DARPA přišla s novou „výzvou”/soutěží: DARPA Triage Challenge. Triage znamená třídění raněných u hromadných neštěstí, přírodních katastrof nebo na bojišti. V zásadě jde o to, že podle úrovně zranění šance na přežití rychle klesá. Mluvíme o zhruba o prvních 15 minutách. Naděje, že v tomto čase bude dostatek kvalifikovaných doktorů je mizivá. Tak nastupují roboti a AI. Update: 27/7/2025 — DTC Systems 2 - W30 (zbývá 9 týdnů) - grid map, pose, audio

• https://triagechallenge.darpa.mil/ — oficiální stránky soutěže

Obsah

• 241213 - Pravidla - Event 2
• 250103 - NDAA 2025
• 250319 - Virtuál zrušen
• 250629 - DTC Systems 2 - Robotika
• 250708 - DTC Systems 2 - W27 (zbývá 12 týdnů) - skryté předpoklady
• 250713 - DTC Systems 2 - W28 (zbývá 11 týdnů) - Helena, Audio, grid_map, ...
• 250720 - DTC Systems 2 - W29 (zbývá 10 týdnů) - nedotažené konce
• 250727 - DTC Systems 2 - W30 (zbývá 9 týdnů) - grid map, pose, audio

Blog

13. prosince 2024 — Pravidla - Event 2

• System Competition
• Data Competition
• (pravidla pro Virtual Competition ještě nějsou k dispozici)

teleoperace NENÍ povolena

tabulka cen

Workshop 2

• Event Qualification: 12/5/2024 - 1/5/2025 (konec za 3 týdny!)
• Event Date: 3/9/2025 - 3/15/2025 (březen!)

3. ledna 2025 — NDAA 2025

Key Restrictions on UGVs

• The FY25 NDAA mandates a study and report on the Department of Defense's (DoD) use of Chinese-made UGVs. Following this assessment, it prohibits the procurement and operation of covered UGVs from foreign adversaries, particularly those identified as posing cybersecurity risks [1][2][3].

• The Act requires the DoD to disassemble and analyze components of commercially available UGVs from foreign adversaries to identify supply chain risks. This includes a comprehensive evaluation of UGVs to enhance the resiliency of the domestic supply chain [3][4].

• A provision directs national security agencies to assess drones from companies like DJI and Autel Robotics, potentially leading to their inclusion on the Federal Communications Commission’s (FCC) Covered List if deemed a national security risk. This would prevent these companies from obtaining FCC authorization for new models [1][5].

• Beginning one year after the enactment of the NDAA, the DoD is prohibited from procuring or operating UGVs manufactured by entities in foreign adversary countries. This aims to phase out reliance on potentially insecure technologies over time [3][4].

• The legislation emphasizes enhanced oversight regarding the procurement processes for UGVs and other unmanned systems, ensuring that any acquisitions align with national security interests [1][2].

• [1] https://www.auvsi.org/auvsi-applauds-fy-2025-ndaa-support-autonomy-uncrewed-systems-supply-chain-security
• [2] https://www.auvsi.org/auvsi-calls-congress-ban-dod-use-unsecure-autonomous-vehicles-national-defense-authorization-act
• [3] https://www.wiley.law/alert-NDAA-Provisions-Impacting-Governments-Contractors-and-Their-Supply-Chains
• [4] https://ucdrones.github.io/regulatory-updates.html
• [5] https://dronelife.com/2024/12/19/ndaa-fy25-key-takeaways-for-the-drone-industry/

19. března 2025 — Virtuál zrušen

29. června 2025 — DTC Systems 2 - Robotika

(osgar) md@md-ThinkPad-P50:~/git/osgar-apps/dtc-systems$ python report_dtc.py Gas 0 0 0
Reporting: Gas
Get Status
b'{"clock":6.889369,"run":"Run2","team":"Test Team","user":"testuser"}'

