This is an interesting overview.

Re auto routing: one of the most important tradeoffs is speed at run-time vs. flexibility to bring in new data, such as "live" traffic speeds.

Re public transit routing: RAPTOR (and its variants) is the most commonly used routing algorithm these days.

Even though this paper says OpenTripPlanner uses RAPTOR, it did not in 2015. However, the more recent OpenTripPlanner 2.x version does now use RAPTOR: https://github.com/opentripplanner/OpenTripPlanner/releases/...

The Transitland Routing API also uses a variant of RAPTOR to provide a hosted transit routing API across the entire United States: https://www.transit.land/documentation/routing-api/

The author list on this paper is kind of the routing equivalent of seeing a song with writing credits to Bob Dylan, John Lennon, Jimmy Page, Thom Yorke and Prince.

Transfer Patterns. This was a fun paper to implement back in the day. I had a local route planner that I wrote in nodejs, and it was always fun trying out research papers.

I wonder what Google, Microsoft, Mapbox et al use these days.