The shape of the data you feed a routing algorithm matters as much as the algorithm you choose.
Every map you have scrolled through is a compromise. The road data behind it holds far more detail than the screen can show at that zoom level, and something has to decide which points survive.
The same decision sits inside routing. A search algorithm reads the graph it is handed, so an oversized graph means slow answers. That is the problem a 2023 study on navigation efficiency set out to address. Finding the shortest path is easy. Finding it in a wasteful graph is not.
The Douglas-Peucker algorithm has thinned out coastlines and borders in cartography for decades. Applied to transit routes it does the same job: keep the points that define the shape of a journey, drop the ones that only add noise.
There is a trade-off to respect. Simplify too aggressively and a route line can cut a corner it should not, which matters when that line is what the user follows on the map. The useful property of Douglas-Peucker is that it is parameterised. You choose how much deviation you are willing to accept.
Muhammad Yasir Anshari Haq, Riswan Septriayadi Sianturi, and Aryo Pinandito (2023). Enhancing the Efficiency of Navigation Systems: Dijkstra Algorithm Optimization Through Integration of Douglas-Peucker Line Simplification Technique. Proceedings of the 8th International Conference on Sustainable Information Engineering and Technology.
That is the real lesson for anyone building location features. Performance work does not have to mean a faster algorithm or a bigger server. Sometimes it means looking at the shape of the data you are feeding in, and deciding that half of it was never load bearing.
