Does the ShortestPathFinder work out an optimal route considering other start points, or is each route generated independent of others?

Each From-To line you put in the ShortestPathFinder is considered separately, but you can have it find an optimal order for the intermediate points within a single line (the Travelling Salesman problem).

However... depending on your network cost and the transformer settings, I would expect it to merge routes anyway if that is the most optimal routing. I think your wish to get the lowest overall length kinda contradicts your wish to merge routes.

Thanks for the response. I would like to find a way for the shortest path finder to generate routes so that there is a lowest total length (where joining onto another route adds no extra distance onto the cumulative count) - which doesn't seem possible if the ShortestPathFinder calculates routes in isolation of other routes

I think you may need to illustrate what you mean. A route is either the shortest route or it isn't.

You can choose to calculate using a weighting instead of an absolute shortest route, but I can't really picture exactly what you are after.

Please forgive the dreadful paint sketch, but I think it helps explain what I'm trying to do. The red points are start points, the black triangle is the destination. The road network is visible in the background.

The black line is the optimal "network" I want to achieve for these points. The orange lines show the routes of shortest distance between the start points and the destination for a few points which would produce a different path to the black. The shortestpathfinder would output the orange paths for these points.

So as you can see the start points on the far left would produce a smaller "network length" if they followed the black line, but would have a shorter path if they followed the orange line.

The start point with blue arrows has a roughly equal distance to the destination if it goes left along the black or right along the orange. In the circumstance like this I would see a more optimal route being the black because this would result in the lower overall network length and would also open up a shorter overall network route for the start point to its right (which would normally go along the orange path as it is a shorter distance).

So what I'm getting at is I want a way for the routes to be calculated so they produce an optimal network length rather than each route having the shortest possible length

Right, if you connect all the start points and the end point to a single line and tell the ShortestPathFinder to change the order of the points if necessary you will get something that is close, but not quite, what you’re looking for.

On the other hand, of you calculate all shortest paths and then run them through a stream order calculation you might get closer (can’t test that right now, no FME at home)

Thanks for your help, I'll give this a try :)

Hi @em_fme From your paint sketch it looks like you might want to try the RMinimumSpanningTree transformer. See https://www.safe.com/blog/2017/09/minimum-spanning-tree-evangelist167/ and https://hub.safe.com/publishers/lizsanderson/transformers/rminimumspanningtree

Thanks, this looks perfect

Can this be constrained to a network? Looks like it connects points but not along network lines?