Robotic Arm

What a robotic arm is

A robotic arm, also called a robot arm or robot manipulator, is a programmable mechanical arm. It has a fixed base at one end and a tool at the other. Between them sit links (rigid segments) connected by joints (parts that bend or rotate). Motors move the joints, sensors report where each joint is, and a controller, which is a small computer, decides how far each motor should turn.

The tool on the end is called the end effector. It might be a two-finger gripper, a suction cup, a welding torch, a spray nozzle or a screwdriver. Swapping it changes what the same arm can do.

The first industrial robot arm, called Unimate, started work at a General Motors plant in 1961, moving hot metal parts. The basic idea has stayed the same since: a repeatable arm that does a physical task the same way every time.

The main parts of a robotic arm

Links and joints

Links are the stiff sections, and joints connect them. A revolute joint rotates, like an elbow or a wrist. A prismatic joint slides in a straight line, like a drawer. Most arms mix the two.

Each independent way a joint can move is called an axis, or a degree of freedom. An arm with six revolute joints has six degrees of freedom.

Actuators

Actuators are the muscles. Electric motors are the most common, usually paired with a gearbox that trades speed for strength. Some arms use compressed air or hydraulic fluid instead.

Sensors

Encoders at each joint measure the angle, so the controller always knows the arm's pose, meaning the position and angle of every part. Many arms add force sensors, which feel how hard the tool is pushing, and cameras, which help the arm find objects that are not always in the same place.

Controller

The controller reads the sensors, runs the program and sends commands to the motors many times per second.

Types of robotic arms

Arms differ mainly in how their joints are arranged. Two numbers appear on almost every spec sheet. Payload is the heaviest load the arm can carry at its tool, including the gripper itself in many specifications. Repeatability is how closely the arm returns to the same point when told to go there again, and it is usually much finer than the arm's absolute accuracy.

Articulated arms

The articulated arm is the one most people picture. Six axes give the tool full freedom to reach any position and point in any direction within its range, which is why it handles so many jobs. Play the robot arm game to get a feel for how reaching, lining up and grabbing work when you control an arm yourself.

SCARA and delta arms

SCARA arms are stiff in the vertical direction and quick in the horizontal plane, so they excel at placing parts onto a flat board. Delta robots are built for speed: their moving parts are very light, and they are used for high-speed picking on packing lines.

Cartesian and gantry systems

These slide along straight rails, like a 3D printer. They are simple, precise and easy to scale up, but cannot reach around obstacles.

Cobots

A collaborative robot, or cobot, is designed to work beside people without a safety cage in many applications. It limits its force, stops on contact and can be taught by physically guiding it.

How a robotic arm moves

Forward and inverse kinematics

Kinematics is the geometry of motion. Forward kinematics takes the angle of every joint and works out where the tool ends up. Inverse kinematics reverses it: given where you want the tool, it finds the angle each joint should take. Inverse kinematics is harder, because there can be several solutions, or none if the target is out of reach.

Closing the loop

Joint encoders feed the actual angles back to the controller, which corrects any difference from the commanded angles. A camera or force sensor extends the loop so the arm adjusts to what it sees and feels.

Why six axes

To place a tool at any point in space you need three numbers for position. To point it in any direction you need three more for orientation. That is six in total, so six independent axes is the minimum for full freedom. Simpler tasks need fewer axes, and arms with seven or more reach around obstacles more flexibly.

How robots are taught: teleoperation and learning from demonstration

The traditional way to program an arm is to write out a fixed list of positions and moves. That works for a car factory where every part arrives in the same place. It works poorly for a warehouse, where boxes vary in size and are never in the same spot twice.

Teleoperation

In teleoperation, a person controls the robot remotely and the robot moves as the person moves. The operator might use handheld controllers, a leader arm the robot mirrors, or a camera that tracks their hands. The person provides the judgment, and the robot provides the reach and strength.

Learning from demonstration

While the person drives, the system records what the robot sees through its cameras and what actions the person takes. After many examples, a machine-learning model can be trained to produce those actions on its own from what the cameras see. This approach is called learning from demonstration, or imitation learning. It lets someone teach a task by doing it, with no programming.

Why people drive the robots

People are good at physical tasks that are hard to describe in code, such as picking up a floppy bag or nudging a crooked box. Driving the robot produces examples of exactly that skill. This is why warehouse and humanoid robots are commonly trained with people operating them first.

The game is a small version of this

The game on this page follows the same loop at toy scale. You steer the robot's two arms with your hands to sort packages on a conveyor, which is teleoperation. You then buy more robots and train them by demonstration, so they sort on their own. Train your own robots in the browser game to see the idea in action.

What robotic arms are used for

• Manufacturing: welding, painting, gluing, screwing and assembling parts, especially in car and electronics plants.
• Packaging and palletizing: placing items into boxes and stacking boxes on pallets.
• Warehouses and logistics: picking items from bins, moving parcels and sorting packages onto conveyors.

Arms are most valuable when a task is repetitive, heavy or dangerous. They are least valuable when the task changes constantly, which is the problem teaching by demonstration tries to ease.

How much a robotic arm costs

The table gives rough price bands for the arm itself. They are approximate and as of 2026, and they vary by country, supplier and what is included.

The arm is rarely the whole bill. A working industrial cell also needs a gripper, safety equipment, fixtures, software and installation, and the total is often several times the price of the arm.

Build your own robotic arm

You can build a working arm at home. The usual route is an open-source design with printable parts, off-the-shelf hobby servo motors and a small controller board. The SO-101 is one popular free design among people experimenting with robot learning.

What you need

You need a 3D printer or a kit with printed parts, servo motors that report their positions, a controller board and a matching power supply. You also need patience for calibration, which means telling the software what each joint position means.

A sensible path

Learn to read and command each joint first, then move the tool to a point using inverse kinematics, then record a few demonstrations and replay them. If you want to feel the control side before buying parts, use the free browser game above to control a robot arm with your hands.

Frequently asked questions