Routing Failure

[tracmark]

Nick - I had a routing failure at TRB. I thought it was a problem with the algorithm, but I have come to believe it is actually a data problem, combined with a UI problem. I tried to make a route (See initial Route Failure)

The "Error: No route" message flashes at the bottom of my screen (I'm using Google Chrome) but disappears quickly. I did not see that message initially (so that is my UI issue) - maybe it needs to be made more visible and stay visible longer, and with a slightly more verbose error message. I was expecting to get a route. When I took off the curb ramp requirement I get a route (see Routing without Ramp Requirement)

Putting back the ramp requirement flashes the error code, but does not remove the route that does not require the curb ramps. The route stays showing on the map. If you miss the flashed error message, you think the routing algorithm is the issue.

Playing with the inclines doesn't give me a "ramp required" route. The issue appears to be that the underlying data does not realize that curb ramps are present. For example, if you look at the "Initial Route Failure" image, you will see that most of the intersections are impassible if that requirement is present. A good example is Pike & Broadway. But a Streetview image of that intersection shows that ramps are present on all four corners. (Ramps at Pike and Broadway)

Without having access to the actual data, I can't tell whether it is AccessMap that can't see the curb ramps because of some type of coding error, or whether they are simply not in the database. I do notice that the sidewalks on a number of streets are also missing from the map between Union and Pike. For example, see Belmont and Boylston (on the lower left side Routing without Ramp Requirement image.) Those streets have sidewalks. (See Sidewalks on Belmont between Union and Pike.)

If I choose really carefully for places where AccessMap thinks there are curb ramps, the routing algorithm works. (see Successful ramp routing.)

So - The issue I thought I had (being unable to route when requiring curb ramps for a route I expected to be successful in my mind) fails either because of missing/incorrect data, because the the curb ramps are incorrectly coded, not because the routing algorithm is having an issue. At least I think that data quality/completeness is the real issue. But we do need a slightly better way of telling people that given the personalized choice, no route is feasible.

Let me know if you have questions.

Mark