Build intuition for sequential decisions, from MDPs to PPO and SAC.
RT, OpenVLA, ACT, Diffusion Policy, and the π family.
Explore how world models can be applied to embodied tasks.
Connect policy learning to continuous control and robotics tasks.
A structured course from PID and LQR to MPC and impedance control.
From models and frame trees to closed-loop planning with MoveIt 2.
Transformers, ViT, vision encoders, and multimodal fusion.
SLAM, foot contact, tactile sensing, and multisensor fusion.
Get started with MuJoCo, Isaac Sim, Gymnasium, and PyBullet.
Coordinate transforms, FK/IK, tf2, URDF, and MoveIt 2.
Low-level communication, actuator protocols, and host–controller links.
Links, joints, motors, gearboxes, and end effectors.
From teleoperation data to imitation learning and policy training.
Access Ryzen AI APUs in your browser with ROCm, JupyterHub, and Code Server.
ACT BF16 training on Radeon GPUs, closed-loop evaluation, and successful rollouts.
An AMD flagship project: reinforcement learning for locomotion and VLA experiments.
Follow eight chapters, from PD control and kinematics to policy training and perception.
Parallel PPO training on GPUs and stable gaits with mjlab and MuJoCo Warp.
From environment setup to continuous control with InvertedPendulum, Reacher, and Pusher.
Train ACT on ALOHA simulation data and evaluate it over multiple episodes.
Preview the course on Isaac Lab training and cross-simulator validation.
Prepare for robot learning with datasets, toolchains, and classical robotics.
From hardware connectivity and safety tests to action playback on a real robot.
Validate a simulation policy before deploying it to a real robot.
Page content converted to Markdown. Use the original page link at the end to explore interactive graphics.
执行器负责把电能变成旋转、位移和力矩。后面的 PID 并不是悬空工作的，它最终调的就是这些真实的运动响应。
右侧可以切换三种常见电动执行器，并观察命令与负载变化后它们在到位、掉速和发热上的差别。
带编码器做闭环纠偏，擅长精确到位与稳定保持。
按脉冲逐格前进，结构简单、低速有力，但负载大时可能失步。
高速高效、功率密度高，通常配合驱动器和减速器进入机器人关节。
当前 目标角度: 12°
负载越大，电机越容易出现掉速、发热或保持误差。