트윅 tweak
트윅 tweak
  • 43
  • 23 980
Exploring the NVIDIA Omniverse World, Including IsaacSim
IsaacSim을 포함하는 NVIDIA의 Omniverse 세계관 알아보기.
핵심 플랫폼: (Nucleus, Connect, Kit, Simulation and RTX Renderer)
00:00 인트로 (intro)
01:54 Use cases (Digital Twins)
10:55 Use cases (Robotics Simulation)
14:27 Use cases (Synthetic Data Generation)
17:50 Use cases (Virtual Factory)
18:44 OpenUSD
21:08 Omniverse Core Platforms
22:54 Nucleus
24:49 Connect
27:28 Kit
34:30 Simulation
35:27 RTX Renderer
42:20 Omniverse Summary
43:36 Omniverse Conclusion
45:26 IsaacSim관련 채용 공고(Job Openings Related to IsaacSim)
47:25 Omniverse Launcher 설치
49:00 Omniverse Launcher 실행
54:52 아웃트로 (outro)
Links
* NVIDIA omniverse main:
www.nvidia.com/en-us/omniverse/
* Omniverse Platform overview:
docs.omniverse.nvidia.com/platform/latest/index.html
* Omniverse documentation
docs.omniverse.nvidia.com/index.html
* Omniverse Kit template video:
th-cam.com/video/asuBZhSHvL4/w-d-xo.html
* Omniverse Download
www.nvidia.com/en-in/omniverse/download/
มุมมอง: 609

วีดีโอ

IsaacGym Franka Cube Stack Task, Reinforcement Learning for Stacking Cubes with a Robotic Arm
มุมมอง 431หลายเดือนก่อน
IsaacGym을 활용하여 Franka Panda 로봇으로 큐브 쌓는 인공지능 강화학습 하기 Github: github.com/tweak4u/IsaacGymEnvsTwk 00:00 오프닝 (opening) 00:30 tweak 00:35 태스크 설명(task description) 01:37 깃허브 (github) 02:16 설정 파일(config files) 10:43 관절 제어 모드와 PD 컨트롤(joint control mode and PD control) 30:35 관절 제어 모드 초기화(joint control mode initialization) 33:35 OSC 모드에서 상태와 행동(state and action in OSC mode) 45:19 Torque 모드에서 상태와 행동(state...
Isaac Sim Preview. 왜 Isaac Sim을 사용해야 하나?
มุมมอง 8802 หลายเดือนก่อน
Isaac Sim 슬쩍 살펴보고, 왜 Isaac Sim을 사용해야 할까? 그 이유 알아보기
IsaacGym Ball Balance 태스크, 테이블을 움직여 공의 중심을 잡는 로봇 강화학습
มุมมอง 3356 หลายเดือนก่อน
하늘에서 떨어지는 공을 받아 테이블 위에서 떨어트리지 않고 유지시키는 밸런스 봇 강화학습하기 00:00 오프닝 (opening) 00:30 tweak 00:35 태스크 소개(task introduction) 01:46 밸런스봇 소개(balance bot introduction) 11:40 파일 및 설정 (files and configs) 16:40 코드를 이용한 밸런스봇 에이전트 파일 생성 (creating a balance bot agent file using code) 24:39 상태(state) 34:03 행동(action) 39:32 보상(reward) 43:04 리셋 조건(reset condition) 44:45 정책 및 행동 추론(policy and action inference) 51:16...
Designing an Adapter to Combine Allegro Hand with UR3
มุมมอง 3527 หลายเดือนก่อน
알레그로 핸드(Allegro hand)와 UR3를 결합시키기 위해 알루미늄 재질의 어댑터를 금속가공하기 위한 설계 방법을 다루는 영상입니다. 어댑터 설계 자료 github: github.com/tweak4u/AllegroHand-UR3-adaptor 알레그로 핸드(Allegro hand) 매뉴얼 주소: wiki.wonikrobotics.com/AllegroHandWiki/index.php/File:V4_AllegroHandUsersManual_1.1.pdf UR3 매뉴얼 주소: www.universal-robots.com/download/?option=16330&query= 00:00 오프닝 00:10 Tweak 00:16 인트로 01:22 UR3 Flange 07:14 알레그로 핸드 mount blo...
구독자 100명 달성 감사인사
มุมมอง 848 หลายเดือนก่อน
2023년 11월 25일 구독자 100명 달성 (첫 영상 업로드 후 172일)
Isaac Gym Humanoid Task: Reinforcement Learning for a Running Humanoid
มุมมอง 9538 หลายเดือนก่อน
강화학습을 이용한 휴머노이드 달리기 관련 논문: Emergence of Locomotion Behaviours in Rich Environments (Deepmind, arXiv, 2017) 관련 영상: th-cam.com/video/hx_bgoTF7bs/w-d-xo.html 00:00 오프닝 (opening) 00:50 tweak 00:55 태스크 목표(task goal) 02:31 파일 및 설정 (files and configs) 06:19 상태소개(state-intro) 06:50 휴머노이드 에이전트(humanoid agent) 19:37 상태(state) 55:42 행동(action) 01:03:04 보상(reward) 01:21:25 리셋 조건(reset condition) 01:22:53 정...
Isaac Gym Franka Cabinet Task: Reinforcement Learning for Robot Arm Drawer Opening (reupload)
มุมมอง 1.2K10 หลายเดือนก่อน
Franka Panda 로봇으로 서랍 여는 인공지능 강화학습하기 00:00 오프닝 (opening) 00:45 tweak 00:50 태스크 목표(task goal) 06:43 파일 (files) 13:37 상태(state) 24:47 행동(action) 32:28 정책(policy) 35:52 보상 인트로(reward-intro) 39:06 변환(transforms) 54:26 거리-회전 보상 (distance, rotation reward) 01:04:40 주변-손가락 거리 보상 (around, finger distance reward) 01:10:03 오픈-패널티 보상 (open, penalty reward) 01:18:54 보상 보너스 (reward bonus) 01:21:44 방어용 보상 (re...
Analyzing the IsaacGym Cartpole Reinforcement Learning Example. (Utilizing the RL Games Library)
มุมมอง 82511 หลายเดือนก่อน
IsaacGym 벤치마크 환경에서 Cartpole 태스크 강화학습 예제 분석 * train.py 코드 분석 * Cartpole MDP 분석 * PPO (proximal policy optimization) 00:00 오프닝 00:37 tweak 00:41 train.py 14:48 상태(state) / 행동(action) 23:43 정책(policy) 37:16 보상(reward) 49:06 RL & PPO 54:38 실습 01:12:50 클로징
Analysis of the Folder Structure of the Isaac Gym Reinforcement Learning Benchmark Environment
มุมมอง 36711 หลายเดือนก่อน
Isaac Gym 강화학습 벤치마크 환경의 폴더 구조를 분석하는 영상 * IsaacGymEnvs, asset, docs, isaacgymenvs * isaacgymenvs.egg-info 00:00 오프닝 00:24 폴더 구조 설명 12:02 실습 23:13 클로징
Isaac Gym RL Benchmark Environments Installation and Analysis
มุมมอง 94011 หลายเดือนก่อน
Isaac Gym 벤치마크 환경의 설치 방법 및 인수 설정 방법 분석 00:00 Intro 00:24 Github access 10:39 Creating an environment 21:11 Running the benchmarks 28:22 Loading trained models, checkpoints 39:49 Configuration and command line arguments 41:27 Tasks 43:05 Domain Randomization 44:54 Reproducibility and Determinism 46:42 Multi-GPU Training 49:39 Population Based Training 51:25 WandB support 53:23 Capture videos 57:...
Isaac Gym Introduction to Benchmark Environments for Reinforcement Learning
มุมมอง 99711 หลายเดือนก่อน
Isaac Gym에서 강화학습을 하기 위한 벤치마크 환경을 소개하는 예제 * (Partially Observable) Markov Decision Process * NeurIPS 2021 Paper: Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning
IsaacGym Example Franka Nut Bolt IK OSC [Basic] Task planning of robotic bolt-nut assembly using FSM
มุมมอง 351ปีที่แล้ว
유한 상태 머신(Finite State Machine)을 이용한 로봇 볼트/너트 조립 태스크 플래닝 * Finite state machine * Robot task planning
IsaacGym Example Convex Decomposition [Basic] Collision Calculation, Between Precision and Lightness
มุมมอง 333ปีที่แล้ว
Convex Decomposition을 이용한 충돌 계산용 mesh 모델 생성 과정에 대한 예제 * Convex Hull * V-HACD and its parameters
Isaac Gym example Spherical Joint [Basic] Spherical joint control using exponential coordinate
มุมมอง 213ปีที่แล้ว
Exponential cooridnate을 이용해서 spherical joint를 제어하는 예제 * Exponential Coordinate, Screw axis, Twist * Lie Group, Lie Algebra, Rodrigues’ formula, etc. * Quaternion to exponential coordinate
Isaac Gym example Terrain creation [Basic] Creating desired terrain from code
มุมมอง 473ปีที่แล้ว
Isaac Gym example Terrain creation [Basic] Creating desired terrain from code
Isaac Gym example actor scaling [Basic] Adjusting the size of the actor assets freely
มุมมอง 180ปีที่แล้ว
Isaac Gym example actor scaling [Basic] Adjusting the size of the actor assets freely
Isaac Gym example, Graphics materials [Basic] Graphics processing steps when loading an asset
มุมมอง 253ปีที่แล้ว
Isaac Gym example, Graphics materials [Basic] Graphics processing steps when loading an asset
An engineer YouTuber Thanks for reaching 10 subscribers
มุมมอง 108ปีที่แล้ว
An engineer TH-camr Thanks for reaching 10 subscribers
Isaac Gym example Graphics up-axis [Basic] Configuration of the reference coordinate system
มุมมอง 147ปีที่แล้ว
Isaac Gym example Graphics up-axis [Basic] Configuration of the reference coordinate system
Isaac Gym example multiple cameras [Basic] Acquiring images from multiple cameras(feat: view matrix)
มุมมอง 363ปีที่แล้ว
Isaac Gym example multiple cameras [Basic] Acquiring images from multiple cameras(feat: view matrix)
Isaac Gym example, apply forces at position [Basic] applying forces to the desired position
มุมมอง 202ปีที่แล้ว
Isaac Gym example, apply forces at position [Basic] applying forces to the desired position
Isaac Gym example apply forces [Basic] applying forces/torques to objects!
มุมมอง 209ปีที่แล้ว
Isaac Gym example apply forces [Basic] applying forces/torques to objects!
Isaac Gym Example: Franka Operational Space Control [Basic] Force Control Using Inverse Dynamics
มุมมอง 418ปีที่แล้ว
Isaac Gym Example: Franka Operational Space Control [Basic] Force Control Using Inverse Dynamics
Isaac Gym example franka ik picking [Basic] Cube picking using inverse kinematics
มุมมอง 482ปีที่แล้ว
Isaac Gym example franka ik picking [Basic] Cube picking using inverse kinematics
Isaac Gym example pytorch interop [Basic] utilizing pytorch tensor
มุมมอง 218ปีที่แล้ว
Isaac Gym example pytorch interop [Basic] utilizing pytorch tensor
Isaac Gym example Kuka Bin [Basic], Preparing for bin picking with Kuka robot and Allegro hand.
มุมมอง 296ปีที่แล้ว
Isaac Gym example Kuka Bin [Basic], Preparing for bin picking with Kuka robot and Allegro hand.
Isaac Gym example large mass ratio [Basic] simulation of object density, volume, and weight.
มุมมอง 177ปีที่แล้ว
Isaac Gym example large mass ratio [Basic] simulation of object density, volume, and weight.
Isaac Gym example projectiles [Basic] Shooting cube missiles using a keyboard and mouse
มุมมอง 198ปีที่แล้ว
Isaac Gym example projectiles [Basic] Shooting cube missiles using a keyboard and mouse
Isaac Gym example visualize transforms [Basic], let's visualize transformations in real-time
มุมมอง 197ปีที่แล้ว
Isaac Gym example visualize transforms [Basic], let's visualize transformations in real-time

