Embodied Intelligence
Embodied Intelligence is a core frontier of artificial intelligence, studying how agents perceive, interact with, and learn from the physical world through their bodies. Unlike purely digital AI, embodied intelligence emphasizes that intelligence must be "embodied" — the body is not merely an I/O device, but the very medium through which cognition occurs (Varela, 1991).
This section comprehensively covers the theoretical foundations, core technologies, software/hardware ecosystems, and industry landscape of embodied intelligence:
Contents:
- Overview — Survey, roadmap, milestones, key conferences
- Theoretical Research — Embodied cognition, perception-action loop, world models, paper reviews
- Robotics Fundamentals — Kinematics, dynamics, motion planning, control theory, SLAM
- Robot Learning — Imitation learning, RL for robotics, Sim2Real, diffusion policy, teleoperation
- Models & Algorithms — VLA models, world models, LLM-driven robotics, open-source models
- Simulation & Software Development — Simulation platforms, simulation assets, world building, ROS2, NVIDIA ecosystem, development toolchains
- Hardware — Sensors, actuators, computing platforms, open-source hardware
- Robot Forms — Humanoid, quadruped, dexterous hands, drones, service robots
- Real-World Deployment — Sim2Real deployment, safety, calibration, real-time systems
- Industry Ecosystem — Company landscape, industry research, market, academia-industry collaboration, challenges