Rescheduling Capability Analysis and Evaluation of Experimental Results

Abdul Samad Shibghatullah 135 30052008 The following subsections present the findings derived from the two experiments. The discussion will be organised according to the evaluation criteria set out in Section 6.2 which are rescheduling capability and rescheduling speed.

6.3.1 Rescheduling Capability

In this research the rescheduling capability was measured based on the number of matches. A match means that a duty which has lost its drivercrew because of a UE finds a drivercrew that is available to take the duty. The matching process is done by CRSMAS. The perfect result is when 100 of the duties find a match. Table 6.2 shows the number of successful matches in the Single Event and Multiple Event experiments. The table shows that the matching rate for Single Event is 47.3 230 matched out of 486 events and for Multiple Events is 42.1 2805 matched out of 6664. The capability of CRSMAS in Single Event is better than in Multiple Events. Table 6.2: The Matched Results of the Two Experiments Single Event Multiple Events Total Number of Events 486 6664 Total Number of Matched 230 2805 Percentage of Matched 47.3 42.1 However, as a whole the matching percentage is still low. The reasons are because of the weaknesses and limitations of CRSMAS, crew schedules and the nature of the UE. The details are presented below: The Weaknesses and Limitations of CRSMAS In Multiple Events the matching rate is lower than Single Event experiment. One of the reasons for this is that CRSMAS did not support more than one round of rescheduling. This is due to the fact that in Multiple Events, many events happen at the same time; therefore it will take more time to do second or further rounds of rescheduling because the process is done manually. Automating the multiple rounds more than one round of rescheduling process would be useful to increase the matching rate. Another limitation of CRSMAS although not concerning the matching rate is that CRSMAS cannot simulate different types of events at the same time. In Multiple Events experiment many events are tested, but the events are the same. This limitation is Abdul Samad Shibghatullah 136 30052008 because in different events there are different rules. In MAS, agents can only negotiate with one set of rules at a time. This is one of the limitations of MAS. One way that may solve this problem is to run the CRSMAS in different windows for different events. The Limitations in Crew Schedules The experimental results of Chapters Four and Five show that the matching percentage of small schedules is much lower. Small schedules have very few duties which is why when UE occurred the probability of finding a match is difficult. However, this problem may be solved by not restricting the crews according to route. For example, if Crew C is not available then any available crew from the same garage is entitled to take C’s duty. But with one condition that is all crews in that garage know all the routes that are managed by the garage. In both experiments the matches were restricted only to the same crew schedules. In the Single Event experiment it was discovered that one important factor that influences the matching possibility is duty spreading. If duties are not spread then the possibility of matching them is lower. For example, the large schedule Appendix C has many duties, but because the spread of duties in an hour is not equal or distributed then the matches were lower than the medium schedule Appendix D. The reason why the schedules are different is because they are not both from the same bus company. Apparently, different companies have different policies in constructing their schedules. By making the duties equally spread or distributed in an hour it increases the possibility of finding a match. Based on the analysis of both experiments the period of relief time also influences the possibility of matching. If the relief time is long then the possibility of finding a match is higher because it can absorb the UE such as lateness and delay. For example, the large schedule Appendix C allocates just enough relief time in the schedule more or less than 45 minutes. As a result, the schedule cannot absorb the lateness event and not many crews are available for rescheduling. In contrast, the medium schedule Appendix D allocates long hours for relief, so the schedule can absorb the lateness event and many crews are available for rescheduling. Longer periods of relief time increase the possibility of finding a match. Abdul Samad Shibghatullah 137 30052008 The Nature of UE The longer the period of a UE, the more difficult it is to find a match. For example, the matching rate for a delay event 80 to 180 minutes - that is very long is very low compared to lateness or short period unavailability events. The reason is that during the long period there has to be more than one rescheduling and the distribution of duties varies from hour to hour. If, in all the hours, the number of duties are large then there is no problem in finding matches although it has to go through many rounds of rescheduling. One way to solve this problem may be to divide long periods into a few short periods and then do rescheduling to all the parts.

6.3.2 Rescheduling Speed