Approximately $100 billion is spent annually on the construction, refit and repair of commercial and naval vessels. New commercial construction is performed at hundreds of shipyards worldwide for commercial shipping companies that carried about 6.8 billion tons of cargo by sea last year (over one ton for every person on the planet).

The problem of developing schedules that make efficient use of shipyard resources is critical, both to control costs and to meet needs regarding available resources, be they time, dry dock space, or skilled manpower. Both construction and maintenance are extremely complex processes, involving thousands of activities with thousands of constraints among them. This complexity makes it virtually impossible for people, even augmented by the commercial scheduling systems available today, to find good schedules.

If the goal in scheduling is to manage resource usage, the most effective way to achieve that goal is for the scheduling system being used to target resource management directly. Experiments at several US shipyards indicate that doing this can save over 10% of total labor costs, and ARGOS is unique in providing this functionality. These savings are achieved without changing the fundamental production process or shipyard facility in any way.

Figure 1 shows a typical project from an actual shipyard. The top histogram shows the results of the ARGOS optimization, while the bottom histogram is the resource staffing profile of the original schedule. Not only is the ARGOS profile smoother and flatter, it also maintains a desirable ramp-up and ramp-down shape. Note that the original profile peaks at around 1900 workers on staff, while the ARGOS profile peaks at around 1500; both schedules start and end at the same time.

The black line in the figure shows the resource level (i.e., the total workforce size) on site for this project. The red areas indicate overtime; any difference between the top of the green area and the black line indicates undertime (or idle time). Note the significant gaps between the resources on staff and those needed in the first half of the original schedule; those gaps correspond to significant idle time expense and have been eliminated in the ARGOS version of the schedule.

Figure 2 shows the same histograms, but now colored by how critical each task is. A task that cannot "slip" without delaying the entire project is colored red, while a task that can slip significantly is colored green. The user controls the way in which ARGOS trades off the optimization criteria of schedule cost and robustness, since these two desiderata often compete with one another. Thus the user can easily create a more (or less) time-critical schedule simply by adjusting a slider and running ARGOS again. In the figure, note that the ARGOS schedule is "greener" than the original schedule, with the critical red tasks at the middle and end of the original schedule having been moved to a time where there is potential slack in their execution. From an operational perspective, this means that the ARGOS schedule, in addition to being cheaper to execute, is more likely to be executed successfully.

Figure 3 displays a final perspective on the same project, and shows that the ARGOS schedule is over 10% cheaper than the original. Further detail provided by the ARGOS GUI is not shown here, but includes a variety of other perspectives on similar information and the recommended start date and time for each of the activities that ARGOS moved in order to achieve its results. These other features were designed after extensive consultations with schedulers at a variety of American yards.