AWS DeepRacer League Update #22

Open Division has exploded this week. Join me as I walk through the events on tracks this week.

AWS DeepRacer

AWS DeepRacer is a 1/18th scale autonomous race car but also much more. It is a complete program that has helped thousands of employees in numerous organizations begin their educational journey into machine learning through fun and rivalry.

Visit AWS DeepRacer page to learn more about how it can help you and your organization begin and progress the journey towards machine learning.

Join the AWS Machine Learning Community to talk to people who have used DeepRacer in their learning experience.

Pro Division

Baja Highway track (Source: AWS DeepRacer Console)

Race info:

  • 64.67 meters long
  • 107 centimeters wide
  • Head-to-bot format with four bot cars cruising at 2 m/s with no lane changes
  • Racers are trying to complete three laps – all laps are separate
  • Each collision results in a five seconds penalty
  • Each driving off the track results in a three seconds penalty
  • Top 16 go on into a live race which will be streamed on Twitch to fight for places in the League Finals and AWS DeepRacer EVOs
All submissions
Update time: 2021-07-24 23:00:00 UTC
Comparing to 2021-07-17 13:00:00 UTC

Racers: 119 (+13)
Total submission attempts: 47035 (+20676)
Best total time: 01:06.262 (-00:01.626) by JJ
Mean total time: 03:00.156369748 (-00:04.060281195)
Best lap: 00:22.065 (-00:00.363) by JJ
Mean best lap: 00:55.505369748 (-00:01.129762328)
Mean average lap: 01:00.051806723 (-00:01.353410258)

Pretty nice turnout this week, 13 racers have joined the rest on Baja Highway. JJ had slightly improved his time. On average each racer managed to cut down 4 seconds from their entries

Top 25 submissions
Update time: 2021-07-24 23:00:00 UTC
Comparing to 2021-07-17 13:00:00 UTC

Racers: 25
Total submission attempts: 35486 (+14067)
Best total time: 01:06.262 (-00:01.626) by JJ
Mean total time: 01:14.741080 (-00:08.519880)
Best lap: 00:22.065 (-00:00.363) by JJ
Mean best lap: 00:24.537040 (-00:02.256200)
Mean average lap: 00:24.913440 (-00:02.839880)

Top 25 have managed even more – 8.5 seconds improvement on average. From the tables comparison I can see that the best time to qualify to the finale has gone down by 3 seconds to ZhengYi-NYCU-CGI who has now taken over the leader’s position. Second spot got taken by Yi-LI-NYCU-CGI. Their colleague Yoway has punched the ticket to Vegas in June’s finale but is still racing for pleasure. Sounds like the fight between NYCU students and JPMC’s Ace and Rogue might dominate the race night (it’s starting at 1:30 for me).

The slowest time currently qualifying into the finale is 20th Penguin-DNPds. Their time, 01:19.633 is a decrease of 17.5 seconds compared to the last qualifying time last week.

The time gap between 1st and 16th racer has gone down from over 16 to under 12 seconds. I’m expecting this to go down even more in the upcoming week though the racers who are usually feared the most have already revealed their cards and so this month might be a bit more peaceful.

Many known names reside in the 16 in the finale spots. Let me remind you that despite having qualified into the pro division this month Ace and Rogue are not new to this game, they have both taken time in the championships in 2020.

It’s worth mentioning that skoota in 22nd has started racing in October 2020 and qualified into the Pro Division last month together with Ace and Rogue, taking the third spot right behind them. It’s good to see leading racers from Open Division be competitive in Pro Division as well.

