search

eclipse-sumo / sumo

Eclipse SUMO is an open source, highly portable, microscopic and continuous traffic simulation package designed to handle large networks. It allows for intermodal simulation including pedestrians and comes with a large set of tools for scenario creation.
https://eclipse.dev/sumo
Eclipse Public License 2.0
2.53k stars 1.43k forks source link

Improve lane-changing model #1045

Closed behrisch closed 3 years ago

behrisch commented 11 years ago

this is for work on MSLCM_JE2013''''''

differences between JE2013 and the default model (LC2013)

open sub-tickets (not bugs)

behrisch commented 11 years ago

counter-lane change deadlocks are less common with 186a148 (aka r14816)

behrisch commented 11 years ago

added features (dontBrake, lastAccel, disableAMBACKBLOCKER1) in 9549ccd (aka r14839) improving performance of A92-scenario

behrisch commented 11 years ago

@namdre commented:

fixed an aspect of #1044 in fec87da (aka r14890) improving performance of A92-scenario considerably

behrisch commented 11 years ago

@namdre commented:

the functionaltiy of JE2013 in fec87da (aka r14890) has been made available as new model "DK2013" in [14893,14894]. Development continues in JE2013.

behrisch commented 11 years ago

@namdre commented:

relaced conditions for LCA_AMBLOCKINGFOLLOWER 21313b1 (aka r14904) improving performance of A92-scenario

behrisch commented 11 years ago

@namdre commented:

activated feature 'lcaRightImpatience' in f717427 (aka r14905)

behrisch commented 11 years ago

@namdre commented:

added feature 'laMinSpeed' in 09c018a (aka r14912) improving performance of A92, Orinoko, Acosta and Pasubio (joined is a bit worse, probably because fewer people use the inside lane of the roundabout)

behrisch commented 11 years ago

@namdre commented:

added feature 'equalBest' in 3fff33c (aka r14913) improving performance of A92, Orinoko (mixed results in iTETRIS)

behrisch commented 11 years ago

@namdre commented:

added feature firstBlocked in 57acca0 (aka r14959) which fixes some counter-lc deadlocks

behrisch commented 11 years ago

@namdre commented:

in 2424033 (aka r15035) resetting laneChanger state before any calls to laneChangeModel. This prevents resetting of messages sent to followers.

behrisch commented 11 years ago

@namdre commented:

in fc934b2 (aka r15051) omplete rewrite of speed adaption in regard to blocking vehicles. The basic idea is to distinguish more cases when being blocked: overtaking/falling behind a leader and cutting in before/being overtake by a follower. Fine grained control over speeds now moves into vSafes and patchSpeed only executes these requested speeds.

also added trigger for keeping right when the neighboring vehicles are faster

behrisch commented 11 years ago

@namdre commented:

in e45c2a8 (aka r15055) now using already existing vSafe requests when deciding about behavior in regard to blockers. This is a big improvement in the A92 scenario

behrisch commented 11 years ago

@namdre commented:

in 702650c (aka r15056) preventing overtaking on the right on uncongested highway, relaxed conditions for cutting in before a follower. This is a big improvement in the A92 scenario.

behrisch commented 11 years ago

@namdre commented:

in 58d7a80 (aka r15057) preventing overtaking on the right on uncongested highway (regardless of own lc-wish), no longer using myAcceleration to predict next acceleration refs #1045. This is another big improvement in the A92 scenario

behrisch commented 11 years ago

@namdre commented:

in 98119e2 (aka r15061) introducing hysteresis function for lookAhead speed. setting minimum lookahead speed to 0 to allow more flexibility when jammed

behrisch commented 11 years ago

@namdre commented:

in 9b782ca (aka r15066) introducing new feature 'furtherBlock' this avoids deadlocks when vehicles have to counter-lc across multiple lanes. also relaxing condtions for moving away from the best lane to the right on highways. This fixes the test stops/overtaking_highspeed

behrisch commented 11 years ago

@namdre commented:

in c7a2280 (aka r15088) fixing another case of counter-lc deadlock (if the follower cannot leave enough space for the leader, the leader must reserve space for the follower instead

behrisch commented 11 years ago

@namdre commented:

in 1835ef7 (aka r15089) fixing computation of overtaking distance, preventing deadlock when there is not enough space to overtake. Big improvement in A92 scenario.

behrisch commented 11 years ago

@namdre commented:

in 564925e (aka r15105) driving cooperatively inside multi-lane roundabouts (see #789). This helps a lot in the acosta scenario

behrisch commented 11 years ago

@namdre commented:

in dd6f8ee (aka r15107) avoiding deadlocks in roundabout (no dist bonus if on the final roundabout edge), honor vSafe requests while on an internal lane, consider slowing down due to myLeftSpace when informing blockers

behrisch commented 11 years ago

@namdre commented:

in 460c84c (aka r15108) now considering myLeadingBlockerLength before when planning speed adaptiong for blockers. some improvements in A92

behrisch commented 11 years ago

@namdre commented:

in 99b7bf7 (aka r15173) left-changing vehicles may now slow down blocking followers over multiple steps as long as they are still fast (> 27m/s). This improves flow in A92 scenario and still prevents vehicles waiting on an on-ramp from slowing down the main flow.

