Machine Learning of Plan Robustness Knowledge About Instances

Sergio Jiménez Celorrio, Fernando Fernández, Daniel Borrajo

Abstract: Classical planning domain representations assume all the objects from one type are exactly the same. But when solving problems in the real world systems, the execution of a plan that theoretically solves a problem, can fail because of not properly capturing the special features of an object in the initial representation. We propose to capture this uncertainty about the world with an architecture that integrates planning, execution and learning. In this paper, we describe the PELA system (Planning-Execution-Learning Architecture). This system generates plans, executes those plans in the real world, and automatically acquires knowledge about the behaviour of the objects to strengthen the execution processes in the future.