ความคิดเห็น

  • @DyK-pz5eq
    @DyK-pz5eq 5 วันที่ผ่านมา

    안녕하세요. 학생들 노트북(내장그래픽)수준에서도 해당 알고리듬을 구현해볼 수 있을까요?

    • @tweak-cd6ss
      @tweak-cd6ss 4 วันที่ผ่านมา

      안녕하세요. 학습을 gpu가 아닌 cpu를 사용하면 가능합니다. 다만 학습을 위한 계산 시간이 훨씬 오래 걸리는 점은 어느 정도 감내하셔야 합니다.

  • @user-ji4mi1gi3z
    @user-ji4mi1gi3z 14 วันที่ผ่านมา

    아 로봇 시뮬레이션 강화학습 보상함수를 이렇게 설계하는 거군요... 쉽지 않네요 ㅎㅎ 돈받고 팔아야 될 내용이네요 ㅎㅎ 감사합니다 후원 열어놓으시면 좋을거 같네요~ 로봇팔+강화학습 issacGym issacSim 공부를 하고싶은데... 선생님 강의 외에 혹시 책이나 모아져 있는 자료가 있을까요? 웹개발 게임개발은 조금 아는데 로봇개발은 아예 처음이라서요 ㅎㅎ; 혹시 ros2? 호환되는 실제 로봇팔 제품도 추천하실만한게 있으시면 알려주시면 감사하겠습니다 혹시... 유료 컨설팅이나 개발대행도 하실까요?

    • @tweak-cd6ss
      @tweak-cd6ss 13 วันที่ผ่านมา

      안녕하세요. IsaacGym은 사실 자료가 많이 없어서 자체 document나 구글링에 많이 의존을 해야 하는 상황입니다. 강화학습은 자료가 워낙 많아서 검색을 잘 해보시거나 아니면 david silver 강의 추천드립니다. 로봇팔 관련 공부는 조금 방대한데.. Craig책 introduction to robotics나, modern robotics 책도 볼만한 것 같습니다. IsaacSim은 아직까지 NVIDIA 공식 document가 가장 나은 것 같습니다. ros2는 대부분의 로봇에 공식/비공식 적으로 많이 호환됩니다. 많이 사용하는 로봇팔은 Franka panda, UR시리즈 로봇 정도 보시면 무난할 것 같습니다. 유료 컨설팅/개발 대행은 하고 있지 않습니다.