behrisch commented 11 years ago

in 9b5fcb6 (aka r15284) avoiding some lane changes to resolve issue #1097

behrisch commented 11 years ago

@namdre changed description from:

this is for work on MSLCM_JE2013''''''

to:

this is for work on MSLCM_JE2013''''''

improvements still missing from the default model (LC2013)

  • [15173]
  • [15284]
behrisch commented 11 years ago

in 0bc42d3 (aka r15861) added JE2013 feature 99b7bf7 (aka r15173) to LC2013, patched keepRight (see #1139) in JE2013 and LC2013

this is for work on MSLCM_JE2013''''''

improvements still missing from the default model (LC2013)

  • [15173]
  • [15284]

to:

this is for work on MSLCM_JE2013''''''

improvements still missing from the default model (LC2013)

  • [15173]
  • [15284] recheck: should depend on distance (see pasubio)

open sub-tickets

behrisch commented 11 years ago

@namdre changed description from:

this is for work on MSLCM_JE2013''''''

improvements still missing from the default model (LC2013)

  • [15173]
  • [15284] recheck: should depend on distance (see pasubio)

open sub-tickets

to:

this is for work on MSLCM_JE2013''''''

improvements still missing from the default model (LC2013)

  • [15173]
  • [15284] recheck: should depend on distance (see pasubio)

open sub-tickets (not bugs)

  • 1136

  • 1266

behrisch commented 11 years ago

@namdre committed 0f3d65a (aka r16464): avoiding division by zero. refs #1316, refs #1045

behrisch commented 11 years ago

@namdre committed b8b6f29 (aka r16494): avoiding division by zero. refs #1316, #1045

behrisch commented 11 years ago

@namdre committed 574840e (aka r16540): refining stayOnBest feature in JE2013 by adding distance threshold. Now improves iTetris stats instead of making it worse. refs #1045

behrisch commented 11 years ago

@namdre changed description from:

this is for work on MSLCM_JE2013''''''

improvements still missing from the default model (LC2013)

  • [15173]
  • [15284] recheck: should depend on distance (see pasubio)

open sub-tickets (not bugs)

  • 1136

  • 1266

to:

this is for work on MSLCM_JE2013''''''

differences between JE2013 and the default model (LC2013)

  • [15173]
  • [15284],[16540] recheck stayOnBest TURN_LANE_DIST 200 improves iTETRIS but makes BS run worse
  • ROUNDABOUT_DIST_BONUS set to 100 (80 in LC2013)

open sub-tickets (not bugs)

  • 1136

  • 1266

behrisch commented 11 years ago

@namdre committed 9965805(../commit/9965805168f24ab4432e3148777117285a03dc3c) (aka r16806): tweaking debug output refs #1045

behrisch commented 11 years ago

@namdre committed 157ac3d (aka r16807): preserving some lane-changing state across simulation steps to avoid edge-ordering depencies (in saveBlockerLength) refs #1045

behrisch commented 11 years ago

@namdre committed 2cd5fda (aka r17057): Some improvements to JE2013 refs #1045

behrisch commented 11 years ago

@namdre changed description from:

this is for work on MSLCM_JE2013''''''

differences between JE2013 and the default model (LC2013)

  • [15173]
  • [15284],[16540] recheck stayOnBest TURN_LANE_DIST 200 improves iTETRIS but makes BS run worse
  • ROUNDABOUT_DIST_BONUS set to 100 (80 in LC2013)

open sub-tickets (not bugs)

  • 1136

  • 1266

to:

this is for work on MSLCM_JE2013''''''

differences between JE2013 and the default model (LC2013)

  • [15173]
  • [15284],[16540] recheck stayOnBest TURN_LANE_DIST 200 improves iTETRIS but makes BS run worse
  • ROUNDABOUT_DIST_BONUS set to 100 (80 in LC2013)
  • [17057]. improves A92 and iTETRIS (mostly) but makes bs_dido worse

open sub-tickets (not bugs)

  • 1136

  • 1266

behrisch commented 11 years ago

@namdre committed af40b91 (aka r17058): respecting option lanechange.overtake-right where required refs #1045

behrisch commented 11 years ago

@namdre changed description from:

this is for work on MSLCM_JE2013''''''

differences between JE2013 and the default model (LC2013)

  • [15173]
  • [15284],[16540] recheck stayOnBest TURN_LANE_DIST 200 improves iTETRIS but makes BS run worse
  • ROUNDABOUT_DIST_BONUS set to 100 (80 in LC2013)
  • [17057]. improves A92 and iTETRIS (mostly) but makes bs_dido worse

open sub-tickets (not bugs)

  • 1136

  • 1266

to:

this is for work on MSLCM_JE2013''''''

differences between JE2013 and the default model (LC2013)

