Closed behrisch closed 6 years ago
actually:
- geometric center of district-edge from nodes (and radius)
- geometric center of district-edge to nodes (and radius)
to:
1495655423789687
save:
- for taz-source edges the geometric center of the to-nodes of all edges along with the radius
- for taz-sink edges the geometric center of the from nodes of all edges along with the raidus
the best center value would be the solution to the https://en.wikipedia.org/wiki/Smallest-circle_problem along with the radius of that circle.
The center of the bounding-box should be a sufficient approximation.
@namdre changed type from "enhancement" to "defect"
The A* heuristic uses getDistanceTo() which always returns 0 if the origin or destination is a district edge.
Instead, each district edge should record its geometric center and radius for computing a lower bound on distance.
to:
The A* heuristic uses getDistanceTo() which always returns 0 if the origin or destination is a district edge. This effectively turns the algorithm into a Dijkstra search.
Instead, each district edge should record its geometric center and radius for computing a lower bound on distance.
@behrisch commented:
Wouldn't bounding boxes instead of circles do as well? (although the distance calculation takes a little bit more effort)
@behrisch committed 46d67a0 (aka r24514): towards a better A* with taz, refs #3144
@behrisch committed 3e60b15 (aka r24519): test for A* with taz, refs #3144
The A* heuristic uses getDistanceTo() which always returns 0 if the origin or destination is a district edge. This effectively turns the algorithm into a Dijkstra search.
Instead, each district edge should record its geometric center and radius for computing a lower bound on distance.
Migrated from http://sumo.dlr.de/ticket/3144