As a planner, you want to make a schedule as optimal as possible. So it makes sense to choose a tool like 'the optimizer'. If you want to use it effectively, you must first create the conditions for this. If you choose to automate optimal planning, it is essential to define the plan rules and goal requirements very precisely. This sounds simple, but in practice it is often a lot more complicated. I'll explain how best to do it.
Let's use a common scenario as an example: imagine that a delivery job has a clear time window. A customer chooses to deliver a package or groceries between 4:00 and 8:00 pm. At first glance, this seems like a simple and clear rule.
Combinations quickly make scheduling more complex
The complexity increases quickly, however, when a shipper or logistics party has to handle or deliver not just one, but multiple package orders at the same time. In doing so, you
naturally want the fewest possible vans to be needed to handle the packages. How do you combine time constraints with the need for the efficient use of resources? That's where the real challenge begins!
Example: to use or not to use an extra van?
An example: You can deliver a certain job at 19:45, but need an extra van and driver to do so. Without that extra van, you won't arrive at the client's place until 20:15. According to strict rules, 20:15 is too late, but using an extra van involves significant costs.
Hard versus soft rule
From a human perspective, there is a quick compromise here: the hard rule of being on time turns into a soft rule. You do your best to be on time, but not at any cost. Of course, translating a soft rule into a mathematical model is not easy. It requires an algorithm to find a balance between the soft rule and the other applicable goals.
For that matter, theoretically there is an obvious solution to this balancing act. You can add penalty points for being late and adjust the goal-setting function. From an optimization standpoint, you obviously want to minimize not only the number of vans and
miles, but also the number of penalty points. This seems a simple and practical solution. It meets the client's requirements, is easy to implement and does not unnecessarily complicate the algorithm.
The challenge of multiple objectives
Then, despite the seemingly simple solution, there is a big catch. The challenge here is not a technical one. A goal-setting function with three objectives has emerged: minimum number of vans, minimum mileage and minimum penalty points. Someone, often the end user, must now determine how to balance these goals. And this is extremely difficult. How many minutes late versus 100 extra miles? How do extra vans compare to mileage savings? And is this trade-off the same every day?"
Algorithm seeks solution for the wrong problem
In many realistic scheduling problems, the number of goals quickly rises to ten or even
more. It then becomes almost impossible for users to determine the correct weighting factors. The result? The algorithm searches for the optimal solution to the wrong
problem.
The simple solution: keep it simple
So what is the right solution? The problem is not a technical one, so the solution doesn't necessarily have to be. Instead of working with soft rules and conflicting objectives within the algorithm, it is better to aim for simplicity. In our example, too late is too late. We do not compromise on that front, even if it means requiring an extra van to prevent a minor violation. This may not seem optimal, but it is. In complex scenarios with multiple
objectives, the model always remains an approximation of reality, not an exact
representation.
There is another reason why this is optimal: in a good planning application, the planner always remains ultimately responsible, not the optimization algorithm. Contrary to the algorithm, the human planner may well decide that being 15 minutes late is allowed, if this avoids the use of an extra van. The algorithm provides the optimal solution within a simplified model world. At the same time, the human planner adds his expertise and considerations, creating the truly optimal solution: the "More Optimal" solution.
Playing with both hard and soft rules
The game that the planner plays on the way to optimal planning can be done neither without hard nor soft rules. With the hard rules, you maintain the simplicity and predictability needed for efficient and reliable optimization. With the soft ones, you respond to other wishes. So the latter are not taboo either. Some planning problems simply require a trade-off between different priorities. The trick is to know when these exceptions are really necessary. It is that game, that balancing act, that we enjoy at More Optimal.