Documentation: Is there any documentation that shows handling real scenarios

[Abhijit-Stylenow]

In the real world, there are lot of conditions and constraints. To name some few:

1. Defined Holidays, defined maintenance etc
2. Time taken between two operations (for movement)
3. Multiple machines availability for same operation (say 5 drilling machines) - distribute load across machines
4. SetUp / SetDown time - some operations need a cleaning process which is dependent on material
5. Repeated operations - e.g. A - B - C - B - C - D - E - D. (here B & C operations are required twice)
6. Sequence of job orders - especially in paint shop, you cannot have Black followed by White.

These are some of the cases.

Can someone point to documentation where you can model and solve such cases?

[Pabloo22]

Hi, thanks for your question.

The library, including the documentation, is still in development. I plan to add more in-depth tutorials, and explaining how to handle those cases would be a great addition.

Currently, the library does not support natively these special cases. However, I believe handling them should be possible, especially when working with dispatching rules. In general, it requires extending current classes via inheritance or composition.

For example, to encode additional information in an operation, you can extend the class by sub-classing the original one:

class CoatingOperation(Operation):
    def __init__(self, machines, duration, color="white"):
        super().__init__(machines, duration)
        self.color = color

Another low-effort feature The Dispatcher class supports a ready_operations_filter function. You can create a custom function to filter out operations based on a custom condition:

def color_sequence_filter(dispatcher: Dispatcher, operations: list[Operation]):
    operations_to_keep = []
    for operation in operations:
        if not isinstance(operation, CoatingOperation):
            operations_to_keep.append(operation)
            continue
        # Assuming a single machine per operation
        machine_id = operation.machine_id
        last_scheduled_operation = dispatcher.schedule.schedule[machine_id][-1]
        if not last_scheduled_operation:
            operations_to_keep.append(operation)
            continue
        if last_scheduled_operation.operation.color != "black" or operation.color != "white":
            operations_to_keep.append(operation)

    return operations_to_keep

This does not support the points 1-4, though. For handling those cases, extending the Dispatcher class would be required with an updated logic.

For point 5, maybe it can be circumvented with no additional logic by creating a duplicated operation if the repeated operations always belong to the same job. If they belong to different jobs, then adding a new attribute that links the same operations and updating the solver's logic would be required again.

If you or any person reading this is interested in contributing to the library by adding support for more real-world constraints, I would be glad to meet up to discuss it in more detail.