Tomcat/Version History

From Robowiki
< Tomcat
Revision as of 19:20, 18 June 2012 by Jdev (talk | contribs) (The penultimate version of the Tomcat:))
Jump to navigation Jump to search

v. 3.57f

Date: 19.06.12
Results: retried

  • bug introduced in version 3.57e with visit logs processing fixed

v. 3.57e

Date: 19.06.12
Results: retried

  • bug in wave hit interval fixed
  • bug with using incorrect enemy pos for my waves launching fixed
  • calculate visit logs efficiency from start of battle (earlier, visit logs are added to processing only after enemy's 4 hits)

Version comment: I planed to make v. 3.57d last version of Tomcat and switch to development of ConceptA, but first version of ConceptA attract my attention to Tomcat's problem with head on targeters. And researching it i found 3 bug in Tomcat fixed in this version. I do not know already, but my current plan is release Tomcat's v. 3.58 with better head on performance and stop development. But who knows - this day shows, that each last version has ways to improvement:)


v. 3.57d

Date: 14.06.12
Results: RoboRumble ‒ APS: 87.41% (3rd), PL: 897-2 (3rd), Survival: 95.61%

  • adaptive data views weighting in gun removed
  • anti-surefer dataview #3 removed

v. 3.57c

Date: 14.06.12
Results: retired

  • Bug with using incorrect danger for BO in gun because caching fixed

v. 3.57b

Date: 27.03.12
Results: RoboRumble ‒ APS: 87.32% (3rd), PL: 896-2 (2nd), Survival: 95.53%

  • Normalize log efficiencies by bullet flight time in movement

v. 3.57a

Date: 23.03.12
Results: RoboRumble ‒ APS: 87.18% (3rd), PL: 859-2 (3rd), Survival: 95.59%

  • My bullets launch position calculation bug fixed
  • Use correct snapshot in movement
  • Some assertions added

v. 3.57

Date: 22.03.12
Results: RoboRumble ‒ APS: 86.5% (3rd), PL: 856-4 (3rd), Survival: 95.04%

  • Bug with using outdated position fixed in TSes, future bullet creation & simple BOs calculation. But fixing this bug in movement decrease performance because some reason...

v. 3.56

Date: 21.03.12
Results: waiting

  • major code clean up
  • LXXRobot* hierarhy rewrited
  • TurnSnapshots rewrited

v. 3.55

Date: 13.02.12
Results: RoboRumble ‒ APS: 87.03% (3rd), PL: 859-1 (2nd), Survival: 95.36%

  • adaptivr views weighting system inplemented in gun
  • anti-surfer #3 view added
  • firstBulletFlight time attribute removed from distance and timeSinceDirChange views

v. 3.54a

Date: 20.01.12
Results: RoboRumble ‒ APS: 87.27% (3rd), PL: 844-1 (2nd), Survival: 95.74%

  • data analysis refactored

v. 3.54

Date: 23.12.11
Results: RoboRumble ‒ APS: 87.26% (3rd), PL: 841-0 (1st), Survival: 95.75%

  • filtering out of entries following each other in time in gun fixed
  • log hit rates length in AEGM increased to 9, 45 & 5000 correspondly

v. 3.53c

Date: 22.12.11
Results: RoboRumble ‒ APS: 87.12% (3rd), PL: 840-1 (2nd), Survival: 95.68%

  • heap sort fixed

v. 3.53b

Date: 22.12.11
Results: retired

  • intersection calculation in RTree optimised: do not calculate intersection for children, if parent intersection is FULL; calculate intersection only on step from parent to child; unnecessary comparsion removed from intersection calculation
  • heap sort fixed

v. 3.53a

Date: 20.12.11
Results: RoboRumble ‒ APS: 87.12% (3rd), PL: 839-2 (3rd), Survival: 95.69%

  • use step by step heap sort to select only needed entries in movement (sort time reduced by 4 times)
  • rectangles overlaping in RTree preventing added (range search time reduced by 20%)

v. 3.53

Date: 19.12.11
Results: RoboRumble ‒ APS: 87.15% (3rd), PL: 840-1 (2nd), Survival: 95.77%

  • RTree implemented
  • PSTree removed
  • Now Range Search works correctly in faster in 3-5 times

v. 3.52b

Date: 15.12.11
Results: RoboRumble ‒ APS: 87.03% (3rd), PL: 840-1 (2nd), Survival: 95.64%

