Is LCSS a good option for loopy routes?

[pricealexandra]

I've been playing around with the mappymatch LCSS matcher. I love how quick it is and the examples are really fantastic. However, when I try uploading traces that aren't particularly straightforward and have a lot of loops or out and backs, it seems very hard to find combinations of parameters that match the trace accurately. My general goal is to "snap" the points in the original trace to the road network (they are usually already pretty close to accurate), but not to simplify the path at all -- I'd still like to preserve loops around blocks and out-and-backs. I'm trying to evaluate if this is user error on my part or if this library/algorithm is just not a great fit for these types of routes. Would love to hear if others have tackled similar problems with success!

[nreinicke]

This is a great question and apologies that I missed this! In our own experience, the LCSS matcher was designed for and validated on high resolution GPS traces collected from vehicles and in those data we don't typically encounter traces with a lot of out-and-backs and loops. So, matching on these kinds of traces is a bit of an uncharted territory for that algorithm and based on your initial results it might not be an ideal fit. We're starting to look at the idea of expanding this package to better support more use cases like mode-specific map matching and perhaps that work will unlock matching traces of this style.

[pricealexandra]

Awesome, thank you for the explanation and totally understandable that this example is pretty far outside the normal use case.