Wouter Kool

Hi! Thanks for notifying. I looked into the issue but am unable to find a way to make the NamedTuple with overriding `__getitem__` work in both Python 3.7 and 3.8....

Hi! Thanks for the interest in our paper! This notebook is mainly for illustrative purposes on the algorithm, which explains why we used a very simple model which is not...

@N-Wouda currently I have kept both the original constructor and a new constructor with the new parameterization to avoid changing all the tests. I wanted to discuss this with you...

I've updated this to be in sync with main. @N-Wouda can you have another look?

> This works fine, but any change in the routes of U and V are now reason to re-evaluate. Even when that change happens many nodes after U or V,...

@leonlan and I just discussed this feature. Some notes: * Let's implement basic support first, we can optimize it further. * We will support multi-trip by adding 'depot' delimiters into...

> > * Representation of a multi-trip solution. How do we include trip delimiters? I think the intermediate depots should be made explicit e.g. `Route #1: 1 2 0 3...

Hi @npadon, @N-Wouda, @joopvdh has made a basic implementation in https://github.com/ortec/PyVRP/tree/multi-trip. It technically supports multi-trip, but has some limitations (e.g. crossover does not preserve multi-trip) and algorithmic/implementation performance is not...

Note that there are multiple levels at which we could do a precheck based on bounds. For example if we only want to defer fullly computing `twBetween`, we may even...

I think there are also some situations where we can skip evaluating the `TWS::merge` since we can reason that the timewarp will be 0 (e.g. when removing a customer from...