Загрузка видео...

Не удалось загрузить видео

На главную

200ms of simulated network latency. Same test. Different physics engine. The first half shows Unreal's Chaos physics. Notice the visible corrections and jitter when interacting with overlapping surfaces. In the second half, I switch to Box3D (debug rendering enabled) under the same network conditions. The corrections largely disappear and...

20,337 просмотров • 4 месяцев назад •via X (Twitter)

Комментарии: 23

Фото профиля Steve Robinson
Steve Robinson4 месяцев назад

*Listening* This looks so coherent. Nice, man. 🔥

Фото профиля Antonio Lattanzio
Antonio Lattanzio4 месяцев назад

Thanks man! 🙏

Фото профиля Anthony Emedobi
Anthony Emedobi4 месяцев назад

This is solid chief 🙌

Фото профиля Antonio Lattanzio
Antonio Lattanzio4 месяцев назад

Thanks man! Hopefully it will become more solid soon 😃

Фото профиля Anthony Emedobi
Anthony Emedobi4 месяцев назад

Sure mate 🫡

Фото профиля Daniele Benegiamo
Daniele Benegiamo4 месяцев назад

The difference is impressive. Great job!

Фото профиля Antonio Lattanzio
Antonio Lattanzio4 месяцев назад

Thanks, man! I believe Chaos probably has a way to avoid this, but compared to Box3D, it feels a bit overengineered, which can make things more frustrating than they need to be

Фото профиля Daniele Benegiamo
Daniele Benegiamo4 месяцев назад

Absolutely. About two years ago, I studied the Chaos implementation to understand how to use it properly, since there was no documentation available. They made some “peculiar” choices...

Фото профиля Daniele Benegiamo
Daniele Benegiamo4 месяцев назад

When they were switching from PhysX to Chaos, the entire engine - for some versions - used a single, large interface as the physics API. A bridge between the two implementations. I was hoping this was the beginning of a plugin-based physics system, but instead...

Фото профиля Barak
Barak4 месяцев назад

This is really really cool! I have an Unreal project and I think this can really help. Thanks for the awesome work - definitely gonna check this out once open source lands! 👑👑

Фото профиля Caleb Longmire
Caleb Longmire4 месяцев назад

Real question is how to make the swap, cause as soon as someone shows me how I sure will.

Фото профиля Antonio Lattanzio
Antonio Lattanzio4 месяцев назад

Please follow me as I will post the open source plugin as soon as is in a decent state :)

Фото профиля Morphtek
Morphtek4 месяцев назад

what solver settings did you have on the chaos ?

Фото профиля Antonio Lattanzio
Antonio Lattanzio4 месяцев назад

I am using the default settings! Solver Iterations • Position Iterations: 4 • Velocity Iterations: 1 • Projection Iterations: 1 Collision & Spacing • Collision Margin Fraction: 0.05 • Collision Margin Max: 2.0 • Contact Search Distance: 3.0 Physics Substepping • Max Physics Delta Time: 0.033333 • Min Physics Delta Time: 0.001333 • Substepping: Disabled (Default) • Cluster Connection Factor: 1.0

Фото профиля 창새, 슈퍼둘기맨
창새, 슈퍼둘기맨4 месяцев назад

우와. 얼마 전에 멋진 파괴효과와 거미 영상을 추천으로 봤었는데 안토니오님 이었군요. 결과물들 정말 멋지게 잘 봤어요. 🤩

Фото профиля Antonio Lattanzio
Antonio Lattanzio4 месяцев назад

정말 감사합니다! 😊 추천으로 보시게 됐다니 정말 기쁘네요. 앞으로도 더 멋진 결과물 보여드릴 수 있도록 열심히 만들겠습니다! 🤩

Фото профиля over
over4 месяцев назад

well its caused chaos for a reason

Фото профиля Filippo Tarpini
Filippo Tarpini3 месяцев назад

Nice. Is Box3D simulation deterministic even at different frame rates, or do both instances need to have the exact same fps?

