1. SDS,  RAN,  CYC, and FCFS.

2. SDM,  SFTAO,  and FWJM.

3. SPT,  FCFS,  and SRPT.

4. LST,  SPT,  and CGO.

5. SPT,  SRPT,  and SRCS.

6. WFFS,  WLAS,  and WFLUS.

The scheduling model was implemented in a SIMAN simul- ation program and the dispatching rules were coded in FORTRAN. Seven performance measures (flow-time, total completed parts, number of tardy jobs, average WIP, maximum number of jig/fixture, machine utilisation, and AGV utilisation), under three separate algorithms were employed to evaluate the scheduling rules. Algorithm 1 used a look-ahead approach, algorithm 2 used the rules found to be effective from previous studies [35], and algorithm 3 used simple rules at each decision point. Tang et al. found that the look-ahead approach outperfor- med the other approaches for flow-time and number of tardy parts. They did not conclude which combination of scheduling rules was the best one    overall.

Goyal et al. [27] investigated the effect of various scheduling rules applied to two decision points, i.e. selection of jobs to enter to the system (loading), and selection of jobs to be processed by the machine (dispatching). Seven scheduling rules and four performance measures (average workstation utilisation, average buffer utilisation, average throughput, and average lateness) were employed to determine the most effective pro- duction schedule for an FMS. However, results indicated  that the best combination of rules for  each  performance measure was different.

Selladurai et al. [75] developed, in the “C” language, a visual interactive computer graphics simulation software for analysing dynamic scheduling of an FMS. The FMS consisted of a load/unload station and eight machines with equal capacity. A local buffer was presented in front of each machining centre. Dispatching rules being analysed were SPT, LPT, EDD, SIO, SLACK, and SLACK/LRO. Performances of each rule on average flow-time, average tardiness, and number of late jobs were reported, but only the average flow-time was  considered as the major performance criterion, and the authors concluded that SIO was an optimal scheduling rule for the system.

Caprihan and Wadhwa [76] studied the impact of the routeing flexibility of an FMS. The FMS, which was modelled by SIMAN IV simulation language with user written C code, consisted of six machines, one load station, and one unload station. Four factors, which were routeing flexibility, number of pallets, dispatching rule, and sequencing rule, were considered. Makespan was the only performance measure employed. They used Taguchi’s experimental design framework to gain quick

insights into the behaviour of the four factors within the FMS environments. However, they concluded that routeing flexibility should not be taken for granted as a  direction  for  perform- ance improvement.

Holthhaus and Ziegler [77] presented a new coordination scheduling approach, which was analysed using a simulation model consisting of 256 machines and only one decision point. Four local dispatching rules, i.e. FIFO, SPT, MOD, and S/OPN, and two global rules, i.e. NINQ and COVERT,  were  used  in the model. In the new approach, a look-ahead policy, namely look-ahead job demanding (LAJD), was used to schedule the jobs of each machine, say machine m, based on look-ahead information introduced by Itoh et al.  [78].  LAJD  considered not only the state  of  a  given machine,  but  also  the states of all machines which preceded machine m. Whenever there was the risk of running idle for one or  more  of  the  machines, LAJD was activated in conjunction with one of the above- mentioned traditional rules. For a specific machine, the risk of running idle was measured by the run-out time of the current total-work-content (i.e. the remaining processing time of the waiting-for-processing jobs and the currently processing job). LAJD was not activated when the  work-content  was  high. They tested the developed procedure on five performance measures, that is, mean flow-time, maximum flow-time, per- centage of tardy jobs, mean tardiness, and maximum tardiness. Although the results showed that the look-ahead policy outper- formed all traditional rules,  there  was  no  single combination of LAJD and a traditional rule outperforming all other combi- nations. They did not conclude which  combination  was  the best  one overall.