[ Please note: The project has been discontinued as of May 31, 2005 and is superseded by the projects of the ii Labs.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. ]

(Related publications: [PUBLink] [PUBLink])

You call the humans to account for the behavior of the planning system. They reply that most resource-based planning systems consider only a limited number of actions, increasing this number only if they are unable to produce a correct plan, the primary optimization goal being plan length. If it is possible to produce a correct plan with a certain number of actions, optimization of other goal properties can begin and plans with a larger number of actions are no longer considered. For your planning problem, the plan catch_a_group & catch_a_group is the only correct plan with a minimal number of two actions, and thus also the optimum with respect to any secondary optimization criterion.

Taking the number of actions or the plan length as the primary optimization criterion seems a very academic approach to you and you wonder if it has any relevance. You notice that the humans are becoming increasingly uncomfortable when they start arguing that optimization of the plan's duration is what is needed in most cases. And the duration would be the same as the number of actions. You no longer trust these humans and demand that they draw a picture of the complete search space. The result does not improve the humans' situation because the plan with the shortest duration is different from the one involving the minimal number of actions (see figure below).

Luckily for the group of humans, your goal is a plan involving the minimal amount of pain, and you reluctantly release the humans as promised because the plan involving the minimal amount of pain does not entail catching a whole group of humans. Somehow you feel dissatisfied at not having broken your promises.

For questions, comments or suggestions, please contact us.

Last update:
May 19, 2001 by Alexander Nareyek