Фото профиля unemployable
unemployable4 месяцев назад

Too much GPU wasted on the ass crack, that's your bottleneck

Фото профиля typeshi
typeshi4 месяцев назад

Is this music necessary

Фото профиля AI Mastery Guide
AI Mastery Guide4 месяцев назад

Deterministic simulation making that much of a difference under latency is a great find.

Фото профиля 돈벌고영
돈벌고영4 месяцев назад

200ms 지연에서 나란히 보니 차이가 꽤 노골적이네요. Box3D 쪽은 물체가 겹친 뒤에도 움직임이 덜 불안해서 비교가 바로 됩니다.

Фото профиля 😏
😏4 месяцев назад

No box3d will allow you to predict which buttons the user pressed during those 200ms

Похожие видео

The sense of presence in VR is something amazing. And it is what convinces me that sooner or later, VR will succeed. VR is now in one of its down moments. That's why, in these harsh days, I wanted to get back to feeling what made me fall in love with this technology: presence and immersion. I took some very good-looking PCVR experiences, like Half-Life: Alyx, or AirCar, and tried them at the maximum resolution possible on my Dell Pro Max T2 Tower, featuring an #NVIDIA RTX 6000 graphics card. Until a few months ago, I was using my old laptop, which started to show throttling issues, so I was used to lowering the render resolution, while this time, I was moving the slider up and up. The results have been astonishing: even a demo like AirCar made me in awe again, and Half-Life: Alyx was as immersive as I have ever tried it. As much as I love standalone VR, because I think it helped a lot in democratizing the technology, the WOW sensation that you can try on PCVR today is still unbeatable. Probably a good compromise between the two can be cloud rendering, but we still have to wait for it to become a reality. The sense of being somewhere else is a wonderful sensation, and everyone I demoed VR to all these years has been amazed by it, too. And every day, something makes me feel "there" (like today), I remember why I love this technology and why one day it will have a higher adoption. Sure, there are a lot of problems to solve, but I'm convinced one day it will happen. When? Well, it is always 5 to 10 years, every year... #VirtualReality #VR #technology [Disclaimer: I mentioned the PC model because I'm a #DellProPrecision ambassador, and I have been given the PC to do tests and experiments and share my results. There is no monetary compensation or sales affiliation related to this post]

TonyVT SkarredGhost

28,965 просмотров • 7 месяцев назад

Experiments in progress. The one on the right has been learning for ~3 hours, the one in the middle for ~1 hour, and the one on the left just started a few minutes ago. The initial motivation for making the physical Atari was just to commit ourselves to a subset of algorithms that can make progress in this setup. This commitment rules out algorithms that require billions of samples to learn (or worse, require multiple environments running in parallel). Atari games are simple enough that we should be able to show learning on them in a short amount of time with no prior knowledge. Since then, I've realized that this setup is also a good way to compare different paradigms in robotics in a principled way. These paradigms are sim2real, learning from tele-operated data, and learning directly on the robots. So far, I have observed that getting sim2real to work reliably is hard. It requires tweaks that don't scale. Policies that can play perfectly in simulation fall apart because of latencies and the messiness of the real world. These aspects could be modeled to improve the simulation, but not without sinking significant human engineering hours. I have higher hopes for learning from tele-operated data, but that requires a human to learn the task first. These experiments are on my to-do list. I have to learn to play some of the games well through the robot. I’m half-decent at playing Pong and Ms Pacman now. Learning directly on robots is looking like the most promising approach. This approach takes away pesky distribution shifts and makes it possible to have algorithms that continually improve with more data and time without any human intervention. It feels great to let experiments run overnight and wake up to find improved policies. With learning on robots, I should, in principle, be able to go on a long vacation and come back to find better policies for complex tasks beyond Atari games. Whether that is possible with current learning algorithms is a different question.

Khurram Javed

52,110 просмотров • 11 месяцев назад