final touchups... trim down pts in synthetic maps

This commit is contained in:
Dustin Carlino 2019-01-31 13:40:11 -08:00
parent 7432ec1b42
commit 8b3e3d8f27
5 changed files with 21 additions and 37 deletions

View File

@ -5,8 +5,8 @@
0,
{
"center": {
"inner_x": 375.2584876887879,
"inner_y": 394.4733648122756
"inner_x": 375.8089,
"inner_y": 394.3162
},
"intersection_type": "Border",
"label": "west"
@ -16,8 +16,8 @@
2,
{
"center": {
"inner_x": 426.5473937988281,
"inner_y": 394.67230224609375
"inner_x": 426.7998,
"inner_y": 394.2691
},
"intersection_type": "TrafficSignal",
"label": null
@ -27,8 +27,8 @@
3,
{
"center": {
"inner_x": 426.9548681897562,
"inner_y": 338.6898394428313
"inner_x": 427.4046,
"inner_y": 340.1785
},
"intersection_type": "Border",
"label": "north"
@ -38,8 +38,8 @@
5,
{
"center": {
"inner_x": 426.9279815253348,
"inner_y": 456.09920803179926
"inner_x": 426.4618,
"inner_y": 457.0412
},
"intersection_type": "Border",
"label": "south"

View File

@ -5,8 +5,8 @@
0,
{
"center": {
"inner_x": 288.5454406738281,
"inner_y": 486.2385864257813
"inner_x": 288.0221,
"inner_y": 487.1246
},
"intersection_type": "StopSign",
"label": null
@ -16,8 +16,8 @@
1,
{
"center": {
"inner_x": 512.0357055664062,
"inner_y": 486.63079833984375
"inner_x": 511.1362,
"inner_y": 486.7896
},
"intersection_type": "StopSign",
"label": null
@ -53,8 +53,8 @@
{
"label": "north",
"center": {
"inner_x": 351.30378272601706,
"inner_y": 450.3966017603446
"inner_x": 353.4848,
"inner_y": 451.9377
}
}
],
@ -63,8 +63,8 @@
{
"label": "south",
"center": {
"inner_x": 439.6870726625188,
"inner_y": 515.314383299414
"inner_x": 440.4084,
"inner_y": 517.7681
}
}
]

View File

@ -23,10 +23,6 @@
- also could be benchmarks; just arbitrary data over time
- also screenshots
- things fixed-pt should solve
- determinism tests failing
- polyline intersection() finding a line hit, but then failing on get_slice_ending_at
- improve test code and explore problems
- big timestep, does follow error blow up?
- alternative to scenario is a sequence of commands for tests

View File

@ -2,12 +2,7 @@
## Geometry
- try fixed pt again, for determinism purposes mostly
- can natively order
- then work on proper SI types, with negative/positive cases handled carefully
- also bounds?
- cant get rid of the ccw intersection check... different answer in some cases that looks bad
- can natively order geom stuff now!
- underlying problems
- bad polyline shifting remains

View File

@ -5,17 +5,10 @@
## Map layer
- pt2d resolution
- handle turns with 0 geometry first.
- maybe Traversable having helpers is also bad. if it can fail for zero-geom turns.
- these seem to exist in the first place due to intersections that ought to be merged! argh.
- savestating test should work now!
- make sure display is always just 2 decimal places... should be though.
- FP math still sucky... debug a polyline, look at the deltas.
- get rid of HashablePt2D?
- then my own physics types
- fixed precision math
- more careful geom types, with negative/positive cases
- also bounds?
- cant get rid of the ccw intersection check... different answer in some cases that looks bad
- maybe also the time to split into different lane types? what's similar/not between them?
- graph querying?