  • check bullet shadows optimisation reverted
  • prediction data precalculating reverted
  • bullet shadows calculating optimisation reverted
  • major code cleanup reverted

v. 3.52a

Date: 15.12.11
Results: retired

  • getFuturePosition optimisation reverted

v. 3.52

Date: 15.12.11
Results: retired

  • updating prediction data for bullets in air reworked
  • getSimilarEntries optimised
  • check bullet shadows optimised
  • prediction data precalculating added
  • getFuturePosition optimised
  • bullet shadows calculating optimised
  • major code cleanup
  • NPE in BulletManager avoiding added

v. 3.51

Date: 10.12.11
Results: RoboRumble ‒ APS: 87% (3rd), PL: 840-1 (2nd), Survival: 95.6%

  • CPU usage profiling added
  • prediction data management reworked: now do not unnecessary calculations and update bullets in air with data from bulletHitBullet and hitByByllet events

v. 3.50a.opts.3

Date: 10.12.11
Results: RoboRumble ‒ APS: 86.93% (3rd), PL: 841-1 (1st), Survival: 95.45%

  • use distance square for points generation loop break
  • dist to enemy danger component calculation optimised
  • enable avoiding collisions with enemy only when distance between is less than 100

v. 3.50opt39

Date: 09.12.11
Results: RoboRumble ‒ APS: 86.83% (3rd), PL: 841-2 (2nd), Survival: 95.39%

  • work with walls (smoothing, distancing etc.) optimisation
  • limits container in PSTree changed from map to array
  • isEnemyRammingNow() optimised
  • angle calculation in DistanceController optimised
  • RobotImage speed & absoluteHeading optimised
  • waves processing optimised
  • MovementDecision optimised - explicit collisions with walls avoiding removed
  • getPointDanger() optimised
  • wall smoothing optimised - do not smoothing if min distance to wall greater, than wall stick
  • avoiding with enemy collisions optimised

v. 3.50

Date: 06.12.11
Results: RoboRumble ‒ APS: 86.9% (3rd), PL: 840-1 (1st), Survival: 95.49%

  • use projected position on first bullet flight time for calculating enemy BO on first and second bullet attributes

v. 3.49

Date: 04.12.11
Results: RoboRumble ‒ APS: 86.93% (3rd), PL: 840-1 (2nd), Survival: 95.49%

  • use 3 best logs in terms of log hit rate - log miss rate and 3 best logs in terms of enemy hit rate

v. 3.48c

Date: 30.11.11
Results: RoboRumble ‒ APS: 86.85% (3rd), PL: 841-0 (1st), Survival: 95.42%

  • revert to v3.48
  • lateral* usage reworked
  • lateral velocity calculation bug refixed

v. 3.48a

Date: 29.11.11
Results: RoboRumble ‒ APS: 86.96% (3rd), PL: 839-2 (2nd), Survival: 95.51%

  • lateral velocity calculation bug fixed

v. 3.48

Date: 28.11.11
Results: RoboRumble ‒ APS: 86.92% (3rd), PL: 840-1 (2nd), Survival: 95.45%

  • reselect orbit direction, when enemy velocity signum changed, instead of acceleration signum
  • do not reselect orbit direction, when distance between changed more than 25px
  • sort predicted bearing offset to points danger calculations optimisation
  • remove calculation danger on second wave and enemy attack angle
  • multiply danger of bullet to bullet power
  • increase same orbit direction multiplier to 0.95
  • points generator clean up and refactoring
  • reuse points generation method in pif method
  • bug -8 acceleration in movement decision applying fixed
  • evaluate second wave surfing for 6 safest points in both directions

v. 3.47b RoboRumble ‒ APS: 86.81% (3rd), PL: 839-1 (1st), Survival: 95.32%

  • movement mechanics reworked - select orbit direction to destination point every tick
  • use exactly same code for real movement and movement prediction

v. 3.47a RoboRumble ‒ APS: 86.86% (3rd), PL: 839-1 (1st), Survival: 95.4%

  • antiSurf2 data view in gun fixed - second lastVisitedGF1 changed to lastVisitedGF2

v. 3.47 RoboRumble ‒ APS: 86.84% (3rd), PL: 838-2 (3rd), Survival: 95.38%

  • WaveSurfingMovement refactored
  • multiply wave surfing implemented

v. 3.46 RoboRumble ‒ APS: 86.71% (3rd), PL: 839-1 (1st), Survival: 95.24%

  • revert to v. 3.44mt1

v. 3.45 RoboRumble ‒ APS: 86.82% (3rd), PL: 839-1 (1st), Survival: 95.35%

  • enemy visits history gathering added
  • added data view to gun with 2 last visited gfs and hi time importance

v. 3.44mt1 RoboRumble ‒ APS: 86.79% (3rd), PL: 839-0 (1st), Survival: 95.33%

  • use visit logs only after 5 hits
  • use no danger for future bullets, after fire time is passed
  • recalculate orbit direction if distance to travel < stop distance + 8
  • danger formula tweaks
  • use min distance to enemy instead of distance to enemy cur pos

