A Neural Network for Solving Job Shop Scheduling Problem

Abstract

Job shop scheduling problem is a ressource allocation problem subject to satisfy precedence constraints and ressource constraints. This problem belongs to combinatorial optimization problems. It ranks among the most difficult known to mathematical community, since it has proved to belong to the class of NP hard problems. The most practical solution algorithms abondon the goal of finding the optimal solution, and instead attempt to find an approximate, useful solution in a reasonable amount of time. Many of these algorithms exploit the problem specific information and hence are less general. However simulated annealing algorithm for job shop scheduling is general and produces better results than heuristics. However its major drawback is that the execution time is high. One approach to reduce the execution time is to develop parallel and distributed models, then neural network are appropriate. Simulated annealing and neural network techniques can be combined to give a stochastic, parallel solution which is encoded in a specific neural network: Boltzman machine. The architecture of a Boltzman machine is similar to discrete Hopfield net and is usally taken to be an array of fully connected neurons. Part of the array is divided into input/output units while the rest is considered as hidden units. In this paper a neural model: Boltzman machine for solving job shop scheduling problem is presented.