Do you know if there’s a way to improve the routing part? I find it often takes really obtuse paths that go out of the way and add time. It shows the more optimal roads as existing, so I’m not sure why. Maybe missing speed limits or something that make it think worse roads are more optimal?
OsmAnd? I’ve found it’s often worse than GMaps for driving but ends up strictly better for walking, cycling, or just telling me where I am right now. OsmAnd for me often goes out of the way in weird ways by car, so right now I use OsmAnd for cycling, walking, and orientation where it’s quite superior (and where I’m less familiar with good routes) and then use it just to tell me where a destination is if I’m driving and figure the route myself. This route-self-figuring isn’t practical for everyone (longer routes, more complicated infra, just not that good with directions), and for that I either recommend something other than OsmAnd like GMaps (GMaps seems to give basically optimal routes for driving; it just falls flat on its face for walking and cycling) or, if you’re feeling like having fun about it and have time to kill, looking along the route to see if there’s missing info (either along the route OSM chooses that would negatively impact it or along the route you would choose that would positively impact it). The level of data granularity on OSM is kind of insane, so I would keep it limited to things that could plausibly impact travel times.
Something to keep in mind is that OSM at its core is the database itself. The renderer, navigation, visual editor, etc. are just plug-and-play – anyone could use the underlying dataset to try their hand at it. For example, a slightly suboptimal route I was talking about earlier happens on OsmAnd but not on the OSM website proper, which gives a perfect route between the points. Thus, I don’t let OsmAnd’s driving bugs dampen my enthusiasm for the project, since OSM stores much more information than GMaps and is thus perfectly capable of enabling routing software to make great choices. I’ve also seen OsmAnd continually improve over my time using it, where now it takes fewer questionable routes (likely both the data getting filled in and the routing algorithm being improved). This is definitely speaking from a position of privilege, though, since I’m not e.g. in a new city where I’m still learning where everything is; I’m instead in a position to take missing info as it comes and fill in the gaps.
I will note that sometimes it really is because some data is missing. For instance, I performed a stress test by creating a bicycling route diagonally from one end of town to the other. Adding data about traffic signals like stop signs meaningfully changed the route and seemed to improve it. With the crazy level of navigation-relevant detail that OSM stores (this also includes highly granular accessibility info), I would be surprised if a routing app hasn’t outpaced GMaps for literally everything in the next 5–10 years.
Do you know if there’s a way to improve the routing part? I find it often takes really obtuse paths that go out of the way and add time. It shows the more optimal roads as existing, so I’m not sure why. Maybe missing speed limits or something that make it think worse roads are more optimal?
OsmAnd? I’ve found it’s often worse than GMaps for driving but ends up strictly better for walking, cycling, or just telling me where I am right now. OsmAnd for me often goes out of the way in weird ways by car, so right now I use OsmAnd for cycling, walking, and orientation where it’s quite superior (and where I’m less familiar with good routes) and then use it just to tell me where a destination is if I’m driving and figure the route myself. This route-self-figuring isn’t practical for everyone (longer routes, more complicated infra, just not that good with directions), and for that I either recommend something other than OsmAnd like GMaps (GMaps seems to give basically optimal routes for driving; it just falls flat on its face for walking and cycling) or, if you’re feeling like having fun about it and have time to kill, looking along the route to see if there’s missing info (either along the route OSM chooses that would negatively impact it or along the route you would choose that would positively impact it). The level of data granularity on OSM is kind of insane, so I would keep it limited to things that could plausibly impact travel times.
Something to keep in mind is that OSM at its core is the database itself. The renderer, navigation, visual editor, etc. are just plug-and-play – anyone could use the underlying dataset to try their hand at it. For example, a slightly suboptimal route I was talking about earlier happens on OsmAnd but not on the OSM website proper, which gives a perfect route between the points. Thus, I don’t let OsmAnd’s driving bugs dampen my enthusiasm for the project, since OSM stores much more information than GMaps and is thus perfectly capable of enabling routing software to make great choices. I’ve also seen OsmAnd continually improve over my time using it, where now it takes fewer questionable routes (likely both the data getting filled in and the routing algorithm being improved). This is definitely speaking from a position of privilege, though, since I’m not e.g. in a new city where I’m still learning where everything is; I’m instead in a position to take missing info as it comes and fill in the gaps.
I will note that sometimes it really is because some data is missing. For instance, I performed a stress test by creating a bicycling route diagonally from one end of town to the other. Adding data about traffic signals like stop signs meaningfully changed the route and seemed to improve it. With the crazy level of navigation-relevant detail that OSM stores (this also includes highly granular accessibility info), I would be surprised if a routing app hasn’t outpaced GMaps for literally everything in the next 5–10 years.