Rank Racer Total Time Avg Lap Best Lap Resets Submissions
1 (↔0) JJ 01:06.262 (-00:01.626) 00:22.087 (-00:00.542) 00:22.065 (-00:00.517) 0 (-0) 3960 (+1346)
2 (–) ZhengYi-NYCU-CGI 01:07.850 (–:–.—) 00:22.616 (–:–.—) 00:22.508 (–:–.—) 0 (–) 125 (–)
3 (⬇1) Karl-NAB 01:08.232 (-00:00.026) 00:22.744 (-00:00.008) 00:22.357 (-00:00.071) 0 (-0) 2561 (+1068)
4 (⬆4) Ernesto 01:08.259 (-00:03.766) 00:22.753 (-00:01.255) 00:22.588 (-00:01.066) 0 (-0) 2087 (+1240)
5 (⬆94) Yi-LI-NYCU-CGI 01:10.476 (-04:26.168) 00:23.492 (-01:28.722) 00:23.412 (-01:24.079) 0 (-73) 223 (+221)
6 (⬇3) JPMC-Ace-Hyderabad 01:10.879 (-00:00.000) 00:23.626 (-00:00.000) 00:23.410 (-00:00.000) 0 (-0) 246 (-0)
7 (⬇3) JPMC-Rogue-Hyderabad 01:10.909 (-00:00.000) 00:23.636 (-00:00.000) 00:23.447 (-00:00.000) 0 (-0) 458 (+32)
8 (⬇3) Yoway-NYCU-CGI 01:11.179 (-00:00.000) 00:23.726 (-00:00.000) 00:23.404 (-00:00.000) 0 (-0) 94 (-0)
9 (⬇3) PolishThunder 01:11.769 (-00:00.000) 00:23.923 (-00:00.000) 00:23.569 (-00:00.000) 0 (-0) 774 (-0)
10 (⬇3) Duckworth 01:11.932 (-00:00.000) 00:23.977 (-00:00.000) 00:23.464 (-00:00.000) 0 (-0) 3317 (+1534)
11 (⬇1) TonyJ 01:12.392 (-00:01.529) 00:24.130 (-00:00.510) 00:23.712 (-00:00.446) 0 (-0) 3434 (+1648)
12 (⬆22) AJM-Model-1 01:12.426 (-01:00.669) 00:24.142 (-00:20.223) 00:23.938 (-00:16.014) 0 (-15) 970 (+957)
13 (⬇4) RobinCastro-DBS 01:13.340 (-00:00.221) 00:24.446 (-00:00.074) 00:24.198 (+00:00.565) 0 (-0) 2513 (+795)
14 (–) rosscomp1 01:13.424 (–:–.—) 00:24.474 (–:–.—) 00:23.718 (–:–.—) 0 (–) 1102 (–)
15 (⬇2) SorinB 01:14.189 (-00:01.895) 00:24.729 (-00:00.632) 00:24.227 (-00:00.835) 0 (-0) 1824 (+1175)
16 (–) GT-DevelopersIO 01:14.764 (–:–.—) 00:24.921 (–:–.—) 00:24.761 (–:–.—) 0 (–) 480 (–)
17 (⬇6) dartjason 01:14.912 (-00:00.923) 00:24.970 (-00:00.308) 00:24.872 (-00:00.043) 0 (-0) 2600 (+1327)
18 (⬇4) Matt 01:15.003 (-00:02.108) 00:25.001 (-00:00.702) 00:24.910 (-00:00.737) 0 (-0) 2279 (+1258)
19 (⬇7) AIDeepRacer 01:15.918 (-00:00.000) 00:25.306 (-00:00.000) 00:25.125 (-00:00.000) 0 (-0) 174 (+157)
20 (⬆7) Penguin-DNPds 01:19.633 (-00:36.490) 00:26.544 (-00:12.163) 00:26.135 (-00:08.193) 0 (-9) 236 (+183)
21 (⬇6) CGS-LargeLattice 01:20.742 (-00:00.000) 00:26.914 (-00:00.000) 00:26.458 (-00:00.000) 0 (-0) 782 (+8)
22 (⬇5) skoota 01:23.940 (-00:05.965) 00:27.980 (-00:01.988) 00:27.588 (-00:01.551) 0 (-0) 1141 (+274)
23 (⬇5) alexj 01:25.323 (-00:05.128) 00:28.441 (-00:01.709) 00:27.838 (-00:02.155) 0 (-0) 95 (+42)
24 (⬇8) Rober 01:26.559 (-00:00.000) 00:28.853 (-00:00.000) 00:27.223 (-00:00.000) 1 (-0) 5 (-0)
25 (⬇4) kimwooglae 01:28.215 (-00:11.033) 00:29.405 (-00:03.677) 00:28.499 (-00:00.755) 0 (-3) 4006 (+1256)

