Learning to Repair Plans and Schedules Using a Relational (Deictic) Representation

PALOMBARINI, JORGE; MARTINEZ, ERNESTO

Computer Aided Chemical Engineering

Elsevier

Lugar: Amsterdam; Año: 2009; p. 1377 - 1382

Unplanned and abnormal events may have a significant impact in the feasibility of plans and schedules which requires to repair them ‘on-the-fly’ to guarantee due date compliance of orders-in-progress and negotiating delivery conditions for new orders. In this work, a repair-based rescheduling approach based on the integration of intensive simulations with logical and relational reinforcement learning is proposed. Based on a deictic representation of schedule states a number of repair operators have been designed to guide the search for a goal state. The knowledge generated via simulation is encoded in a relational regressión tree for the Q-value function defining the utility of applying a given repair operator at a given schedule state. A prototype implementation is discussed using a representative example of 3 batch extruders processing orders for 4 different products. The learning curve for the problem of inserting a new order vividly illustrates the advantages of logical and relational learning in rescheduling.