Reactive thermostatistical search (RTS) is used within this paper to optimi
se parameter settings for discrete event simulation problems; the performan
ce of its implementation within the Witness discrete event simulation packa
ge is benchmarked against that of the optimiser used within the ProModel pa
ckage. RTS is a new technique, based upon simulated annealing, which incorp
orates elements of tabu search into the SA process, by learning from its ex
perience of the problem domain and modifying its search strategy accordingl
y. (C) 1999 Elsevier Science Ltd. All rights reserved.