  • @user-fr7wz1xw8b
    @user-fr7wz1xw8b 17 วันที่ผ่านมา

    안녕하세요 좋은 영상들 너무 감사드립니다. 아이작 짐 자료가 많이 없는데, 트윅님의 영상이 매우 큰 도움이 되고 있어요! 현재 어려움을 겪는 부분이 있는데, 아무리 구글링을 해봐도 해결책을 찾지 못해 혹시 도움 받을 수 있을까 하여 질문드립니다. 아이작 짐에서 dof가 없는 rigid body 두 개를 로드한 후, 한 물체가 다른 물체를 피해서 움직이는 position(경로)를 학습시키려고 합니다. set_actor_root_state_tensor 함수를 사용하여 pos을 지정해주고 있는데요, 문제는 이 함수가 에셋의 충돌을 무시하고 pos을 지정해서 두 에셋이 겹치는 현상이 생깁니다. 이 함수를 사용하지 않으면 충돌이 정상적으로 발생합니다. 혹시 에셋이 겹치는 현상을 해결할 수 있는 설정 옵션이나, 다른 함수를 사용하는 방법이 있을까 하여 질문드립니다. 감사합니다.

    • @tweak-cd6ss
      @tweak-cd6ss 17 วันที่ผ่านมา

      안녕하세요. set_actor_root_state_tensor는 rigid body의 상태를 강제로 직접 조정하는 함수로써 주로 reset을 할 때 object를 특정 위치/자세로 teleport 시킬 때 사용합니다. position을 언급한 것으로 보아 policy의 action이 대상 object state의 위치(x, y, z)를 직접 조정하는 것으로 보이는데요, 이렇게 되면 만약 충돌이 발생한 시점에서 다음 simulation step에 원래 서로 튕기는 모션이 발생해야 하는데 특정 위치로 강제하기 때문에 정상적인 충돌이 발생하지 않을 수도 있을 것 같네요. 만약 제가 이해한 상황이 맞으면, 제가 생각하는 해결책은 policy의 action을 대상 object의 위치 말고 속도와 연결시키거나, apply_rigid_body_force_tensors함수를 이용하여 물체에 힘을 가하는 방식으로 해결할 수 있을 것 같습니다. 물체 위치를 직접적으로 teleport시키는 방식의 action은 자연스럽지 않아 보입니다.

    • @user-fr7wz1xw8b
      @user-fr7wz1xw8b 17 วันที่ผ่านมา

      ​@@tweak-cd6ss 답변 감사드립니다. 제가 원하는 것은 물체의 pos를 학습시키고, 해당 pos로 로봇의 엔드이펙터를 pd control하는 것이었습니다. 그래서 경로에 대한 pos값이 필요한데, 그러면 학습은 velocity나 force에 대해서 하고, 찾아낸 최적 policy를 통해 움직이는 물체의 pos을 따로 저장하여 pd control에 사용해도 유효한 결과를 얻을 수 있을까요? agent는 vel이나 force에 대한 최적값을 학습하는데 정작 사용하는 값은 pos라 해당 경로가 최적값이 아닐 수 있겠다는 점 때문에 고민이 됩니다.

    • @tweak-cd6ss
      @tweak-cd6ss 17 วันที่ผ่านมา

      로봇 EE를 가이드 해주는 pointer의 최적 경로를 학습하는 걸로 이해되는데, 가이더가 단순 pointer 역할만 하는 것이라면 물리 속성의 영향력을 최소화 시킬 수 있도록 가이더의 속성값을 설정해주면 어느 정도는 pos에 준하는 성능이 나올 것 같습니다. 즉, 가이더의 gravity를 disable해주고, density 값을 작게 설정한 뒤, velocity/force로 움직이면 충돌 검출을 하면서 physical effect는 최소화하고, 경로 찾기 문제에만 focus할 수 있는 가이더가 학습될 것 같습니다.

    • @user-fr7wz1xw8b
      @user-fr7wz1xw8b 17 วันที่ผ่านมา

      @@tweak-cd6ss 답변 정말 감사드립니다. 조언해주신 내용 고려하여 수정해보겠습니다. 항상 좋은 영상에 감사드립니다.

  • @hubertkim7021
    @hubertkim7021 20 วันที่ผ่านมา

    자세한 영상 감사합니다!! 궁금한 점이 한가지 있는데요. 혹시 설명 해 주실 수 있을까요? 리워드 설명하실 때 보면 EE가 접근하고, 문고리를 사이에 두고 그리퍼 위치하고, drawer를 당기는 리워드는 있는데, 그 drawer를 당기기 전에 grip 실행하는 리워드는 못 봤는데 혹시 그 부분은 어느 항목에 들어있는지 알 수 있을까요?

    • @tweak-cd6ss
      @tweak-cd6ss 20 วันที่ผ่านมา

      일단 dist, rotation, around_handle reward에 의해 그리퍼를 문고리에 위치하도록 만들어주면 그 다음에 영향을 미치는 것은 finger distance reward입니다. 이것에 의해 grip이 발생하게 되는데, left finger의 z축 값과 문고리 좌표의 z축 값 사이의 거리가 작을 수록 높은 reward를 얻도록 설계되어 있고, right finger도 마찬가지 입니다. 다만, 이 reward의 최종 계산 값이 적용되기 까지 두 개의 조건을 만족해야 하는데, left finger의 z축이 문고리 좌표 z축 위치보다 높게 있어야 하고, right finger의 z축은 문고리 z축 위치보다 낮게 있어야 조건이 성립되고, 그 이후부터는 finger와 문고리 사이의 거리가 가까워야 높은 reward를 얻도록 설계되어 있습니다. 01:04:40 이 부분을 참고하시면 되겠습니다. 참고로 finger distance reward의 0.04(meter)라는 값은 left/right finger의 각각의 가동 범위입니다.

    • @hubertkim7021
      @hubertkim7021 20 วันที่ผ่านมา

      ⁠@@tweak-cd6ss빠른 답변 정말 감사합니다. 이해가 되었습니다. 기본적으로 rewarding에서 접근 후 손가락으로 잡는 방식으로 유도를 하는 군요. 그렇다면, 이전에 기본편에서 state machine 설명하셨는데, 그걸 여기에 적용할 수 있을것 같은데, 어떻게 생각하시는지 여쭤봐도 될까요? 접근하는 것과 그립, 풀링을 각각 다른 state로 설정하면 cascaded if의 방향으로 리워딩을 안해도 되지 않을가 생각되어서요.

