Antonio Garrido, César Guzmán, Eva Onaindia
Universidad Politècnica de València,
This paper appears in: 28th Workshop of the UK Planning and Scheduling Special Interest Group, PLANSIG. Brescia, Italy.
Issue Date: Dec. 1-3, 2010
Abstract - download article
The execution of a plan in a highly dynamic real-world environment entails facing unexpected events that produce discrepancies between the observed and the predicted state. In a situation like this, we need to fix the flaws, and typically we have two possible options: replan from scratch or repair, i.e. adapt the original plan to the new context. This paper presents an effective method to support the decision making between repairing or replanning. Particularly, we have designed a method that estimates the cost of a bridge from the observed state to any reachable goal state in the original plan. We have also endowed this adaptation method with an anytime behaviour to improve the plan quality in terms of both problem metric and stability. The paper also presents some experimental results to evaluate the approach.