v. 3.44 RoboRumble ‒ APS: 86.11% (4th), PL: 838-1 (1st), Survival: 95.01%

  • divide point danger by (1 - robot's shadowed interval length)

v. 3.43 RoboRumble ‒ APS: 85.35% (6th), PL: 835-4 (3rd), Survival: 94.22%

  • bot effective interval doubled in gun

v. 3.42 RoboRumble ‒ APS: 85.23% (7th), PL: 839-2 (1st), Survival: 94.09%

  • best logs management reworked
  • use 4 best logs from each category

v. 3.41p RoboRumble ‒ APS: 84.94% (7th), PL: 836-3 (3rd), Survival: 93.86%

  • bug fix in wave to surf selection

v. 3.41o2 RoboRumble ‒ APS: 85.02% (7th), PL: 834-4 (4th), Survival: 93.95%

  • everything, that has influence optimised. Now Tomcat is fast almost as nano:)

v. 3.41a RoboRumble ‒ APS: 84.85% (7th), PL: 835-2 (3rd), Survival: 93.97%

  • bug with using incorrect turn snapshot to recalculating prediction data fixed
  • bug with not reselecting orbit direction after bullets on air updated fixed
  • use stop distance + 8 to determine time to stop

v. 3.41 RoboRumble ‒ APS: 84.78% (7th), PL: 829-6 (6th), Survival: 93.59%

  • bug fix in danger calculation function - now distance to center and attack angle really has influence
  • distance to enemy component added to danger function
  • remove danger from future waves, if time to fire more than time to stop or predicted fire time passed

v. 3.40 RoboRumble ‒ APS: 84.8% (7th), PL: 832-3 (3rd), Survival: 93.7%

  • bullet shadows processing reworked - now in WS use all predicted bullets, but at prediction time shadowed bullets add with danger 0.01
  • lxx bullet search in bullet manager fixed

v. 3.39b RoboRumble ‒ APS: 84.79% (7th), PL: 830-5 (5th), Survival: 93.67%

  • reselect orbit direction, when bullet shadows changed
  • bug fix with not removing shadows of intercepted bullets

v. 3.39 RoboRumble ‒ APS: 85.01% (7th), PL: 830-3 (3rd), Survival: 93.75%

  • bullet shadows implemented

v. 3.38 RoboRumble ‒ APS: 84.67% (7th), PL: 830-3 (3rd), Survival: 93.38%

  • Movement challenger mode added
  • enemy turn rate attributes added to common gun views

v. 3.37 RoboRumble ‒ APS: 84.61% (7th), PL: 828-5 (4th), Survival: 93.36%

  • Increase desired distance from 450 to 650

v. 3.36 RoboRumble ‒ APS: 84.32% (8th), PL: 828-5 (5th), Survival: 93.18%

  • revert to v3.34

v. 3.35 RoboRumble ‒ APS: 84.28% (8th), PL: 825-8 (7th), Survival: 93.08% RETIRED

  • add data view to gun with high time importance and attributes: speed, acceleration, lateral dir, bearing

v. 3.34 RoboRumble ‒ APS: 84.31% (7th), PL: 825-6 (5th), Survival: 93.17%

  • use kNN search for data analysis in gun

v. 3.33 RoboRumble ‒ APS: 83.34% (15th), PL: 824-7 (6th), Survival: 92.33%

  • use 5 bullets per log instead of 3 in movement

v. 3.32 RoboRumble ‒ APS: 83.03% (18th), PL: 821-10 (9th), Survival: 92.15%

  • update prediction for bullets on air after bullet intercepted, hitted and missed events

v. 3.31 RoboRumble ‒ APS: 82.74% (19th), PL: 823-8 (8th), Survival: 91.78%

  • my distance last 10 ticks attribute added to hit logs

v. 3.30 RoboRumble ‒ APS: 82.67% (20th), PL: 815-15 (13th), Survival: 91.77%

  • revert to v3.29
  • fill visit logs in wave broken events
  • calculate all logs bearing offsets on prediction time
  • getRobotWidthRadians optimised

v. 3.29 RoboRumble ‒ APS: 81.87% (28th), PL: 817-14 (12th), Survival: 91.13% RETIRED

