EXCALIBUR
Adaptive Constraint-Based Agents in Artificial Environments

[SYMBOLIC SRC]   [Structures]   [Selection]   [SRC-H1]   [SRC-H2]   [SRC-H3]   [SRC-H4]   [SRC-H5]

[ Please note: The project has been discontinued as of May 31, 2005 and is superseded by the projects of the ii Labs. There won't be further updates to these pages. ]

Selecting an Improvement Heuristic

(Related publication: [PUBLink])

To select one of the constraint's improvement heuristics, the general applicability of the heuristics is first checked (see following sections for applicability conditions). Only applicable heuristics are considered for the following choice process.

The choice of a heuristic is made in exactly the same way as the choice of a modification alternative in the Orc Quest example. The chosen heuristic is applied.

An applied heuristic can return a negative success value, which means that the heuristic was unsuccessful and no change has been made. In this case, the choice process is restarted. This restart means that the usual update of the preference values is skipped. Instead, the failed heuristic's preference value is temporarily divided by two (though no value may fall below one) because the probability that the heuristic will return a negative success value again is higher than before. If one of the improvement heuristics has been successfully applied, all adaptations of the preference values that were done for the restarts are undone.


[SYMBOLIC SRC]   [Structures]   [Selection]   [SRC-H1]   [SRC-H2]   [SRC-H3]   [SRC-H4]   [SRC-H5]

For questions, comments or suggestions, please contact us.

Last update:
May 20, 2001 by Alexander Nareyek