Automated Task Rescheduling using Relational Markov Decision Processes with Logical State Abstractions

• Autores: Palombarini, Jorge; Martínez, Ernesto
• Año de publicación: 2012
• Idioma: inglés
• Tipo de recurso: documento de conferencia
• Materia: Ciencias Informáticas; Rescheduling; Relational Markov Decision Process; Manufacturing Systems; Reinforcement Learning; Abstract States
• Institución: Universidad Nacional de La Plata

Generating and representing knowledge about heuristics for repair-based scheduling is a key issue in any rescheduling strategy to deal with unforeseen events and disturbances. Resorting to a feature-based representation of schedule states is very inefficient and generalization to unseen states is highly unreliable whereas the acquired knowledge is difficult to transfer to similar scheduling domains. In contrast, first-order relational representations enable the exploitation of the existence of domain objects and relations over these objects, and enable the use of quantification over objectives (goals), action effects and properties of states. In this work, a novel approach which formalizes the rescheduling problem as a Relational Markov Decision Process integrating first-order (deictic) representations of (abstract) schedule states is presented. The proposed approach is implemented in a real-time rescheduling prototype, allowing an interactive scheduling strategy that may handle different repair goals and disruption scenarios. The industrial case study vividly shows how relational abstractions provide compact repair policies with less computational efforts.
Sociedad Argentina de Informática e Investigación Operativa