Smaller isochrones look fine, but larger ones (>50 minutes or so) cause a lot of errors where a hole does not fit in a shell, and the resulting isoline contains very long incorrect straight segments.
This may be due to a difference in method with Laurent's original isochrone code. We're throwing only the intersection vertices into the grid, and we're not walking along long edges to get points out in the middle of them.
Smaller isochrones look fine, but larger ones (>50 minutes or so) cause a lot of errors where a hole does not fit in a shell, and the resulting isoline contains very long incorrect straight segments.
This may be due to a difference in method with Laurent's original isochrone code. We're throwing only the intersection vertices into the grid, and we're not walking along long edges to get points out in the middle of them.
See https://github.com/opentripplanner/OpenTripPlanner/blob/master/src/main/java/org/opentripplanner/routing/spt/SPTWalker.java