Sabuncuoglu and Hommertzheim [67,68] studied the prob- lems of scheduling machines and AGVs using an FMS simul- ation model in a single  criterion environment.  The influences of machine and AGV scheduling rules on the mean flow-time criterion were investigated. Machine scheduling rules were SPT.TOT,  SPT/TOT,  LPT.TOT,  LPT/TOT,  LWKR, MWKR,

FOPR, MOPR, FCFS, FAFS, and RANDOM. AGV scheduling rules were FCFS, LOQS, STD, LQS, MWKR, and FOPR. The FMS consisted of six machine centres with limited  input buffers, one inspection station, one washing centre, one loading, and one unloading station. Two AGVs were employed to transport  parts  through  the  system.  The  FMS  was  modelled

using the SIMAN  discrete  simulation  language  and animated in CINEMA. The scheduling rules were tested with  three factors, i.e. different utilisation levels, different  queue capacities, and different AGV speeds.  The  results  indicated that scheduling AGVs was as important as  scheduling machines. The due-date of parts was not considered in the simulation model, and consequently no due-date-based rule or measure of performance was employed. Sabuncuoglu [69] then extended the studies under new experimental conditions. The same FMS was employed, but  the  objective  was  to measure the sensitivity of the rules to changes in processing time distributions, various levels of breakdown rates, and types of AGV priority scheme. Although similar results were obtained to those of the previous work [67,68], Sabuncuoglu concluded that scheduling of material  handling  systems  is  as important as  the  machining subsystem.

O’Keefe and Kasirajan [4] investigated the interaction between nine dispatching rules and four next station selection rules in a relatively large dedicated FMS with a simulation model using the RENSAM package. The FMS was modelled with constant operation times and no machine breakdowns or AGV failures. The model contained 16 workstations with local buffers, nine load/unload stations, three AGVs, and six part types. Dispatching rules used in the model were  FIFO, SIO, LIO,  FRO,  MRO,  SIO/TOT,  LIO/TOT,  SLACK,  and   SIOx.

Next station selection rules consisted of  NS,  WINQ, NINQ, and LUS. The only performance measure was weighted flow- time. The smallest value of  weighted  flow-time  they  found was related to SIO/TOT combined with WINQ. The best next station selection rule was  WINQ  and  the  worst  was  LUS. The main shortcoming of this study was its single criterion environment. Although due-date had been considered and two due-date-based rules were applied, no due-date-based perform- ance  was measured.

Rohleder and Scudder [70] made a simulation model, with only one decision point, to investigate the influence of ten scheduling rules on the net present value (NPV) in a JIT production system. Scheduling rules were OPCR, ODD, OPSLK,  CR,  EDD,  MOD,  TSLK,  MDD,  SPT,  and  LWKR.

Performance measures were mean system inventory,  NPV, mean tardiness, percentage of tardy jobs, average starting time of operations, number of jobs in process  (WIP),  number  of jobs finished but not  shipped (NFGS),  and  average  total jobs in system (WIP NFGS). Due-date of jobs was assigned using TWK. The model was run with three due-date tightness (K 3, 6, and 9). The results showed  that  different  scheduling rules had different influences on performance measures. For example, job-based allowance rules dominated the mean system inven- tory.

Rachamadugu et al. [71] studied the influence of sequencing flexibility on the performance of rules used, to schedule oper- ations in manufacturing systems, using a simulation model consisting of ten machines. The performance of ten scheduling rules was examined including FIQ, FIS,  SPT,  EDD,  MDD, CR, EODD, MODD, OCR, and MSUC. The performance measures were mean flow-time, average tardiness, and pro- portion of tardy jobs. The results showed that the performances of all rules were improved, while levels of sequencing flexi- bility were increased. It was demonstrated that the  performance