Number of pages | 56 |
Publication status | Published - Dec 1974 |
Publication series
Name | Series: MIP-R- |
---|---|
No. | 109 |