  • [15173]
  • [15284],[16540] recheck stayOnBest TURN_LANE_DIST 200 improves iTETRIS but makes BS run worse
  • ROUNDABOUT_DIST_BONUS set to 100 (80 in LC2013)
  • [17057]. improves A92 and iTETRIS (mostly) but makes bs_dido worse

open sub-tickets (not bugs)

  • 1136

  • 1266

  • 1435

behrisch commented 11 years ago

@namdre committed ed3f548 (aka r17066): fix #1434, refs #1045

behrisch commented 11 years ago

@namdre changed description from:

this is for work on MSLCM_JE2013''''''

differences between JE2013 and the default model (LC2013)

  • [15173]
  • [15284],[16540] recheck stayOnBest TURN_LANE_DIST 200 improves iTETRIS but makes BS run worse
  • ROUNDABOUT_DIST_BONUS set to 100 (80 in LC2013)
  • [17057]. improves A92 and iTETRIS (mostly) but makes bs_dido worse

open sub-tickets (not bugs)

  • 1136

  • 1266

  • 1435

to:

this is for work on MSLCM_JE2013''''''

differences between JE2013 and the default model (LC2013)

  • [15173]
  • [15284],[16540] recheck stayOnBest TURN_LANE_DIST 200 improves iTETRIS but makes BS run worse
  • ROUNDABOUT_DIST_BONUS set to 100 (80 in LC2013)
  • [17057]. improves A92 and iTETRIS (mostly) but makes bs_dido worse

open sub-tickets (not bugs)

  • 1136

  • 1266

  • 1435

  • 1440

behrisch commented 11 years ago

@namdre committed 2bbd7da (aka r17100): callibrating obligatory lane changing (to get a nicer picture for the publication) refs #1045

behrisch commented 11 years ago

@namdre changed description from:

this is for work on MSLCM_JE2013''''''

differences between JE2013 and the default model (LC2013)

  • [15173]
  • [15284],[16540] recheck stayOnBest TURN_LANE_DIST 200 improves iTETRIS but makes BS run worse
  • ROUNDABOUT_DIST_BONUS set to 100 (80 in LC2013)
  • [17057]. improves A92 and iTETRIS (mostly) but makes bs_dido worse

open sub-tickets (not bugs)

  • 1136

  • 1266

  • 1435

  • 1440

to:

this is for work on MSLCM_JE2013''''''

differences between JE2013 and the default model (LC2013)

  • [15173]
  • [15284],[16540] recheck stayOnBest TURN_LANE_DIST 200 improves iTETRIS but makes BS run worse
  • ROUNDABOUT_DIST_BONUS set to 100 (80 in LC2013)
  • [17057]. improves A92 and iTETRIS (mostly) but makes bs_dido worse
  • [17100] still needs testing on BS

open sub-tickets (not bugs)

  • 1136

  • 1266

  • 1435

  • 1440

behrisch commented 11 years ago

@namdre committed 6e2c602 (aka r17102): typo in [17100] refs #1045

behrisch commented 11 years ago

@namdre changed description from:

this is for work on MSLCM_JE2013''''''

differences between JE2013 and the default model (LC2013)

  • [15173]
  • [15284],[16540] recheck stayOnBest TURN_LANE_DIST 200 improves iTETRIS but makes BS run worse
  • ROUNDABOUT_DIST_BONUS set to 100 (80 in LC2013)
  • [17057]. improves A92 and iTETRIS (mostly) but makes bs_dido worse
  • [17100] still needs testing on BS

open sub-tickets (not bugs)

  • 1136

  • 1266

  • 1435

  • 1440

to:

this is for work on MSLCM_JE2013''''''

differences between JE2013 and the default model (LC2013)

  • [15173]
  • [15284],[16540] recheck stayOnBest TURN_LANE_DIST 200 improves iTETRIS but makes BS run worse
  • ROUNDABOUT_DIST_BONUS set to 100 (80 in LC2013)
  • [17057]. improves A92 and iTETRIS (mostly) but makes bs_dido worse
  • [17100],[17102] calibration constant changed, still needs testing on BS

open sub-tickets (not bugs)

  • 1136

  • 1266

  • 1435

  • 1440

behrisch commented 11 years ago

@namdre changed component from "other (also unclassified)" to "microsim - lane change"

behrisch commented 11 years ago

@namdre committed 395b81e (aka r18630): fix #1380, refs #1045

behrisch commented 11 years ago

@namdre committed 3274c6e (aka r19212): fix #1971, refs #1045

behrisch commented 11 years ago

@namdre committed 4a73eff (aka r19215): adapting [19212] formula to take ego speed into account refs #1971, refs #1045

behrisch commented 11 years ago

@namdre committed 68c8a21 (aka r19324): tweaking debug output refs #1045

behrisch commented 11 years ago

@namdre commented:

fixing jumpy lane-changing in 05dce1f (aka r19329) see #1856

behrisch commented 11 years ago

@namdre commented:

another fix to inadvertent lane-changing in 94086f0 (aka r19356) (Thanks to Thomas Lockhart for providing an error scenario)

behrisch commented 11 years ago

JE2013 has been folded into LC2013