Virtual Tunnel Circuit Finals Videos

Issue #258 resolved
Martin Dlouhy created an issue

Hello Tunnel Circuit teams,

are you willing to share some videos from your runs? The proposal [1] for direct replay of simulations during evaluation was not accepted (yet?), so it is up to us 🙂

[1] https://community.subtchallenge.com/t/transparent-virtual-track-evaluation/1145

Comments (17)

  1. Martin Dlouhy reporter

    Here is our first video for Robotika team, run 01 based on extracted OSGAR logfile:

    https://youtu.be/WiunGpoDvUE

    Score 0 points.

    Note robot A10F2400L and B60F1500R running almost simultaneously although A should start at time 10s (simulated time) and B at time 60s.

    I suppose that it is World A “Circuit Board”

    https://robotika.cz/competitions/subtchallenge/virtual-tunnel-circuit/competition-worlds.jpg

    p.s. unfortunately B, C and D did not dump any log for run 01

  2. Martin Dlouhy reporter

    Robotika 02, logs from robots B and D (there are 4 robots A, B, C, D):

    https://youtu.be/HsycOeeyqtY

    https://youtu.be/zm7ktpMHpWg

    Score 0 points, although robot D found Survivor, but in the 3rd exploration wave “died” on stones …

    It confirms that the times of robots are not synchronized and hopefully it will solve topic /clock (as mentioned in #145). You can see collision of robots A and B at the tunnel entry …

  3. Martin Dlouhy reporter

    Robotika 03, again available logs for robots B and D. It is more-or-less the same as 2nd run except it finally scored thanks to a good luck and return to start with info about Survivor:

    https://youtu.be/XX_HBa2I4ZE

    Note, that in this run robot A started after B … should be opposite with 50s delay … next time.

  4. Martin Dlouhy reporter

    World2 was disaster (12 times, no OSGAR log)

    RobotA, run 04

    Python3: stdout 0:15:00.102140 (49.4 -5.9 0.0) [('scan', 240)] 
    ...
    Python3: stdout 0:47:25.972494 (49.4 -5.9 0.0) [('scan', 111)] 
    Python3: stdout 0:48:00.179047 (-498.3 717.8 0.0) [('acc', 613), ('image', 258), ('pose2d', 350), ('rot', 613), ('scan', 134), ('sim_time_sec', 8)]
    Python3: stdout 0:49:00.069503 (-498.3 717.8 0.0) [('acc', 1273), ('rot', 1273), ('scan', 243)] 
    

    Pure robot … it simply was not receiving sensor data!

  5. Martin Dlouhy reporter

    RobotB, run 04

    Python3: stdout 0:10:59.525324 (43.6 -8.4 0.0) [('acc', 659), ('pose2d', 625), ('rot', 659), ('sim_time_sec', 12)] 
    Python3: stdout 0:11:00.062752 (43.3 -8.4 0.0) [('acc', 6), ('image', 3), ('pose2d', 7), ('rot', 6), ('scan', 2), ('sim_time_sec', 1)] 
    

    this was after entering the tunnel …

    No wonder that Robotika scored 0, 0, 0 in “Tree of Life” (of Death 😞 ), sigh.

  6. Martin Dlouhy reporter

    RobotC, run06 - one more random sample:

    Python3: stdout 0:37:02.190023 (-95.7 92.3 0.0) [('acc', 3), ('image', 1), ('origin', 296), ('rot', 3), ('scan', 17)] 
    Python3: stdout 0:38:00.201117 (-95.7 92.3 0.0) [('acc', 2179), ('image', 93), ('origin', 474), ('rot', 2179), ('scan', 52)] 
    Python3: stdout 0:39:00.114822 (-95.7 92.3 0.0) [('acc', 1231), ('origin', 492), ('rot', 1231)] 
    

    it is statistics of received messages related to ROS topics per minute (i.e. 3x IMU data, 1 image, 296 origin reports, 17 scans, no odometry … for example

  7. Martin Dlouhy reporter

    OK, here is one more video from run 10, robot C:

    https://youtu.be/yJhcvtdr-Oo

    No wonder this tunnel is named “Are we there yet?” 🙂

    Interesting points:

    1. robot B did not move at all
    2. robot A got crazy in the tunnel (robot C is passing by it)

    and here is the reason of craziness (I would be also mad)

