Embodied AI Engineering: World Models, Foundation Models for Robotics, and the Architecture of Physically Intelligent Systems (AI Infrastructure, Hardware & Compiler Engineering Series) - Tapa blanda

Sinopsis

Master the Next Frontier of Robotics and Physically Intelligent Systems

Physical AI is the defining engineering revolution of this decade. As humanoid robots from Figure, Agility, Tesla, and Unitree transition from lab prototypes to commercial deployment, the demand for engineers who understand embodied AI systems has skyrocketed. Embodied AI Engineering is the definitive, hands-on guide to implementing foundation models, world models, and vision-language-action (VLA) pipelines on physical machines.

This book bypasses abstract academic derivations to deliver mid-level engineering depth, concrete architectural patterns, and practical deployment strategies. You will learn how to bridge the gap between digital AI models and physical actuators operating in unpredictable, real-world environments.

• World Models & Latent-Space Planning: Deep dive into RSSM, DreamerV3, and video prediction architectures for predictive physics and planning.
• Imitation Learning & Diffusion Policy: Train robust policies using behavioral cloning, DAgger, and Action Chunking Transformers (ACT).
• VLA Architectures: Demystify state-of-the-art models including RT-2, OpenVLA, and π0 from Physical Intelligence.
• Physical Tokenization: Learn how transformer models tokenize actions, 3D observations, and proprioceptive sensor feedback.
• Sim-to-Real Transfer: Utilize NVIDIA IsaacSim, Isaac Lab, and Genesis to train policies safely before deployment.
• Edge Inference & Latency: Optimize 7-billion-parameter models for low-latency, real-time control on NVIDIA Dragonwing and Qualcomm platforms.

Whether you are a machine learning engineer, software developer transitioning to robotics, or a hardware engineer adopting modern AI, this book provides the engineering blueprints, safety validation frameworks, and deployment pipelines to build the next generation of humanoid robots.

"Sinopsis" puede pertenecer a otra edición de este libro.