Report {'casualty_id': 1, 'team': 'Robotika', 'system': 'Matty M01', 'location': 
  {'latitude': 10, 'longitude': 20, 'time_ago': 1}}
b'{"run":"Run2","team":"Test Team","user":"testuser","system":"Matty M01","clock":8.900963,
"report_id":"e2ca8f47-da17-4f9b-b258da52389ec7dd","report_timestamp":"2025-06-26T13:33:44.083908",
"report_status":"accepted","casualty_id":1}'
Traceback (most recent call last):
  File "report_dtc.py", line 128, in 
    print(score(artf_type, args.x, args.y, args.z))
  File "report_dtc.py", line 104, in score
    report_artf(artf_type, x, y, z)
  File "report_dtc.py", line 63, in report_artf
    assert response.status_code == 201, response.status_code
AssertionError: 200

• Robotem rovně 2025 - test neuronky, baterky, drncání, AP přes mobil
• Field Robot Event 2025 (Itálie) - horko, prach, stres, noční jízda
• RoboOrienteering 2025 - neznámý terén, GPS, hledání obětí (kužely), výpadek napájení, hodinová jízda

8. července 2025 — DTC Systems 2 - W27 (zbývá 12 týdnů) - skryté předpoklady

Zombie s lidarovým vzorem

13. července 2025 — DTC Systems 2 - W28 (zbývá 11 týdnů) - Helena, Audio, grid_map, …

Demo M02+M03

from threading import Thread

class Node(Thread):

m03-dtc-day-250713_170143.log
    140,168,400 100%    5.90MB/s    0:00:22 (xfr#1, to-chk=16/26)
m03-matty-followpath-250713_163357.log
         20,199 100%   31.87kB/s    0:00:00 (xfr#2, to-chk=11/26)
m03-matty-followpath-250713_163547.log
    808,649,552 100%    5.69MB/s    0:02:15 (xfr#3, to-chk=10/26)
m03-matty-followpath-250713_163908.log
  1,285,921,633 100%    5.94MB/s    0:03:26 (xfr#4, to-chk=9/26)
m03-matty-followpath-250713_164134.log

• mobility: https://youtu.be/MRd-9Yo-vDE
• gravel: https://youtu.be/R8eb7rZRSk0
• forest: https://youtu.be/xvHPZC8R630

20. července 2025 — DTC Systems 2 - W29 (zbývá 10 týdnů) - nedotažené konce

DTC User Interface

Audio

schema zapojení audio

ALSA lib confmisc.c:855:(parse_card) cannot find card '0'
ALSA lib conf.c:5178:(_snd_config_evaluate) function snd_func_card_inum returned error: No such file or directory
ALSA lib confmisc.c:422:(snd_func_concat) error evaluating strings

Pose Estimation

Noční vidění

27. července 2025 — DTC Systems 2 - W30 (zbývá 9 týdnů) - grid map, pose, audio

Stonehenge

OAK-D depth data

• 1 (Batch Size): This indicates that the model processed a single image at a time. If you were processing multiple images simultaneously, this number would change to reflect the number of images in your batch .
• 56 (Features per Detection): This dimension contains the information for each detected object, specifically designed for pose estimation . It's typically broken down as follows:
• First 5 values: These represent the bounding box coordinates and confidence score for a detected object (e.g., a person). This usually includes x, y, width, height, and confidence .
• Remaining 51 values (17 keypoints x 3 values per keypoint): These are the keypoints for pose estimation. YOLO models trained on datasets like COCO typically detect 17 keypoints (e.g., nose, eyes, shoulders, elbows, wrists, hips, knees, ankles) . Each keypoint is represented by three values: its x coordinate, y coordinate, and a visibility score or confidence . So, 17 keypoints * 3 values/keypoint = 51 values.
• 8400 (Number of Potential Detections/Anchors): This is the total number of anchor boxes or potential detections the model considers across the input image. Each of these 8400 entries contains the 56 features described above .

• Non-Maximum Suppression (NMS): To eliminate redundant bounding box detections that overlap significantly.
• Confidence Thresholding: To filter out detections with low confidence scores.
• Converting Coordinates: The raw keypoint coordinates might be normalized or relative to the bounding box, and often need to be converted to actual image coordinates for visualization or further use .