    • @tweak-cd6ss
      @tweak-cd6ss 19 วันที่ผ่านมา

      일단 FSM은 적용이 가능할텐데, 적용한다면 policy 학습 과정에서 search space가 기존의 continuous space에서 제한된 state로 discrete하게 줄어서 빠른 학습은 가능할 것 같습니다(마치 grid world 문제와 같이... towardsdatascience.com/reinforcement-learning-implement-grid-world-from-scratch-c5963765ebff). 이 경우 reward의 형태가 서랍이 열리면 +1, 그렇지 않은 경우 0으로 아주 간단(sparse)하게 설정할 수 있을 것 같긴 합니다(state 사이의 motion은 algorithmic planning으로 수행). 다만 manual한 state modeling 및 그에 따른 정교한 goal pose설정이 필요하고, 태스크가 거대해질 수록 확장성이 제한될 것 같습니다. 또한 policy가 주어진 태스크를 수행하기 위한 global optimal에 가까운 solution을 찾을 확률이 줄어들 수도 있을 것 같습니다. 로봇의 motion이 사용자가 정의한 state machine내의 motion 분포로 제한될 테니까요(태스크가 어느 정도 단순한 경우). 이 외에도 replanning과 같은 동적 상황에서의 작업 수행 능력은 아무래도 반응성 등이 떨어질 것으로 보입니다. 태스크가 복잡한 경우에는 오히려 단순 reward shaping만으로는 solution 자체를 찾을 확률이 FSM에 비해 현저히 떨어지는데, 그에 따라 low-high level policy로 나눠서 low-level policy는 imitation learning으로 pre-train, high-level policy는 sparse reward로 푸는 방법도 있습니다.

    • @hubertkim7021
      @hubertkim7021 18 วันที่ผ่านมา

      @@tweak-cd6ss 와 깊은 insight 감사합니다. 제가 따라가야 할 부분이 많군요. 저는 이 예제를 돌려보면서 rewarding 으로 손잡이에 접근시키기까지도 training iteration이 좀 오래걸리는 구나 싶어서 초반state는 아예 graph search algorithm으로 접근시키고, 이후 좀 복잡한 task인 'grip' 부분만 따로 reward로 학습시키고, 'pull' 같은 부분은 말씀하신 discrete task로 하면 되지 않을까 생각했거든요. 늘 귀한 답변 감사드립니다. 말씀해주신 키워드로도 좀 더 알아보겠습니다.

  • @kyw0615
    @kyw0615 21 วันที่ผ่านมา

    multi gpu 쓸 때 질문이 있는데요. 대부분의 메인보드가 두 개의 gpu를 사용하면 pcie 배속을 하나는 16x랑 하나는 4x로 지원을 하더라고요. 다른 경우는 8x 8x정도가 있고요. 그리고 워크스테이션용 cpu가 아닌 이상 16x 16x를 다 받아줄 수도 없잖아요. 말이 좀 길었는데요. 궁금한 건 isaac gym의 경우 pcie 배속에 성능이 얼마나 영향을 받나요? 이번에 큰 맘 먹고 ai용 pc 구매할 예정이라 질문해봤습니다. 아 그리고 영상 잘 봤습니다. 유익하네요.

    • @tweak-cd6ss
      @tweak-cd6ss 20 วันที่ผ่านมา

      pcie 배속을 비교/변경하며 isaacgym에서 multi-gpu를 이용한 벤치마크 실험을 해본적은 없습니다. 제가 사용하는 PC의 메인보드는 두 개 GPU활용 시 x8/x8로 제한되는 모델인데요, 질문에 정확히 답변이 되는 실험은 아니겠으나 Humanoid 태스크를 1000epoch 학습 시킴에 있어서 single-gpu는 약 523초, multi-gpu(RTX6000 ada 2개)는 약 552초 가량 소요되어 overhead로 인하여 multi-gpu가 학습에 더 많은 시간이 소요되었습니다(state기반, not image). pcie 배속에 따른 성능 측정을 위해서는 x4레인으로 강제로 다운 시켜서 multi-gpu 학습 후 비교하면 될 것 같은데, 다른 PC는 사양이 달라서 정확한 비교가 안될 것 같고, 지금 PC는 재부팅이 불가능한 상황이라 더 이상의 실험은 어렵네요... 개인적인 생각으로는 isaacgym과 연동하여 학습할 모델이 큰 이미지를 다루고 거대한 모델을 다루는 경우, x8/x8, multi-gpu 조합이 유리해 보이고, 그게 아니라면 x4/x16-single-gpu로 셋팅하여 first gpu는 display/evaluation용, second gpu는 x16 slot에 물려서 학습 전용으로 구성하는 방법도 좋아 보입니다. 아무래도 x4 <-> x16간에 레인 차에 따른 multi-gpu overhead 및 속도 저하가 예상되기 때문입니다. 만약 저라면 그냥 x8/x8로 갈 것 같습니다~

  • @user-bh9dj7jd5r
    @user-bh9dj7jd5r 23 วันที่ผ่านมา

    ros2를 먼저 공부하고 isaac sim을 공부해야겠군요

  • @user-bh9dj7jd5r
    @user-bh9dj7jd5r 23 วันที่ผ่านมา

    isaacsim 공부 하는데 도움 되는 곳이 있을까요?

    • @tweak-cd6ss
      @tweak-cd6ss 23 วันที่ผ่านมา

      한국 유튜브 채널도 간혹 보이기는 합니다만, 아직까지는 구글링(영문) 및 NVIDIA 공식 문서가 공부하기에 가장 좋아보입니다. 막히는 부분은 Forum에서 검색하시는 것도 좋을 것 같네요. forums.developer.nvidia.com/

    • @user-bh9dj7jd5r
      @user-bh9dj7jd5r 23 วันที่ผ่านมา

      @@tweak-cd6ss 감사합니다!

  • @kyw0615
    @kyw0615 24 วันที่ผ่านมา

    isaac sim에서 작업한 걸 isaac gym으로 옮기는 거랑 그 반대 작업을 쉽게 하는 법이 있는지 궁금하네요.

    • @tweak-cd6ss
      @tweak-cd6ss 24 วันที่ผ่านมา

      저는 IsaacSim --> IsaacGym으로 변환해본 경험은 없습니다. IsaacGym 벤치마크 환경에서 IsaacSim기반 학습 환경인 IsaacLab으로 migration하는 과정은 isaac-sim.github.io/IsaacLab/source/migration/migrating_from_isaacgymenvs.html 여기서 확인이 가능하니 참고하시면 좋을 것 같습니다.

  • @user-ze9xy4zj7r
    @user-ze9xy4zj7r 24 วันที่ผ่านมา

    안녕하세요 tweak님, 옴니버스관련해서 공부중이었는데 영상이 너무 많은 도움이 되었네요. 혹시 여쭤보고 싶은게 있는데 메일이나 다른 연락 드릴수 있는 방법이 없는지 여쭤보고 싶습니다.!

    • @tweak-cd6ss
      @tweak-cd6ss 24 วันที่ผ่านมา

      안녕하세요. 채널 더보기에서 다음의 이메일 주소 확인이 가능합니다. mindstone39@gmail.com

  • @mrbinggrae5954
    @mrbinggrae5954 25 วันที่ผ่านมา

    유니티도 파이썬으로 개발 가능합니다. 검색하면 나와요!

    • @tweak-cd6ss
      @tweak-cd6ss 24 วันที่ผ่านมา

      아 그렇군요. 제가 2017년도 즈음에 사용했을 때는 C#과 javascript만 활용 가능했었던 것 같은데 검색해보니 이후에 python지원이 되었나 보네요.

