Joints and Kinematics

Hold your arm out straight in front of you. Now bend your elbow and rotate your wrist. Your hand just moved through space to a new position — yet you did not think about the geometry at all. Your nervous system handled the math automatically. A robot arm has no nervous system, so its controller must compute that geometry explicitly every single time it wants to place the tip somewhere. That computation is called kinematics, and it is the core of how any multi-joint robot arm works.

Links and Joints: The Building Blocks

A robot arm is built from alternating links and joints. A link is a rigid segment — a piece of metal or carbon fiber that does not bend. A joint connects two links and allows them to move relative to each other. The type of joint determines what kind of motion is allowed. The two most common joint types in robot arms are revolute joints and prismatic joints. A revolute joint rotates — like your elbow, it swings one link relative to the other around a fixed axis. A prismatic joint slides — it extends or retracts along a straight line, like a piston. Most industrial robot arms use revolute joints almost exclusively because rotation is easier to power with a motor and compact to package.

Forward Kinematics: From Angles to Position

Forward kinematics answers this question: if I know the angle of every joint in the arm, where is the tip of the arm in space? It is called forward because it moves in the direction of cause to effect — joint angles cause the tip to end up somewhere, and we compute where. Imagine a simple two-joint arm lying flat on a table. The first link is 30 cm long and its joint is at the base. The second link is 20 cm long. If the first joint is set to 60 degrees and the second joint is set to 45 degrees, basic trigonometry can tell you the (x, y) coordinates of the tip. With more joints and three dimensions, the math gets more involved, but the idea is identical. Forward kinematics always has exactly one answer: given these joint angles, the tip is at this position. There is no ambiguity.

Inverse Kinematics: From Position to Angles

Inverse kinematics answers the opposite question: I want the tip to be at this point in space at this orientation — what angles must all the joints be set to? It is the problem the controller actually needs to solve. You tell the robot to pick up an object at a known location; the controller runs inverse kinematics to figure out where each joint must go. Inverse kinematics is much harder than forward kinematics for two reasons. First, there may be multiple valid solutions. If your arm has more than three joints, there are usually several different combinations of joint angles that all place the tip at the same position. The robot must pick one — typically the one closest to the current configuration, or the one that avoids hitting an obstacle. Second, for some target positions there is no solution at all: the target is out of reach or requires a configuration the joints cannot achieve. Engineers use numerical solvers — iterative algorithms that make a guess, measure the error, and refine the guess — to solve inverse kinematics in real time for complex arms.

Why This Matters in Real Robots

Every time a factory robot welds a car frame, picks a package off a conveyor, or performs surgery, inverse kinematics is running in the background — typically hundreds of times per second to keep the tip following a smooth path even as the joints move. The solver must be fast enough to keep up with the desired motion and robust enough to handle the multiple-solution problem gracefully. Some robots sidestep the mathematics with a technique called motion capture: a human demonstrates the desired path by moving the arm directly while it records joint angles. The robot then replays those recorded angles. This works well for repetitive tasks but cannot adapt to a changed environment without re-recording.