Загрузка видео...

Не удалось загрузить видео

На главную

🌎World models can predict, but controlling real robots from imagination sees a long-standing failure due to hallucination. 🧠Introducing Uncertainty-Aware RWM: a black-box, end-to-end neural dynamics model with long-horizon uncertainty propagation. 🎯

60,269 просмотров • 9 месяцев назад •via X (Twitter)

Комментарии: 16

Фото профиля Chenhao Li
Chenhao Li9 месяцев назад

❗️World models hallucinate under distribution shift: in low-data regions, small errors compound over long autoregressive rollouts. 🧠Policy optimization then exploits these hallucinations, achieving high reward in imagination while failing catastrophically in the real world.

Фото профиля Chenhao Li
Chenhao Li9 месяцев назад

🌎Uncertainty-Aware Robotic World Model (RWM-U) extends RWM by modeling not only what will happen, but how reliable those predictions are. 🎯We augment autoregressive world models with ensemble-based uncertainty estimation, explicitly capturing epistemic uncertainty from limited or biased offline data.

Фото профиля Chenhao Li
Chenhao Li9 месяцев назад

⭐️Each ensemble member in RWM-U predicts a Gaussian distribution over the next observation. 🎯The predicted variance captures aleatoric uncertainty, while disagreement across ensemble means estimates epistemic uncertainty from limited or biased offline data.

Фото профиля Chenhao Li
Chenhao Li9 месяцев назад

⭐️In long autoregressive rollouts, RWM-U’s epistemic uncertainty closely tracks true prediction error, providing a reliable signal of when the model becomes untrustworthy.

Фото профиля Chenhao Li
Chenhao Li9 месяцев назад

We bring MOPO to long-horizon world models, and to PPO. ✅MOPO-PPO trains policies entirely in imagination, penalizing uncertain transitions to avoid hallucination. No real environment interaction ⏩fast, fully offline learning.

Фото профиля Chenhao Li
Chenhao Li9 месяцев назад

💾Starting from pure offline data, with no online environment interaction (not even a simulator), we train policies that are directly deployable on real hardware.

Фото профиля Chenhao Li
Chenhao Li9 месяцев назад

⭐️Though completely offline, RWM-U and MOPO-PPO achieve comparable performance as state-of-the-art simulator-based online policies.

Фото профиля Chenhao Li
Chenhao Li9 месяцев назад

🌎A key advantage of model-based RL is that the world model can be trained directly on real-world data. 🔝By incorporating real data into model learning, we reduce the sim-to-real gap and achieve final policy performance that surpasses sim-to-real baselines.

Фото профиля Chenhao Li
Chenhao Li9 месяцев назад

🧠Policy conservativeness is controlled by the uncertainty penalty weight during imagination. 🛑When offline data has limited task relevance, increasing the penalty yields more conservative and safer policies.

Фото профиля Chenhao Li
Chenhao Li9 месяцев назад

🌎Uncertainty-Aware Robotic World Model Makes Offline Model-Based Reinforcement Learning Work on Real Robots 👥Chenhao Li @breadli428, Andreas Krause, Marco Hutter @leggedrobotics 🎯Webpage: 📄Paper:

Фото профиля ζ Pedram ζ
ζ Pedram ζ9 месяцев назад

Nice

Фото профиля Sohom Mukherjee
Sohom Mukherjee9 месяцев назад

love seeing this. great work addressing robot hallucination with better models.

Фото профиля Lukas Die Kunst
Lukas Die Kunst9 месяцев назад

The black-box neural dynamics approach here intrigues me. How does its uncertainty propagation compare to traditional control systems in real-world robotics applications?

Фото профиля Damian
Damian9 месяцев назад

Dm

Фото профиля Rahul Raghav
Rahul Raghav9 месяцев назад

how does uncertainty propagation impact real-world robot control stability?

Фото профиля Eddie
Eddie9 месяцев назад

so they predict hallucinations now? sounds like just another overhyped ai buzzword waiting to fail in real applications.

Похожие видео