robotics

Robotics is the engineering discipline that builds machines able to sense and act on the physical world out of collections of sensors, actuators and chips distributed across several PCBs. A **robot** is a machine that perceives its environment, decides and acts on it with some autonomy: an arm that welds, a warehouse vehicle that dodges pallets, a drone that holds its position in the wind.

Robotics is the engineering discipline that builds machines able to sense and act on the physical world out of collections of sensors, actuators and chips distributed across several PCBs. A robot is a machine that perceives its environment, decides and acts on it with some autonomy: an arm that welds, a warehouse vehicle that dodges pallets, a drone that holds its position in the wind.

Sensor data flows inward, from a sensor to an MCU and on to an MPU, and control signals flow outward to the actuators. Each pass through that cycle is a closed loop: the action changes the world and the change comes back in through the sensors, which makes every robot a cyber-physical system (the converse fails: a power grid regulating its frequency is a CPS with no arms).

The drone is the best teaching robot because it forgives nothing: unstable without control, cheap noisy sensors, little weight to spare for computers, decisions in milliseconds. Factory robots sit at the other end, often repeating one trajectory in a closed cell with per-axis position control and a fixed program; autonomy pays only when the environment changes.

The inward half is perception (turning camera, lidar and radar data into objects and free space) and state estimation; a robot that moves through an unknown place must also work out where it is while it maps it (SLAM), or keep track of its motion from its own sensors (navigation).

The outward half is motion planning (a path that avoids obstacles and respects the vehicle's dynamics) and the feedback loops that follow it, down to the motor currents.

Learning to act from a body's own experience, where every action returns a consequence, is embodied AI; why walking and grasping proved harder for machines than chess is Moravec's paradox.

Many robots working together become a fleet, with its own questions of architecture, communication and task allocation.

Most commercial drones and mobile robots do not fly or drive with neural networks: they use cascaded PID loops and an EKF, and learning enters in the upper layers (vision, planning). The processes that make up a robot usually talk through ROS 2, while the fast loops stay on microcontrollers.