Open Division

Baja Turnpike (Source: AWS DeepRacer Console)

Race info:

  • 40.56 meters long
  • 107 centimetres wide
  • Simpler and shorter than the pro track but still pretty challenging with a variety of straights and turns
  • Time Trial format
  • Competitors need to complete three laps and total time counts for their classification, each lap starts from halt
  • Each off-track results in five seconds penalty.
  • Racers fight for qualification to the Pro group which is won by the top 10%
  • They will also receive a Pro Welcoming Kit which is a bit of a mystery
All submissions
Update time: 2021-07-24 23:00:00 UTC
Comparing to 2021-07-17 13:00:00 UTC

Racers: 1920 (+623)
Total submission attempts: 9836 (+4244)
Best total time: 01:07.131 (-00:14.076) by Janesoki
Mean total time: 04:00.164610417 (-00:03.149971696)
Best lap: 00:21.781 (-00:02.675) by Janesoki
Mean best lap: 01:13.826510417 (-00:01.126105620)
Mean average lap: 01:20.054536458 (-00:01.049997852)

623 racers have joined. Six hundred and twenty three! That’s more than a total number of racers in the Open Division in May. At 1920 it’s the second biggest recorded virtual race in history after Toronto Turnpike in October 2019. We might be looking at a change of a leader in this statistic with only 63 racers difference at this point.

The activity is primarily fuelled by the Udacity AWS Machine Learning Scholarship Program students. They are going through the AWS Machine Learning Foundations Course which involves DeepRacer. I’m expecting that quite a few of them will stay to race in the Pro Division next month. We are seeing increasingly difficult and challenging questions in our community and we need to cool some heads as well as it’s easy to get overly enthusiastic about DeepRacer League. Keep it cool, play it smart.

Current number of racers to qualify into the Pro Division is 192. This means that the last one to qualify is zoomzoomzoom with a time of 02:34.367. This is an improvement of 0.56 seconds compared to abw who were 130th last week. Abw have not improved their time and are now in 200th spot.

Top 25 submissions
Update time: 2021-07-24 23:00:00 UTC
Comparing to 2021-07-17 13:00:00 UTC

Racers: 25
Total submission attempts: 1458 (+733)
Best total time: 01:07.131 (-00:14.076) by Janesoki
Mean total time: 01:29.770400 (-00:10.927480)
Best lap: 00:21.781 (-00:02.675) by Janesoki
Mean best lap: 00:28.557280 (-00:02.925240)
Mean average lap: 00:29.923080 (-00:03.642520)

Best total time this week has gone down by over 14 seconds and mean improvement in top 25 was almost 11 seconds. mikeRacer who was leading last week is now in sixth. Janesoki has managed to eliminate the reset and cut down on a few more seconds, Boltronic in the fourth who has been leading for a long time has gained a single reset but still managed to cut five seconds from the total time. Looks like the model is pushed to the edge of stability with its action space.

We even have Udacity students in the top 10. Open Division has never been so active since it was announced. I can see difficult questions, some local training stacks being set up. The spirits are high, learning abundant and DeepRacers go crazy in simulation.