    • @mrbinggrae5954
      @mrbinggrae5954 24 วันที่ผ่านมา

      @@tweak-cd6ss 저도 최근에 AI 공부하다가 우연히 알게 되었어요. 파이썬 생태계에도 침범하고자 여러가지를 넣어놨더군요.

  • @mrbinggrae5954
    @mrbinggrae5954 25 วันที่ผ่านมา

    컴터가 구리면 공부도 할 수 없는 그분... ㅠ

    • @tweak-cd6ss
      @tweak-cd6ss 24 วันที่ผ่านมา

      개인이나 소규모 랩은 점점 장비 허들이 높아 지는 추세네요..

  • @hubertkim7021
    @hubertkim7021 27 วันที่ผ่านมา

    귀한 영상 너무 감사합니다.

  • @hubertkim7021
    @hubertkim7021 27 วันที่ผ่านมา

    감사합니다~!

  • @hubertkim7021
    @hubertkim7021 27 วันที่ผ่านมา

    넘 감사합니다. 강의 올려주신 덕분에 일일이 라인별로 찾아가면서 공부할 시간을 많이 덜게됩니다. 😆

  • @user-zr3ke7tx9v
    @user-zr3ke7tx9v 28 วันที่ผ่านมา

    모든 영상 다 챙겨보고 있고 많은걸 받아갑니다 항상 감사합니다 ㅎㅎ 이제 아이작 짐에서 아이작 랩으로 옮겨가시는 건가요? 아니면 아이작 짐을 계속 다루실건가요? 더이상 아이작 짐은 지원하지 않는다고 엔비디아 페이지에 나와있길래 트윅님의 앞으로의 영상 내용이 궁금합니다!!

    • @tweak-cd6ss
      @tweak-cd6ss 28 วันที่ผ่านมา

      IsaacGym도 당분간은 계속 다룰 생각입니다. IsaacSim이 좋기는 하지만, 복잡하고 무겁다는 단점이 있습니다. 반면에 IsaacGym은 미들급 사양의 PC에도 원활하게 돌릴 수 있고, 알고리즘을 빠르게 검증하는 데에 앞으로도 쓸모가 있을 것 같습니다. pybullet이 여전히 쓰이는 것 처럼요. simulately.wiki/docs/simulators/IsaacGym/ 여기를 보시면 IsaacGym이 최근 연구에도 여전히 널리 사용되고 있습니다. Isaac 시리즈 이외에 다른 새로운 컨텐츠에 대한 exploration도 항상 고민중입니다 :)

  • @hubertkim7021
    @hubertkim7021 28 วันที่ผ่านมา

    귀한영상 감사합니다~! 주로 외국 영상만 찾아보다가 한국분이 디테일한것 까지 설명해주시는거 보니 속이 다 시원하네요.

    • @tweak-cd6ss
      @tweak-cd6ss 28 วันที่ผ่านมา

      좋게 봐주셔서 감사합니다 ^^

  • @seunghyeonjoe
    @seunghyeonjoe หลายเดือนก่อน

    트윅님 매번 양질의 영상잘보고있습니다!! : ) 다름이 아니라, 영상을 보다 궁금한점이 생겨서 질문 드립니다. 다른곳에서도 보면 pid보다 pd제어를 많이쓰는 경우를 볼수있는데, pd제어가 더 많이 쓰이는 이유를 알수있을까요?? 감사합니다.

    • @tweak-cd6ss
      @tweak-cd6ss หลายเดือนก่อน

      안녕하세요. 대부분의 경우 PID보다는 PI, 또는 PD와 같이 두 개의 term으로만 구성해서 사용하게 되는데, 응용에 따라서 굳이 세 개의 term을 모두 사용할 필요가 없기 때문입니다. 일단 변수가 많아지면 그만큼 튜닝이 어려워지고 제어기가 복잡해집니다. 즉, 내가 적용하고자 하는 응용에 대해서 불필요한 요소는 배제하고 반드시 필요한 요소만 활용하여 제어기를 구성하는 것이 시스템 복잡도, computational cost, tuning 작업 등에서 유리하므로 응용에 맞게 반드시 필요한 요소만 포함하는 것이 좋은데, 많은 응용에서 pd컨트롤러 만으로 충분하기 때문입니다. 예를 들어 휴머노이드 예제 같은 경우, 앞으로 달리는 모션을 생성함에 있어 하나의 지점에 정확한 타겟에 오랫동안 위치하는 것이 중요한게 아니라, 목표하는 위치로 빠르게 보내면서도 그렇다고 너무 불안정하지 않도록 적당히 안정적으로 관절을 움직이는 것이 중요합니다(안정적이며 빠른 응답성). 이 경우에 PD 컨트롤러만으로 충분히 원하는 바를 달성할 수 있기 때문에 굳이 i제어기를 포함시키지 않는 것입니다. 포함시켜봐야 계산할 것만 늘어나고 튜닝 과정도 복잡해 지기 때문입니다. i-제어기의 주요 목적 중 하나는 목표 위치에 대한 누적 오차를 줄이는, 즉 정상상태에러(steady-state error)를 없애는 것인데, 이는 목표 위치를 아주 정확하게 오랜 시간 동안 유지해야 하는 응용에 대해서는 중요하겠으나, 기민하면서도 적당히 안정된, 그리고 빠른 반응이 중요한 응용(e.g., humanoid, dexterous manipulation, drone 등)에서는 큰 역할을 수행하지 못하고 오히려 불리함을 야기 시킬 수 있습니다. 또한 noisy한 센서나 환경인 경우, 누적 에러가 과도해져서 integral term의 영향력을 증가시키고 이로 인하여 phase shift/lag(input signal과 output signal의 위상 차이)과 같은 문제가 발생하여 시스템 반응성이 나빠지고 오히려 시스템을 불안정하게 만들 수 있습니다(drift, bias issues). 저도 PID제어를 전부 활용하여 제어해본 경험은 없으나, 예상컨데 강화학습이나 imitation learning으로 학습할 때, policy action을 통해 next step에서 움직여야 할 motion의 반응성이 낮아지거나 불안정해 지면 불안정한 시스템에 의한 불안정한 데이터가 수집될 것이고, 결과적으로 학습에 불리하게 작용할 것으로 생각됩니다.

  • @tknots8675
    @tknots8675 หลายเดือนก่อน

    좋은 강의 공유해주셔서 정말 감사드립니다! 목소리가 되게 좋으셔서 집중이 잘되네요!!ㅎㅎ

  • @user-kurdsmjs
    @user-kurdsmjs หลายเดือนก่อน

    어? 혹시 road balance 대표님 아니신가요? 목소리가 똑같으신데…?

