How long does it take for Patchworks to reach a solution?

The scheduler is started when the "ON" button in Patchworks is pressed. At this point the scheduler begins the process of selecting treatments and building up the harvest schedule. The process has started, but how does one know when to hit the off button; when is the simulation finished?

As the scheduler executes, it continues to test various treatment timing combinations and continues to search for improvements to the objectives function. Initially, the changes made by the scheduler will yield large gains to the overall state of the solution. As the simulation proceeds and the solution develops the amount of improvement over a given period of time will begin to diminish. At first the scheduler will choose treatments and timings that fill in major gaps in the most important targets, but as these are satisfied the scheduler will then begin to fine-tune the schedule to meet secondary and tertiary objectives. Adjustments to the treatment schedule will be made to balance the values of targets between periods or targets (for example, by swapping the timings of two similar treatments), perhaps only achieving a minuscule decrease in the objective function value.

Eventually the magnitude of these improvements diminishes to being almost nonexistent. Theoretically, the value of the solution will asymptotically converge towards an optimal value for a given set of targets (see Figure 25, “Convergence of the Objective Function”). Although in many situations the scheduler can continue to run almost indefinitely and still yield minor improvements, the value of these improvements will likely be less than the error inherent in the model, and less than the cost-benefit of the increased solution time.

At some point a decision must be made to halt the scheduler and examine the results. The recommended approach in Patchworks is to set a threshold value for for improvement to the objective function over a given amount of effort (i.e. less than 0.1% decrease in value over 400,000 iterations). This type of condition can be set using API commands and the Patchworks Console. In a typical session, the condition is set, the scheduler is started and will run until the condition is met, at which point the scheduler will pause and save reports.

The time to convergence will vary depending on the size of the dataset and the complexity of the targets, but takes only a few minutes for the sample dataset.

Figure 25. Convergence of the Objective Function

When improvement in the objective function value is minimized, Patchworks has reached a stable solution and the simulation is finished.