Scheduling with neural networks - the case of the hubble space telescope

Scope and Purpose: This paper shows how artificial neural networks can be seamlessly merged with unique evidential reasoning techniques to form a flexible and convenient framework for rep­ resenting and solving complex scheduling problems with many "hard" and "soft" consttaints. The techniques described form the core of SPIKF.. an operational scheduling environment for long­ range scheduling of astronomical observations with the orbiting NASA/ESA Hubble Space Tele­ scope. The methodology of SPIKE, which is currently being adapted to some other space mis­ sions, is fairly general and could be brought 10 bear on a wide range of practical scheduling prob­ lems. Abstract: Creating an optimum long-term schedule for the Hubble Space Telescope is difficult by almost any standard due to the large number of activities, many relative and absolute time con­ sttaints, prevailing uncertainties and an unusually wide range of timescales. This problem has mo­ tivated research in neural networks for scheduling. The novel concept of continuous suitability functions defmed over a continuous time domain has been developed 10 represent soft temporal re­ lationships between activities. All consttaints and preferences are automatically translated into the weights of an appropriately designed artificial neural network. The consttaints are subject to prop­ agation and consistency enhancement in order 10 increase the number of explicitly represented consttaints. Equipped with a novel stochastic neuron update rule, the resulting ODS-network, ef­ fectively implements a Las Vegas-type algorithm to generate good schedules with an unparalleled efficiency. When provided with feedback· from execution the network allows dyrumic schedule revision and repair. Keywords: artificial neural networks-combinatorial optimization-constraint satisfaction problems-evidential/uncertainty reasoning-graph problems: maximum independent set, min­ imum vertex cover-hemistic search-nonmonotonic reasoning-scheduling-stochastic al­ gorithms .