    • @tweak-cd6ss
      @tweak-cd6ss หลายเดือนก่อน

      아쉽게도 아니네요^^;

    • @user-kurdsmjs
      @user-kurdsmjs หลายเดือนก่อน

      @@tweak-cd6ss 그렇군요.. 어찌되었건 Issac gym 강의 유튜브에서 보기 쉽지 않은데 한국어로 이런 고품질의 강의를 공유해주셔서 정말 감사드립니다. 학습에 있어 유용하게 활용하겠습니다. 감사합니다

  • @uniforcestelian7917
    @uniforcestelian7917 หลายเดือนก่อน

    감사합니다.. 올려주신 isaacgym 튜토리얼로 환경구축하는데 큰 도움 받았는데요, isaacsim도 기대가 됩니다. isaacgym에서 isaacsim으로 실험환경을 옮겨가는것을 생각중인데, 해당부분에 대한 엄두가 잘 안나네요, isaacgym에서 isaacsim으로 코드 변환에 대한 튜토리얼도 만들어 주실 계획이 있을까요?

    • @tweak-cd6ss
      @tweak-cd6ss หลายเดือนก่อน

      원래 계획은 없었는데 최근에 저도 기존에 구현했던 환경을 isaacsim으로 옮기는 것도 좋을 것 같다는 생각은 하고 있었습니다. 특히 photorealistic 영상이나 isaacgym의 미흡한 기능을 isaacsim이 많이 채워줄 것으로 기대합니다. 해당 내용도 고려해보겠습니다

  • @user-zr3ke7tx9v
    @user-zr3ke7tx9v หลายเดือนก่อน

    너무너무 감사합니다!!

  • @hubertkim7021
    @hubertkim7021 หลายเดือนก่อน

    와 엄청 디테일하게 알려주시네요. 감사합니다!

  • @user-ix3zc3bp4k
    @user-ix3zc3bp4k หลายเดือนก่อน

    감사합니다!!

  • @thanhbaotran6412
    @thanhbaotran6412 2 หลายเดือนก่อน

    hello, can you help me please? i replace franka's gripper and use my own gripper, but, then, the robotarm does not move like that: drive.google.com/file/d/1bsxbowOqlTEDCM_bDKZ0E4iT_VUDGoNc/view?usp=sharing i cannot paste the image in here, so i uploaded it to drive please help me to fix it

    • @tweak-cd6ss
      @tweak-cd6ss 2 หลายเดือนก่อน

      I cannot see the image you uploaded, but if the robot is not moving properly after replacing the gripper, it is likely that the URDF is not correctly configured. I recommend verifying that the URDF is functioning correctly using ROS2 and RViz packages.

    • @thanhbaotran6412
      @thanhbaotran6412 2 หลายเดือนก่อน

      @@tweak-cd6ss so, the moving of robot arm and gripper will depend on which? code on the franka_cube_osc.py or on my URDF file? thank you

    • @thanhbaotran6412
      @thanhbaotran6412 2 หลายเดือนก่อน

      @@tweak-cd6ss hello, can you please help me to check it? I verified, everything is okay, but i still dont know why it doesnt work, i stucked on it for a long time, so, please help me to test it. thank you so much. thank you so much. drive.google.com/drive/folders/1xOuSQ2JFOmqTemeSGg_iWTrXSkV-ZjiR?usp=drive_link i uploaded the python file and folders have urdf and meshes file, my urdf file names franka_sus.urdf in franka_description folder. thank you so much

    • @thanhbaotran6412
      @thanhbaotran6412 2 หลายเดือนก่อน

      @@tweak-cd6ss hello, can you please help me to check these file, i verified but it doesnt work, i dont know why is it. i stuked in this for a long time, please help me to check, test and fix it please. thank you so much for your help. drive.google.com/drive/folders/1xOuSQ2JFOmqTemeSGg_iWTrXSkV-ZjiR?usp=sharing . i put my own urdf file with name franka_sus.urdf in the franka_description folder. please help me.

    • @tweak-cd6ss
      @tweak-cd6ss 2 หลายเดือนก่อน

      "Your uploaded file appears to be based on the ROS1 build system. I am currently not using ROS1, and it would take considerable time for me to verify your file. If the control works correctly in RViz, there shouldn't be significant issues running it in IsaacGym, but other aspects like paths may need verification. I'm sorry I couldn't be of more help."

  • @charles26842
    @charles26842 2 หลายเดือนก่อน

    Hello, After trying to extract the file to my NVidia local host server, I run into an error where it says "failed to copy some files". Error code "Error 0x80070522", do you know how to resolve this issue? Your help would be greatly appreciated, thank you sir.

    • @tweak-cd6ss
      @tweak-cd6ss 2 หลายเดือนก่อน

      The error code "0x80070522" indicates that you do not have the necessary permissions to perform the action you're attempting, which typically involves copying files to a restricted location. This issue often occurs when trying to move files to system directories or other protected locations without the appropriate permissions. Running as an administrator or trying various permission-related methods might be helpful.

    • @charles26842
      @charles26842 2 หลายเดือนก่อน

      @@tweak-cd6ss Thank you, also thank you for replying to older TH-cam videos, you are the BEST ! I was finally able to get it to extract using 7-zip with command window. I ran into another error where after creating a python environment and installing -e. I had a new error where it says PhysX SDK and lib were not found/created. Have you come across this issue? Sense I am on windows 11 do you think that could be causing issues ? Thank you again for your help : )

    • @tweak-cd6ss
      @tweak-cd6ss 2 หลายเดือนก่อน

      Isaac Gym does not officially support installation on Windows. It is only installable on Ubuntu 18.04 or later versions. To install Isaac Gym on Windows, you need to use WSL (Windows Subsystem for Linux) to set up an Ubuntu environment, and then proceed with the Isaac Gym installation within that environment.

    • @charles26842
      @charles26842 2 หลายเดือนก่อน

      @@tweak-cd6ss I have been reading on the forum post and saw that Isaac Sim 4.0 supports unified RL platform and Multi-GPU training. Do you think Isaac Sim is now an easier way to do RL in ? Also, have you ever done RL in Isaac Sim or was it only in Gym ? Thank you again btw.

    • @tweak-cd6ss
      @tweak-cd6ss 2 หลายเดือนก่อน

      Isaac Sim currently provides an RL training framework based on Multi-GPU, along with well-organized manuals, making it relatively easy to perform desired tasks. This offers several advantages over Isaac Gym, but like everything, it comes with its own set of pros and cons. The current RL framework based on Isaac Sim demands higher computer specifications and is structurally more complex. Although well-structured frameworks like Isaac Orbit have recently been introduced, they are still heavy and relatively complex in structure. Based on my experience using RL with Isaac Sim and Isaac Orbit, while they offer richer features and photorealistic images, I personally use Isaac Gym more frequently. Isaac Gym is more intuitive, lightweight, and easier to debug. Additionally, due to these advantages, there seem to be more research cases utilizing Isaac Gym in top-tier robotics conferences such as RSS, ICRA, and CoRL.

  • @simon277
    @simon277 2 หลายเดือนก่อน

    Thank you very much for the video. I got a question while designing a similar mount for our robot. So why 15 mm in thickness? Can i use 10 mm for weight reduction?

