Physical AI & Robotics Visuals
How robots sense, move, and learn, from feedback loops and coordinate frames to motion planning and learned policies, taught through numbered SVG sequences
The Sense–Think–Act Loop
How robots observe changing environments, plan actions, and correct for delays and disturbances in closed loops.
See it visuallyRobot Anatomy and Degrees of Freedom
Distinguish joint space from task space, calculate mobility with Grübler's formula, and map physical workspace limits.
See it visuallyActuators, Gearing and Torque Limits
Follow motor current through a gearbox to joint torque, then check speed, heating, backlash and whether a reachable load can actually be held.
See it visuallyHow Robots Sense
Trace encoder counts, wheel slip, external range and a weighted position update, then examine inertial bias, stereo depth and missing echoes.
See it visuallyCoordinate Frames and Transforms
Convert coordinates across sensor and robot frames using homogeneous transformation matrices in SE(2) and SE(3).
See it visuallyForward and Inverse Kinematics
Predict end-effector positions with forward kinematics and solve multi-branch joint angles with analytical inverse kinematics.
See it visuallyJacobians and Singularities
Turn joint rates into tip velocity, invert a checked two-link example, and see why straightening an arm removes a direction of motion.
See it visuallyFeedback Control and PID
Trace error, proportional torque, retained integral effort and rate damping, then check saturation, conditional integration and derivative noise.
See it visuallyDynamics and Impedance Control
Connect force to acceleration, account for gravity, and build a virtual spring and damper that yields under contact before actuator saturation breaks the model.
See it visuallyState Estimation and the Kalman Filter
Follow a scalar position estimate through prediction, noisy observation and a weighted update, then see why smaller reported variance can still be wrong.
See it visuallySLAM: Mapping While Localizing
Build a map and a robot trajectory together from noisy motion and landmark observations, then trace loop closure and a false match that deforms the estimate.
See it visuallyMotion Planning
Account for robot size, search a checked grid with A*, extend an RRT and time a path, then test smoothing collisions and stopping clearance.
See it visuallyModel Predictive Control and Legged Locomotion
Follow a quadruped's contact schedule, predict motion from ground forces, score a horizon and replan, then see how missing grip breaks the prediction.
See it visuallyRobot Learning: Imitation to RL
Follow a block-pushing policy from demonstrations and behavior cloning to DAgger, reward optimization, simulation, and the limits of transfer.
See it visuallyVision-Language-Action Models
Follow an instruction and camera observation into robot actions, then compare action bins, continuous generation, chunk execution, and feedback limits.
See it visually