View Issue Details
| ID | Project | Category | View Status | Date Submitted | Last Update |
|---|---|---|---|---|---|
| 0004911 | Spring engine | General | public | 2015-07-20 16:06 | 2016-06-09 16:47 |
| Reporter | lhog | Assigned To | Kloot | ||
| Priority | normal | Severity | major | Reproducibility | always |
| Status | resolved | Resolution | fixed | ||
| Product Version | 100.0 | ||||
| Summary | 0004911: pathfinding issues on IsIs Delta, probably on some others | ||||
| Description | Constructor consistently chooses wrong path to approach construction site. Goes in endless loop. | ||||
| Steps To Reproduce | 1) Watch the replay 2) Do same | ||||
| Tags | No tags attached. | ||||
| Attached Files | |||||
| Checked infolog.txt for Errors | |||||
|
|
There are still issues on 100.0+git I've uploaded a demo |
|
|
I assume you mean the issue at 3:20, which is a very old design hack: the set of nodes allowed to be explored is limited by a circular constraint to keep CPU use under control in pathological cases when start and goal are within max-resolution search distance as the crow flies, but separated by half the map in path-space. Fixing that under the current PFS design will be a bit hairy. |
|
|
Fix 7e93dd876e5406d5851419139f8b7b14d6136147 committed to develop branch: fix 0004911 (CircularConstraint magically does not need to be active), repo: spring changeset id: 6624 |
|
|
Apparently the original issue still exists, attaching a new demo |
|
|
Hmm, there seem to be two things going on there: 1) cost flipflopping (very old problem) 2) paths crossing impassable cliffs (news to me, will see if I can find out why) |
|
|
can it be that in a lower resolution the node is pathable while in a higher one it isn't? (due to the cliff being narrow or something) |
|
|
That fortunately isn't possible (crossing edges have inf-cost) but the problem is more subtle, see screenshot: start position maps to low-res block <3,2> (markers are block centers) goal position maps to low-res block <7,3> low-res path is {<3,2>, <4,2>, <5,3>, <6,3>, <7,3>} So we get a seemingly valid path, but our goal is on the other side of the cliff running through <7,3> and we need to backtrack. To fix this, goal blocks should be rejected if entered from a direction s.t. a high-res path to the actual goal position does not exist. e: looks like a relatively easy change, will test for a bit |
|
|
seems to be fixed by https://github.com/spring/spring/commit/7e6b6ee753aa94b7c65930313aa9f176348f345a |
|
|
ok, closing since the original "endless loop" condition should now be much rarer and requires a major redesign to tackle fully. |
| Date Modified | Username | Field | Change |
|---|---|---|---|
| 2015-07-20 16:06 | lhog | New Issue | |
| 2015-07-20 16:06 | lhog | File Added: 20150720_165853_IsisDelta_v02_100.sdf | |
| 2016-02-16 01:00 | hokomoko | File Added: 20160216_015957_IsisDelta_v02_100.0.1-1162-gefb8d7epathbug.sdf.gz | |
| 2016-02-16 01:00 | hokomoko | Note Added: 0015792 | |
| 2016-02-16 02:13 | Kloot | Note Added: 0015794 | |
| 2016-02-16 15:22 | Kloot | Assigned To | => Kloot |
| 2016-02-16 15:22 | Kloot | Status | new => assigned |
| 2016-02-16 18:11 | Changeset attached | => spring develop 7e93dd87 | |
| 2016-02-16 18:11 | Anonymous | Note Added: 0015802 | |
| 2016-02-16 18:11 | Anonymous | Status | assigned => resolved |
| 2016-02-16 18:11 | Anonymous | Resolution | open => fixed |
| 2016-02-16 18:24 | hokomoko | Note Added: 0015803 | |
| 2016-02-16 18:24 | hokomoko | Status | resolved => feedback |
| 2016-02-16 18:24 | hokomoko | Resolution | fixed => reopened |
| 2016-02-16 18:24 | hokomoko | File Added: 20160216_192536_IsisDelta_v02_100.0.1-1164-g7e93dd8pathissue.sdf.gz | |
| 2016-02-16 18:46 | Kloot | Note Added: 0015804 | |
| 2016-02-16 18:52 | hokomoko | Note Added: 0015805 | |
| 2016-02-16 18:52 | hokomoko | Note Edited: 0015805 | |
| 2016-02-16 18:53 | hokomoko | Note Edited: 0015805 | |
| 2016-02-16 19:45 | Kloot | Note Added: 0015806 | |
| 2016-02-16 19:45 | Kloot | File Added: mantis4911.jpg | |
| 2016-02-16 19:47 | Kloot | Note Edited: 0015806 | |
| 2016-02-16 20:15 | Kloot | Note Edited: 0015806 | |
| 2016-02-16 22:46 | hokomoko | Note Added: 0015810 | |
| 2016-02-16 23:30 | Kloot | Note Added: 0015812 | |
| 2016-02-16 23:30 | Kloot | Status | feedback => resolved |
| 2016-02-16 23:30 | Kloot | Resolution | reopened => fixed |
| 2016-06-09 16:47 | hokomoko | Relationship added | related to 0005272 |