    • @tweak-cd6ss
      @tweak-cd6ss 2 หลายเดือนก่อน

      First of all, the material specified in the drawing uploaded to Git is AL 7075. The notation of 7074 is a typo. If you are making an adapter with a thickness of 10mm from AL 7075 material, 10mm thickness will be sufficient. The yield strength of heat-treated AL 7075 is approximately 450 MPa, and the stress on the adapter for a 1.5kg Allegro hand is roughly 5 MPa according to a simple calculation, which is enough to support the robot hand. However, if the goal is simply to reduce weight, drilling dummy holes could be one option.

    • @simon277
      @simon277 2 หลายเดือนก่อน

      @@tweak-cd6ssThank you very much. Using 7 series aluminum sounds sturdy enough. Good luck to your projecets!

  • @user-yd1rj8fy1o
    @user-yd1rj8fy1o 2 หลายเดือนก่อน

    항상 잘 보고있습니다. 감사합니다. 개인적으로 api로 sim 환경 구축하는 것 까지만 해도 굉장히 좋은 영상이 될 것 같습니다.

    • @tweak-cd6ss
      @tweak-cd6ss 2 หลายเดือนก่อน

      참고하겠습니다. 감사합니다

  • @user-zr3ke7tx9v
    @user-zr3ke7tx9v 2 หลายเดือนก่อน

    안녕하세요 언제나 양질의 영상을 올려주셔서 정말로 감사합니다 아이작심에서의 시뮬레이션을 현실로 옮기는 방법인 sim2real을 트윅님이 만드신 예제를 통해 설명하는 동영상도 좋을 것 같아요 ㅎㅎ

    • @tweak-cd6ss
      @tweak-cd6ss 2 หลายเดือนก่อน

      많은 분들이 sim2real에 관심이 많으신 것 같네요. 고려해보도록 하겠습니다

  • @hoeunsu3
    @hoeunsu3 2 หลายเดือนก่อน

    잘 보고 있습니다 감사합니다 :)

  • @hjj5600
    @hjj5600 2 หลายเดือนก่อน

    혹시 Isaac Gym에서 개인이 제작한 모델을 넣고 학습하는 법은 어디서 참고해야 할까요?

    • @tweak-cd6ss
      @tweak-cd6ss 2 หลายเดือนก่อน

      커스텀 모델 import 관련 영상은 아직 없습니다 ^^;

    • @hjj5600
      @hjj5600 2 หลายเดือนก่อน

      @@tweak-cd6ss 그렇군요. 답변 감사드립니다. 항상 영상 잘 보고 있습니다

    • @hjj5600
      @hjj5600 2 หลายเดือนก่อน

      @@tweak-cd6ss 혹시 추후 그런 영상을 제작할 계획은 있으신건가요?

    • @tweak-cd6ss
      @tweak-cd6ss 2 หลายเดือนก่อน

      다른 분들도 요청해주시는 분들이 종종 계셔서 고려해보도록 하겠습니다.

  • @donghokang5933
    @donghokang5933 2 หลายเดือนก่อน

    Robot learning 쪽에 중점을 둔 Orbit 에 관한 영상도 좋을거 같아요. Orbit 을 Isaac Lab 으로 쓸거라고 하네요.

    • @tweak-cd6ss
      @tweak-cd6ss 2 หลายเดือนก่อน

      어차피 Isaac Sim 기반이니 Orbit 영상도 좋을 것 같네요.

  • @thanhbaotran6412
    @thanhbaotran6412 2 หลายเดือนก่อน

    hello, can you help me to take the image of this project ( not graphics.py) by using the camera API in isaacgym please? thank you so much.

    • @tweak-cd6ss
      @tweak-cd6ss 2 หลายเดือนก่อน

      Hello, I will soon be uploading an example of how to use camera images in Isaac Gym.

  • @user-il3hr2lt8s
    @user-il3hr2lt8s 3 หลายเดือนก่อน

    Now that I have trained model parameters, how can I verify them in real objects? Any good ideas? If you can help me, I will be very grateful.

    • @tweak-cd6ss
      @tweak-cd6ss 3 หลายเดือนก่อน

      If you have trained a state-based model, you can verify it by placing the same robot and objects used in the simulation environment in the same positions in the real world. However, the placement needs to be relatively precise. If you have trained using image-based learning, you can attempt verification in the real environment by appropriately placing the same robot and objects, but positioning the camera accurately may be challenging. To solve such issues, you can apply domain randomization during the training process, which involves varying the positions of the camera and objects.

  • @user-il3hr2lt8s
    @user-il3hr2lt8s 3 หลายเดือนก่อน

    Is there any example of a robotic arm opening a cabinet with camera information added?

    • @tweak-cd6ss
      @tweak-cd6ss 3 หลายเดือนก่อน

      There are no default or benchmark examples based on camera observations. However, you can create such an environment by replacing the state with image observations and incorporating a CNN into the initial part of the policy neural network.

    • @user-il3hr2lt8s
      @user-il3hr2lt8s 3 หลายเดือนก่อน

      @@tweak-cd6ss Thank you very much for your reply, because I am not particularly familiar with reinforcement learning networks, I would like to ask if there are any examples that you have written for reference in the method of adding a layer of CNN to the initial network

  • @user-il3hr2lt8s
    @user-il3hr2lt8s 3 หลายเดือนก่อน

    franka_actor = self.gym.create_actor(env_ptr, franka_asset, franka_start_pose, "franka", i, 1, 0),cabinet_actor = self.gym.create_actor(env_ptr, cabinet_asset, cabinet_pose, "cabinet", i, 2, 0)What do the 1 and 2 corresponding to the sixth parameter in these two lines of code mean?

    • @tweak-cd6ss
      @tweak-cd6ss 3 หลายเดือนก่อน

      The numbers are documented as "param6 (int) - a bitwise filter used to mask off collision among elements in the same collisionGroup." For more detailed information, please refer to the document at isaacgym/docs/index.html.

  • @BeenBeen-eb8pp
    @BeenBeen-eb8pp 3 หลายเดือนก่อน

    hello, thank you for your helpful video. but I have a question, i hope you can help me answer it. I have my own robot arm and gripper with urdf file, and i dont want to use franka and its gripper, so how will i do to replace to my own robot arm and gripper? thank you so much

    • @tweak-cd6ss
      @tweak-cd6ss 3 หลายเดือนก่อน

      You can add your own robot to the simulation environment by modifying the asset file path. For example, in the franka cabinet example, replace the 'franka_asset_file' in the _create_env() function with the path to your urdf file.

    • @BeenBeen-tv7pi
      @BeenBeen-tv7pi 3 หลายเดือนก่อน

      @@tweak-cd6ss thank you so much for your reply. but i still confuse about urdf file, please help me answer it. i saw that, in the code file, it has the path to gripper urdf and robot arm urdf seperately. so, should i make a link between arm and gripper in the 3D file, then export to 1 urdf file includes both of arm and gripper? or i will import path of gripper urdf and arm urdf file seperately like in the code file? thank you so much.

    • @tweak-cd6ss
      @tweak-cd6ss 3 หลายเดือนก่อน

      I recommend combining the separate URDF files first and then importing the merged file into the simulator. This approach will give you more stable control results. You can find information on how to combine URDF files by searching on Google.