  • bug with not use visit logs fixed
  • limits for range search in movement fixed
  • avoid of filtering out all entries in targeting added

v. 3.28 RETIRED

  • minor optimisations
  • minor refactoring
  • new tree for range search

v. 3.27 RoboRumble ‒ APS: 82.51% (22nd), PL: 820-10 (9th), Survival: 91.63%

  • turn snapshots by time comparing optimised
  • unused sort data by distance removed
  • linear all entries storing removed from PSTree
  • reselect orbit direction on bullet prediction data updated
  • increase future danger interval to +\- 120 degrees with step 1 degree

v. 3.26 RoboRumble ‒ APS: 82.53% (22nd), PL: 816-14 (12th), Survival: 91.66%

  • antiram tweaks
  • first wave danger weight increased by 50%
  • logs generation fixed

v. 3.25 RoboRumble ‒ APS: 82.42% (22nd), PL: 818-11 (11th), Survival: 91.52%

  • add ram avoiding (to allow turn)
  • attack angle boundaries fixed (was little attack angle for escape and big for coming near)

v. 3.24 RoboRumble ‒ APS: 82.37% (23rd), PL: 816-13 (12th), Survival: 91.52%

  • add uniform 0.2 danger to future bullets with fire delay more than 2 ticks
  • increase min attack angle to 70 degrees
  • decrease attack angle base to 30 degrees
  • danger calculation changed to "not discrete"
  • opponent attack angle to danger function added
  • smooth walls only in orbit direction
  • distance to wall threshold decresead by 22 "pixel"
  • wall sitck buffer removed
  • wall stick decreased to 120 "pixel"
  • if wall smooth needed, then always smooth angle at least on 4 degrees
  • multiply danger of prediction same as prev prediction on 0.9
  • increase robot effective width to 75% for full danger and 255% for half danger
  • skip bullet with travel time less than 2 tick if there another bullets
  • add breaking if robot going directly to opponent
  • distance to enemy removed from danger formula
  • use clear distance to center in danger formula
  • ramming detection fixed

v. 3.23 RoboRumble ‒ APS: 82.18% (25th), PL: 817-12 (11th), Survival: 91.54%

  • bug fix in robot width radians calculation

v. 3.22 RoboRumble ‒ APS: 82.22% (25th), PL: 821-7 (6th), Survival: 91.69%

  • use of enemy miss data to log selection algorithm added

v. 3.21 RoboRumble ‒ APS: 81.41% (32nd), PL: 813-14 (11th), Survival: 90.82%

  • change log selection algorithm

v 3.20 RoboRumble ‒ APS: 82.06% (25th), PL: 814-12 (11th), Survival: 91.42%

  • change log selection formula

v 3.19.183 RoboRumble ‒ APS: 81.82% (27th), PL: 810-16 (13th), Survival: 91.05%

  • revert to version 3.17.169

v 3.18.178 RoboRumble ‒ APS: 81.7% (31st), PL: 802-24 (20th), Survival: 90.89%

  • always use of my lateral speed attr in visit logs added
  • removing visit logs with less than 2 attributes
  • time since my dir change attr added to hits and visits log
  • gun data analise optimized

v 3.17.169 RoboRumble ‒ APS: 81.92% (27th), PL: 804-21 (18th), Survival: 91.2%

  • another stupid bug fix: movement data did not updated by bulletHitBullet events

v 3.16.167 RoboRumble ‒ APS: 81.36% (32nd), PL: 798-18 (15th), Survival: 90.7%

  • getFuturePosition optimised
  • stupid LinkedList addAll and sorting removed
  • effective bearing offset interval in targeting reduced to 75%
  • angle step in targeting reduced to 0.5 rad

v 3.15.155 RoboRumble ‒ APS: 81.42% (32nd), PL: 798-18 (14th), Survival: 90.86%

  • sort data by time in movement data analysis added
  • hypercube calculation for movement data analyses improved (use different side lengths for different dimensions)

v 3.14.153 RoboRumble ‒ APS: 80.46% (40th), PL: 778-38 (36th), Survival: 89.63%

  • new movement buffers managment system - auto selecting buffers (hits and visits) and progressive visits buffer enabling
  • bullets affecting area in WS increased to 155% of robot width radians

v 3.13.152 RoboRumble ‒ APS: 79.6% (45th), PL: 774-42 (38th), Survival: 88.95%

  • continue of zero lateral velocity bug fix (issue #3)

v 3.12.151 RoboRumble ‒ APS: 79.52% (46th), PL: 771-44 (40th), Survival: 88.84%

  • zero lateral velocity bug fix (issue #3)
  • anti ram fire power selection tweak

v 3.11.146 RoboRumble ‒ APS: 78.65% (52nd), PL: 743-73 (65th), Survival: 87.44%

  • default bullet power increasing reverted

v 3.10.145 RoboRumble ‒ APS: 77.69% (57th), PL: 738-77 (68th), Survival: 86.62% RETIRED

  • fire selection bug fix (issue #2)
  • default fire power increaded to 2.05

v 3.9.142 RoboRumble ‒ APS: 78.01% (57th), PL: 739-77 (66th), Survival: 85.33% RETIRED

  • power management tweaked for more powerful bullets

v 3.8.139 RoboRumble ‒ APS: 78.95% (50th), PL: 750-66 (63rd), Survival: 87.64%

  • initial release