    ...
    0:17:00.174657 (115.5 36.3 -0.0) [('acc', 1175), ('pose2d', 1174), ('rot', 1175), ('scan', 235), ('sim_time_sec', 23)]
    0:18:00.071696 (139.0 36.3 -0.0) [('acc', 1174), ('pose2d', 1173), ('rot', 1174), ('scan', 235), ('sim_time_sec', 24)]
    0:19:00.167036 (154.7 47.1 -0.0) [('acc', 1166), ('pose2d', 1167), ('rot', 1166), ('scan', 233), ('sim_time_sec', 23)]
    0:20:00.112116 (154.7 70.5 -0.0) [('acc', 1153), ('pose2d', 1171), ('rot', 1153), ('scan', 234), ('sim_time_sec', 24)]
    0:21:00.103524 (155.1 92.8 -0.0) [('pose2d', 1164), ('scan', 233), ('sim_time_sec', 23)]
    0:22:00.040778 (155.6 115.8 -0.0) [('pose2d', 1152), ('scan', 230), ('sim_time_sec', 23)]
    0:23:00.225396 (156.0 139.0 -0.0) [('pose2d', 1159), ('scan', 232), ('sim_time_sec', 23)]
    0:24:00.096273 (156.5 162.0 -0.0) [('pose2d', 1151), ('scan', 231), ('sim_time_sec', 23)]
    ...
    0:53:00.057601 (168.7 791.2 -0.0) [('pose2d', 1038), ('scan', 207), ('sim_time_sec', 21)]
    0:54:00.026260 (169.1 811.5 -0.0) [('pose2d', 1049), ('scan', 210), ('sim_time_sec', 21)]
    0:55:00.071333 (169.5 832.3 -0.0) [('pose2d', 1040), ('scan', 208), ('sim_time_sec', 21)]
    0:56:00.042111 (169.9 853.2 -0.0) [('pose2d', 1046), ('scan', 209), ('sim_time_sec', 21)]
    0:57:00.013416 (170.3 874.3 -0.0) [('pose2d', 1054), ('scan', 211), ('sim_time_sec', 21)]
    0:58:00.119442 (170.7 895.7 -0.0) [('pose2d', 1070), ('scan', 214), ('sim_time_sec', 21)]
    0:59:00.086211 (170.8 900.1 -0.0) [('pose2d', 220), ('scan', 216), ('sim_time_sec', 5)]
    1:00:00.056517 (170.8 900.1 -0.0) [('scan', 214)]
    1:01:00.139475 (170.8 900.1 -0.0) [('scan', 214)]
    1:02:00.042254 (170.8 900.1 -0.0) [('scan', 213)]
    1:03:00.064216 (170.8 900.1 -0.0) [('scan', 218)]
    1:04:00.009180 (170.8 900.1 -0.0) [('scan', 219)]
    1:05:00.098884 (170.8 900.1 -0.0) [('scan', 220)]
    1:06:00.197615 (170.8 900.1 -0.0) [('scan', 225)]
    1:07:00.028404 (170.8 900.1 -0.0) [('scan', 222)]
    1:08:00.147307 (170.8 900.1 -0.0) [('scan', 220)]
    ...
    

    so after 20 minutes (real time) we lost IMU and after 1hour also odometry …

  8. Martin Dlouhy reporter

    Here is the last video I am able to extract from the final runs in Virtual Tunnel Circuit 15/B60F1500R (long and boring):

    https://youtu.be/mvNtzEm7ME0

    Due to dangerous situation:

    0:19:14.677479 Pitch limit triggered termination: (pitch 57.0)

    it temporarily switched to second side but the robot was not able to recognize that it is trapped on a island. When it returned to staging area it for unknown reason "jumped of the cliff". 😞 Never the
    less it scored regularly 1 point, which I suppose was robot D (shorter exploration to the right).

    So in total 7 OSGAR logs of 60 simulated robot runs (I uploaded only 6 because one was almost identical in repeated simulation).

  9. Nate Koenig

    I don't believe there is a "solution" to this issue. It seems like general discussion that doesn't need to be solved.

  10. Martin Dlouhy reporter

    🙂 well, actually I tried to post question there and the recommendation there was to post it here instead, which I already did before … but never mind. The original topic was to share videos with other teams, but others are probably not interested, so lets close it.

    I would use #261 as follow up

  11. Log in to comment