Rank Racer Total Time Avg Lap Best Lap Resets Submissions
1 (⬆3) Janesoki 01:07.131 (-00:19.885) 00:22.377 (-00:06.628) 00:21.781 (-00:02.675) 0 (-1) 257 (+237)
2 (⬆9) Bmth-3-Sens-AI 01:14.924 (-00:24.240) 00:24.974 (-00:08.080) 00:24.723 (-00:06.781) 0 (-4) 132 (+96)
3 (–) SSC-Tuatara 01:16.736 (–:–.—) 00:25.578 (–:–.—) 00:25.330 (–:–.—) 0 (–) 29 (–)
4 (⬇1) Boltronic 01:17.300 (-00:05.157) 00:25.766 (-00:01.719) 00:23.057 (-00:04.292) 1 (+1) 4 (+3)
5 (⬆4) Yuliya 01:17.960 (-00:16.132) 00:25.986 (-00:05.378) 00:24.132 (-00:06.793) 1 (+1) 515 (+71)
6 (⬇5) mikeRacer 01:21.207 (-00:00.000) 00:27.069 (-00:00.000) 00:26.935 (-00:00.000) 0 (-0) 4 (-0)
7 (⬇5) Webstar34 01:21.270 (-00:00.000) 00:27.090 (-00:00.000) 00:26.328 (-00:00.000) 0 (-0) 5 (-0)
8 (–) uoa 01:23.848 (–:–.—) 00:27.949 (–:–.—) 00:27.268 (–:–.—) 0 (–) 7 (–)
9 (–) SLC 01:29.962 (–:–.—) 00:29.987 (–:–.—) 00:28.862 (–:–.—) 0 (–) 7 (–)
10 (⬇5) pa-ss 01:30.308 (-00:00.000) 00:30.102 (-00:00.000) 00:29.540 (-00:00.000) 0 (-0) 12 (+3)
11 (–) Weelz 01:30.476 (–:–.—) 00:30.158 (–:–.—) 00:25.474 (–:–.—) 2 (–) 10 (–)
12 (⬇6) baby-driver 01:30.937 (-00:00.000) 00:30.312 (-00:00.000) 00:29.863 (-00:00.000) 0 (-0) 23 (+8)
13 (⬇6) RMIT-badger 01:32.249 (-00:00.993) 00:30.749 (-00:00.331) 00:30.212 (-00:00.331) 0 (-0) 37 (+15)
14 (⬇6) AE86-AAW 01:33.816 (-00:00.000) 00:31.272 (-00:00.000) 00:30.543 (-00:00.000) 0 (-0) 6 (-0)
15 (⬆22) Racer-1313 01:33.997 (-00:34.275) 00:31.332 (-00:11.425) 00:29.991 (-00:07.742) 0 (-6) 47 (+33)
16 (⬆8) Xavier 01:34.068 (-00:18.721) 00:31.356 (-00:06.240) 00:29.406 (-00:03.925) 1 (-4) 183 (+145)
17 (⬇7) Dang 01:34.733 (-00:00.000) 00:31.577 (-00:00.000) 00:28.810 (-00:00.000) 2 (-0) 16 (+3)
18 (–) Weeelz 01:36.167 (–:–.—) 00:32.055 (–:–.—) 00:31.525 (–:–.—) 3 (–) 9 (–)
19 (↔0) robracer 01:36.791 (-00:14.028) 00:32.263 (-00:04.676) 00:31.402 (-00:01.810) 0 (-1) 8 (+1)
20 (⬆46) Ben-URCAlpha 01:37.597 (-00:43.447) 00:32.532 (-00:14.482) 00:29.869 (-00:16.925) 1 (+1) 87 (+81)
21 (–) CGS-on-your-12-hopefully 01:37.632 (–:–.—) 00:32.544 (–:–.—) 00:32.277 (–:–.—) 0 (–) 24 (–)
22 (⬇10) Hennessey-Venom 01:40.033 (-00:00.000) 00:33.344 (-00:00.000) 00:31.136 (-00:00.000) 4 (-0) 14 (-0)
23 (⬇10) LuckyMan 01:40.675 (-00:00.000) 00:33.558 (-00:00.000) 00:31.402 (-00:00.000) 1 (-0) 5 (-0)
24 (⬇10) R1999TDR 01:41.589 (-00:00.000) 00:33.863 (-00:00.000) 00:33.797 (-00:00.000) 0 (-0) 11 (+1)
25 (–) Toby1 01:42.854 (–:–.—) 00:34.284 (–:–.—) 00:30.269 (–:–.—) 2 (–) 6 (–)

Make sure to mark 12th of August in your calendar to join us on Twitch watching the Finale Race. Before that happens however, expect an “End of Month” update on our blog. Thank you for your time, see you next week!

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.