Research > Scheduling > Argos Research

ARGOS - Scheduling Optimization System

PROBLEM STATEMENT
Existing scheduling systems, such as ARTEMIS, SAP, or PRIMAVERA, rely on “makespan minimization” techniques to develop schedules. Specifically, they try to schedule tasks as early as possible, subject to constraints and resource availability profiles. Manual intervention is typically required to deal with overloaded resources and other difficulties, and the process of scheduling a single project can take months. This approach relies on the conventional wisdom that it can never be wrong to get work done early, and the assumption that a short schedule is likely to be efficient, since otherwise the inefficiently utilized resources could be loaded up to get more done earlier.

On Time Systems’ (OTS) research has shown that this conventional wisdom is, in fact, misleading. In mass production environments, makespan minimization often is a useful approach, since other jobs can help smooth out the resource loading artifacts the process induces. In other industries, where it is not uncommon to have one, or at most a few, projects in process at any one time, makespan turns out to be a poor stand-in for more complex goals.

OTS has developed a radically new approach to ship construction scheduling, one that addresses shipbuilding’s unique needs directly. The resulting system, ARGOS, is capable of scheduling hundreds of thousands of tasks in a single day, instead of months, without need for human intervention. The resulting schedules typically exhibit a 10–20% reduction in construction labor costs when compared with those in use today. Conversely, in situations where throughput is limited by the available manpower pool, ARGOS makes it possible to progress 10 to 20% more work. All of these savings are achieved without changing the fundamental production process in any way.

Strengths and Weaknesses of Existing (Baseline) Scheduling Technology

Traditional schedulers, as mentioned above, mix human and machine efforts to try to find a legal (meets all constraints) schedule that fits within available resource profiles. Essentially, the process involves something like the following: Tasks are aligned with their earliest allowable start dates, then, starting with Day 1, tasks are assigned in order of their priorities, until no further resources are available. Tasks not receiving resources are postponed one day, and the process repeats. When tasks slip to the end of their float, they are scheduled as overload at the place within their feasible window that results in minimal overload (so far).

Two features are immediately obvious from this description. First, tasks are placed early in the schedule because they can be, not because it makes intrinsic sense. Second, the assignment of priorities to tasks––which must be done manually––crucially affects the resulting schedule. However, there is no reason to believe that it always makes sense to do important tasks first, and it is easy to construct examples where the opposite is true. Existing approaches take so long, are so labor intensive, and produce schedules so far from optimum, precisely because of the need to manually discern prioritizations that force the scheduler to avoid bad placements. In the final analysis, however, even if humans could understand the ramifications of the changes they make to such complicated processes, prioritization is an imperfect substitute for actually addressing the real goals of the company and scheduling accordingly.

Existing commercially available schedulers are essentially makespan minimizers: they try to get the necessary work done as quickly as possible. In many domains––such as assembly-line production––some form of makespan minimization is the most effective solution technology known. The fact that “there will be another one along any minute” allows the scheduler to fill in gaps in resource utilization by interleaving two or more units of production. Many industries, however, are not mass production operations; existing schedulers—designed as they are for a completely different environment—do not serve them well.

Comparison of Baseline with ARGOS Technology for Shipbuilding

ARGOS represents a completely new approach to scheduling that OTS devised when our collaboration with shipyards began to reveal the fundamental flaws of applying existing technologies (even the state-of-the-art algorithms we had developed) to shipyards. Designed from the ground up to address the peculiar needs of shipbuilding, ARGOS is capable of keeping the multiple (and sometimes conflicting) criteria under which yards operate, explicitly in focus, addressing them directly in its algorithms. These criteria can include makespan, but they may also include inventory, work in progress, level resource usage, and many others. We have demonstrated that this approach has dramatic consequences in terms of the efficiency of the resulting schedules.

The following table shows the results obtained by ARGOS scheduling the construction of a single Virginia Class submarine. The relevant baseline costs, using the yard’s existing scheduling system, are approximately $155M; it took the yard roughly 6 weeks to produce that schedule using existing techniques.

Iteration Time Savings 1 2 min 8.4% $13.0M 7 10 min 11.4% $17.7M 20 34 min 11.8% $18.2M Ultimate ~24 hrs 15.5% $24.0M

ARGOS is iterative, producing good schedules very quickly and then refining them as time allows. The process typically reaches a point of diminishing returns after a few iterations and can be interrupted with a useful result if time is of the essence. In this case, the system was able to reduce labor costs by $24M within a day; in half an hour, it could save 75% of that amount.

ARGOS is capable of handling much larger problems without significant degradation in its speed or solution quality. The following table, which represents production for a whole yard over a five-year period, demonstrates this. Here, the relevant baseline cost for comparison is approximately $630M.

Here, the savings are reduced somewhat by the fact that the production of multiple boats over the five years did provide some mitigating effects in the baseline, which was able to compensate for some inefficiencies in the production of single boats through overlaps.

Our estimates are that, if ARGOS were applied to all new Navy construction, annual savings could be expected to be approximately $500M. Numbers for refit and repair are more difficult to obtain, but the percentage savings (10–20%) appear to be comparable.