|Top Previous Next|
We have collected some tips / general advice:
•How long time does a calculation take? Key factors in increased processing time are total number of jobs and especially number of jobs assigned to each resource period. You should try to keep number of jobs per resource period below 100. Total number of jobs can not exceed 7500 and depending upon other settings, the limit may be lower. For tasks that big you may be looking at run times of couple of days. Consider aggregating jobs in advance if possible or split the task into smaller geographical areas. Other factors such as geographical spread, time windows, "tightness" of jobs etc. also play a major role.
•When should a calculation be stopped? From an optimization point of view when there has been "several" iterations without improvements. Use the StopRatio property in the XML to aid in defining this.
•As we make changes to FleetEngine (improving it / adding features etc) it may also happen that future results for the same input gets worse. This is inevitable in a system, which relies on randomized heuristics.
•Deciding if FleetEngine can handle a specific optimization task can sometimes be tricky, since the requirements to a solution may not be known in detail. Don't underestimate the resources to investigate this part !
More technical advice
•Use as few job time windows as possible. If the only resource period is from 8:00 to 11:00, there is no need to specify that the job should be executed between 7:00 and 12:00. It only makes the calculations slower. Generally use as few restrictions as possible to describe the problem.
•We recommend starting with not so many jobs (if you have many), since that will make it faster to detect if a task need to be formulated differently or a restriction is missing.
•Cost may increase (!) during an optimization process, if previously unassigned jobs get assigned. This is not an error.