    • @BeenBeen-tv7pi
      @BeenBeen-tv7pi 3 หลายเดือนก่อน

      @@tweak-cd6ss thank you so much for your answer. hi, but i still confuse sth, please help me answer it. so, my task is using computer vision and robot arm to pick a object in the dense objects, i will train policy to take the best position of object. but i still have not know how to use, transfer the training result into the real world, or how to transfer from the sim to real as know as sim2real. thank you so much

    • @tweak-cd6ss
      @tweak-cd6ss 3 หลายเดือนก่อน

      If your goal is to train a policy for a robotic pick-and-place task and transfer it to a real environment, the paper "Transferring End-to-End Visuomotor Control from Simulation to Real World for a Multi-Stage Task," published in CoRL 2017, could be helpful. It provides a comprehensive overview of the process of transferring simulated robot manipulation skills to the real world. I recommend reading the paper thoroughly and watching the accompanying video clip for a better understanding.

  • @user-eu8xd1ww5p
    @user-eu8xd1ww5p 3 หลายเดือนก่อน

    дякую

  • @user-il3hr2lt8s
    @user-il3hr2lt8s 3 หลายเดือนก่อน

    When I set enableDebugVis: True, FrankaCabinet reported NameError: name 'quat_apply' is not defined. Have you ever encountered this error?

    • @tweak-cd6ss
      @tweak-cd6ss 3 หลายเดือนก่อน

      The issue appears to be related to the library path. The function 'quat_apply' is defined in isaacgym/python/isaacgym/torch_utils.py. Please ensure that your Isaac Gym installation is correct.

    • @user-il3hr2lt8s
      @user-il3hr2lt8s 3 หลายเดือนก่อน

      @@tweak-cd6ss Thank you very much for your prompt reply, I successfully solved this problem. But one more question for you, have you tried the Robotiq85 type of gripper to open and close in the Isaac Gym?How do I do that?

    • @tweak-cd6ss
      @tweak-cd6ss 3 หลายเดือนก่อน

      Please refer to this GitHub github.com/aai4r/aai4r-pouring-skill/tree/release. It includes the process of simulating a pouring task with a robotiq85 gripper attached to a UR3 robot in the Isaac Gym environment. However, since this is our on-going project, the code is not yet neatly organized.

  • @user-io9io3mh7p
    @user-io9io3mh7p 4 หลายเดือนก่อน

    안녕하세요 영상 잘 봤습니다. 제가 실습을 진행해보려고 따라서 명령어를 작성하였는데 fbx no module이란 오류가 발생해서 혹시 해결법을 알고 계신가요?

    • @tweak-cd6ss
      @tweak-cd6ss 4 หลายเดือนก่อน

      안녕하세요. 해당 오류는 겪어본 적이 없어서 잘 모르겠네요.

  • @wendypark3848
    @wendypark3848 4 หลายเดือนก่อน

    많이 배우고 갑니다. 감사합니다

  • @SantiDiazC
    @SantiDiazC 5 หลายเดือนก่อน

    Thank you for this video series about Isaacgym. I followed most of the videos even though my Korean is not that good (auto-translate captions), and It helped me to understand how the examples work which is not very clear in the documentation. Thank you so much!

    • @tweak-cd6ss
      @tweak-cd6ss 5 หลายเดือนก่อน

      I’m glad to hear that it was helpful :)

  • @ddd-zi8cd
    @ddd-zi8cd 5 หลายเดือนก่อน

    로봇팔이 올라갔다가 내려갔다가

  • @ddd-zi8cd
    @ddd-zi8cd 5 หลายเดือนก่อน

    마찰력은 어디갔노

  • @songminjumeokbap
    @songminjumeokbap 5 หลายเดือนก่อน

    너무 잘 보고있습니다. 양질의 영상 정말 감사합니다. 많이 공부하고 배우겠습니다~!!!

  • @user-nf9ew2fd2c
    @user-nf9ew2fd2c 6 หลายเดือนก่อน

    python joint_monkey.py 실행 시 Segmentation error (core dumped) 에러가 나는데 혹시 해결 방법 아시나요? Physics Engine: PhysX Physics Device: cuda:0 GPU Pipeline: disabled Segmentation fault (core dumped)

  • @user-jx9rl1ny3h
    @user-jx9rl1ny3h 6 หลายเดือนก่อน

    안녕하세요! 좋은영상 감사합니다 :) Isaacgym API에 tendon properties가 추가되었던데 혹시 이부분에 대해서도 영상으로 제작할 계획이 있으신가요?

    • @tweak-cd6ss
      @tweak-cd6ss 6 หลายเดือนก่อน

      안녕하세요. 해당 부분이 추가되었는지 몰랐네요. 알아보고 필요하다 판단되면 영상 제작을 고려해보겠습니다~

  • @user-jx9rl1ny3h
    @user-jx9rl1ny3h 7 หลายเดือนก่อน

    좋은영상 갑사합니다! 혹시 이 결과물도 Isaac gym 내에 import하는 방법도 영상으로 올리실 계획이 있을까요?

    • @tweak-cd6ss
      @tweak-cd6ss 7 หลายเดือนก่อน

      말씀하신 영상은 별도의 제작 계획은 원래 없습니다만, 제작을 고민해 보도록 하겠습니다.

    • @user-jx9rl1ny3h
      @user-jx9rl1ny3h 7 หลายเดือนก่อน

      ​@@tweak-cd6ss감사합니다 :)

  • @dinglibao1401
    @dinglibao1401 7 หลายเดือนก่อน

    Thanks for sharing. What tool do we usually use to create/view urdf files?

    • @tweak-cd6ss
      @tweak-cd6ss 7 หลายเดือนก่อน

      Numerous options are available for constructing and inspecting URDF files, such as utilizing SolidWorks, Onshape, and various online viewers. Check out these resources for guidance: SolidWorks and Onshape tutorials: th-cam.com/video/TJeCpGnX508/w-d-xo.html, th-cam.com/video/T7X_p_KMwus/w-d-xo.html. Foxglove Studio for URDF visualization: foxglove.dev/urdf. Alternatively, you can explore a variety of results by searching "URDF creator, viewer" on Google to find the tool that best suits your needs.

    • @dinglibao1401
      @dinglibao1401 7 หลายเดือนก่อน

      Really helpful. Thanks a lot@@tweak-cd6ss

  • @dinglibao1401
    @dinglibao1401 7 หลายเดือนก่อน

    Thanks for sharing. Is it possible to share the repo?

    • @tweak-cd6ss
      @tweak-cd6ss 7 หลายเดือนก่อน

      Absolutely. I've shared the repository on my GitHub page. Take a look at github.com/tweak4u/IsaacGymEnvs/blob/main/isaacgymenvs/tasks/franka_cabinet.py. Everything matches the original source code except for a few lines dedicated to coordinate visualization.

  • @AubreyWintringham
    @AubreyWintringham 7 หลายเดือนก่อน

    구독자 100명 축하드립니다! 너무 큰 도움이 되고 있습니다! 저희도 아이작 짐을 차근차근 따라할 컨텐츠가 없었는데 이렇게 올려주시니 너무 큰 도움이 됩니다!

    • @tweak-cd6ss
      @tweak-cd6ss 7 หลายเดือนก่อน

      감사합니다 :)